WO2023250013A1 - Non-contact systems and methods to estimate pavement friction or type - Google Patents

Abstract

System and method employ (i) pavement texture data comprising direct laser distance measurement data acquired through non-contact sensors and (ii) machine learning models to estimate friction/skid resistance as well as pavement type. The determined friction/skid resistance estimates may be aggregated to generate a friction or skid resistance map for the roadway or associated roadway network, the friction/skid resistance map is then used to determine a maintenance or repair schedule or event for the roadway or associated roadway network. System and method may employ a second non-contact measurement (e.g., high-speed camera) to provide a classification of the pavement type, e.g., to be used in the friction/skid resistance estimation. The determined pavement type may be used to generate a pavement-type map for the roadway or associated roadway network, the pavement type map can also be used alone or in combination with the friction or skid resistance map.

Description

Non-contact Systems and Methods to Estimate Pavement Friction or Type

RELATED APPLICATION

[0001] This PCT international application claims priority to, and the benefit of, U.S. Provisional Patent Application No. 63/353,981, filed June 21, 2022, entitled “Non-contact Systems and Methods to Estimate Pavement Friction,” which is incorporated by reference herein in its entirety.

BACKGROUND

[0002] The tire pavement interaction is essential for the safety of motorists. Pavement management systems are in place through various transportation agencies and other responsible agencies in the United States and most countries to assess the friction of roadways and allocate resources for improving pavement friction. These agencies generally conduct evaluations and monitoring of the friction condition of their roadways on an annual basis for high-trafficked roads.

[0003] The most common equipment for friction evaluation employs a trailer having a skid device comprising a test wheel configured with a fixed or variable slip braking system to continuously contact and measure, at highway travel speeds, frictional properties (i.e., skid resistance) of a road surface. The measure of skid resistance is expressed in a skid number or other equivalent properties. The skid device typically includes a smooth or ribbed reference tire installed on the trailer that includes a water distribution system connected to a water tank installed on a front moving vehicle. The trailer is then towed behind the front vehicle at highway speed, e.g., 80 km/h (50 mph). Once the vehicle and trailer reach the intended test speed, the water system sprays test water in front of the test wheel to mimic wet weather conditions while the braking system is engaged; the water also reduces the wear on the test wheel.

[0004] There is high maintenance on the contact measuring component of the skid device because of the wear from this contact-based measurement. The measurement is also limited in the distance that can be measured due to the limited capacity to carry the consumables (i.e., water) used during the test.

[0005] There is also a lack of real-time data on pavement types in a given roadway system. Routine maintenance may be performed for a stretch of the roadway employing different types of pavements or repair material. [0006] There is a benefit to improving the current testing of pavement friction and type.

SUMMARY

[0007] An exemplary system and method are disclosed that employ (i) pavement texture data comprising direct laser distance measurement data acquired through non-contact sensors and (ii) machine learning models to estimate friction/skid resistance as well as pavement type. The determined friction/skid resistance estimates can be aggregated to generate a friction or skid resistance map for the roadway or associated roadway network, the map can be used to determine a maintenance or repair schedule or event for the roadway or associated roadway network. Examples of non-contact sensors include line laser scanners, stereographic cameras, line-scan lasers, stereo imaging, photogrammetry, acoustic sensors, or other instrumented sensors that can detect texture wavelengths up to 50 mm. In some embodiments, the non-contact sensors can acquire macrotexture wavelengths (50 - 0.5 mm) and/or a portion of microtexture wavelengths (less than 0.5 mm) at highway speeds (e.g., at least 80 km/h (50 mph)). Though the sensors can acquire data at highway speeds of 80 km/h (50 mph), and higher, it should be appreciated that data acquired up to the rated instrumentation speed, which could be between 0.001 mph and 70 mph, would provide an accurate estimate of friction/skid resistance.

[0008] The exemplary system and method may employ a second non-contact measurement (e.g., high-speed camera) to provide the classification of the pavement type, e.g., to be used in the friction/skid resistance estimation. In some embodiments, the pavement type can be used to generate a pavement-type map for the roadway or associated roadway network, the map can also be used alone or in combination with the friction or skid resistance map to determine a maintenance or repair schedule or event for the roadway or associated roadway network.

[0009] The term “friction/skid resistance,” as used herein, refers to a parameter that provides a characterization of roadway friction, e.g., coefficient of friction or some measure of friction resistance and/or skid resistance. Skid resistance is the force developed when a tire that is prevented from rotating slides along the pavement surface. In some embodiments, the skid resistance is the skid number (SN), determined by performing pavement friction testing using the locked-wheel method in which the skid number is equal to the drag force required to slide a locked test tire at a given speed, divided by the effective wheel load and multiplied by 100. [0010] The basic relationship between the forces acting on the vehicle tire and the pavement surface as the vehicle steers around a curve, changes lanes, or compensates for lateral forces follows the equation, F_s = 7²/l 57? — e, where F_s is the side friction, V is the vehicle speed in mph, R is the radius of the path of the vehicle’s center of gravity in ft., and e is the pavement super-elevation in ft./ft.

[0011] The non-contact sensors can be deployed in any service vehicle. In some embodiments, the non-contact sensors can be deployed in commercial vehicles to provide realtime monitoring of roadway pavement conditions.

[0012] By not using a test wheel that directly contacts the pavement surface or using water during the test acquisition, the exemplary system and method can be employed at a reduced cost and with reduced wear. In addition, the exemplary system and method can operate at an extended range (as compared to prior contact-based measurement systems) utilizing non-contact texture profile data. The extended range can be used to increase the network coverage to provide a more frequent and/or more comprehensive evaluation of the network of roadways.

[0013] In an aspect, a method is disclosed to determine friction or skid resistance of a roadway, the method comprising: obtaining, by a processor, non-contact measurement data having pavement macrotexture and microtexture of the roadway, wherein the non-contact measurement data is continuously acquired via one or more non-contact sensors (e.g., in-line laser) by a vehicle housing the non-contact sensor (e.g., at a speed of at least 30 miles per hour or static vehicle); determining, by the processor, using the non-contact measurement data, one or more pavement associated parameters, or associated values, selected from the group consisting of pavement amplitude parameters, pavement statistic parameters, hybrid pavement parameters, and/or pavement spectral parameters; determining, by the processor, a pavement type via at least one of (i) a first classifier using the one or more pavement associated parameters or associated values or (ii) a second classifier configured to determine pavement type based on a second noncontact measurement data (e.g., acquired via high-speed camera); and determining, by the processor, a value for the friction or skid resistance of the roadway using a model (e.g., a trained machine learning model and regression model) defined for the pavement type using at least one of outputs of the first classifier or outputs of the second classifier, wherein the value for the friction or skid resistance is employed to (i) determine a maintenance or repair schedule or event for the roadway or associated roadway network or (ii) generate a friction or skid resistance map for the roadway or associated roadway network. [0014] In another aspect, a method (e.g., for a self-driving vehicle) is disclosed to determine friction or skid resistance of a roadway, the method comprising: obtaining, by a processor, noncontact measurement data having pavement macrotexture and microtexture of the roadway, wherein the non-contact measurement data is continuously acquired via one or more non-contact sensors by a vehicle (e.g., the self-driving vehicle) housing the non-contact sensor measuring the roadway (e.g., at a speed of at least 30 miles per hour); and transmitting, by the processor, the non-contact measurement data to an analysis system, wherein the analysis system is configured to: determine using the non-contact measurement data, one or more pavement associated parameters, or associated values, selected from the group consisting of pavement amplitude parameters, pavement statistic parameters, hybrid pavement parameters, and/or pavement spectral parameters; determine a pavement type via at least one of (i) a first classifier using the one or more pavement associated parameters or associated values or (ii) a second classifier configured to determine pavement type based on a second non-contact measurement data; and determine a value for the friction or skid resistance of the roadway using a model (e.g., a trained machine learning model and regression model) defined for the pavement type using at least one of outputs of the first classifier or outputs of the second classifier, wherein the value for the friction or skid resistance is employed to (i) determine a maintenance or repair schedule or event for the roadway or associated roadway network or (ii) generate a friction or skid resistance map for the roadway or associated roadway network.

[0015] In some embodiments, the method further includes receiving, by the processor, friction or skid resistance map for the roadway or associated roadway network from an analysis system; and updating, by the processor, a control operation of a vehicle system (e.g., braking system) using the received friction or skid resistance map.

[0016] In some embodiments, the method further includes receiving, by the processor, friction or skid resistance map for the roadway or associated roadway network from an analysis system; and generating, by the processor, at a display of the vehicle, a notification or indicator to a user of the vehicle associated with the received friction or skid resistance map.

[0017] In some embodiments, the non-contact measurement data is acquired from a line-laser scanner. [0018] In some embodiments, the one or more pavement-associated parameters or associated values include a kurtosis parameter that indicates a presence of extremely high peaks or deep valleys.

[0019] In some embodiments, the one or more pavement-associated parameters or associated values include a profile solidity factor parameter that is a ratio between a maximum depth of identified valleys and a maximum height of an acquired 2D scan associated with a non-contact sensor of the non-contact sensors.

[0020] In some embodiments, the one or more pavement-associated parameters or associated values include a mean cross-width parameter that is a measure of an average distance between points where an acquired 2D scan associated with a non-contact sensor of the non-contact sensors crosses from above to below a baseline horizontal plane at a zero elevation.

[0021] In some embodiments, the one or more pavement-associated parameters or associated values include a cross-width variance that measures a variance of a distance between points where an acquired 2D scan associated with a non-contact optical sensor of the non-contact optical sensors crosses a determined mean of the acquired 2D scan.

[0022] In some embodiments, the one or more pavement-associated parameters or associated values include a 2-Pt slope variance measure of a slope between two consecutive points.

[0023] In some embodiments, the one or more pavement-associated parameters or associated values include at least one of: (i) a kurtosis parameter that a presence of extremely high peaks or deep valleys, (ii) a profile solidity factor parameter that is a ratio between a maximum depth of identified valleys and a maximum height of an acquired 2D scan associated with a non-contact sensor of the non-contact sensors, (iii) a mean cross width parameter that is a measure of an average distance between points where the acquired 2D scan, (iv) a cross width variance that measures a variance of a distance between points where the acquired 2D scan crosses a determined mean of the acquired 2D scan, (v) a 2-Pt slope variance measure of a slope between two consecutive points, or (v) a combination thereof.

[0024] In some embodiments, the value for the friction or skid resistance of the roadway is determined using a trained machine learning model, as the first classifier, comprising a decision tree classifier (e.g., and regression model).

[0025] In some embodiments, the trained machine learning model was trained using data selected from the group consisting of: chip seals with high macrotexture (e.g., HM-CS); dense fine mixes (e.g., DFM); chip seals with low macrotexture (e.g., LM-CS); open mixes or PFCs (e.g., OM); dense coarse mixes (e.g., DCM); stone matrix asphalt (e.g., SMA); finish-graded concrete; or a combination thereof.

[0026] In some embodiments, the method further includes transmitting, by the processor, the determined value for the friction or skid resistance and a corresponding positioning data to a global analysis system, wherein the global analysis system is configured to aggregate the determined value for the friction or skid resistance for the corresponding positioning data along with determined values for the friction or skid resistance of other positioning data to generate a friction or skid resistance map for a given geographic area.

[0027] In some embodiments, the method further includes transmitting, by the processor, the determined value for the friction or skid resistance and a corresponding positioning data to a global analysis system, wherein the global analysis system is configured to aggregate the determined value for the friction or skid resistance for the corresponding positioning data along with determined values for the friction or skid resistance of other positioning data to generate a road risk map for a given geographic area.

[0028] In some embodiments, the friction or skid resistance map or the road risk map generated for the given geographic area are subsequently transmitted, in whole or in part, to vehicles traveling through the given geographic area.

[0029] In some embodiments, the global analysis system is a cloud-based system.

[0030] In some embodiments, the step of determining pavement type based on the second non-contact measurement data comprises obtaining, by a processor, second non-contact measurement data, wherein the second non-contact measurement data is continuously acquired via one or more second non-contact sensors (e.g., high-speed camera) by the vehicle housing the one or more second non-contact sensors; and determining, by the processor, using the second non-contact measurement data, a pavement type via the second classifier, wherein the second classifier was trained for different pavement types.

[0031] In some embodiments, both (i) the value for the friction or skid resistance and (ii) the determined pavement type is employed to (iii) determine the maintenance or repair schedule or event for the roadway or associated roadway network and/or (iv) generate a pavement type map for the roadway or associated roadway network. [0032] In some embodiments, the pavement-type map includes indicators of patched roadway repairs.

[0033] In some embodiments, the second non-contact measurement data is acquired from one or more high-speed cameras.

[0034] In another aspect, a method is disclosed to determine a pavement type of a roadway, the method comprising: obtaining, by a processor, non-contact measurement data having a set of images of the pavement, wherein the measurement data is continuously acquired at high speed via one or more non-contact high-speed sensors by a vehicle housing the non-contact high-speed sensor; determining, by the processor, a pavement type via a classifier using the non-contact measurement data, wherein the determined pavement type is employed to (i) determine a maintenance or repair schedule or event for the roadway or associated roadway network, (ii) generate a pavement type map for the roadway or associated roadway network and/or (iii) perform subsequent analysis to determine a friction or skid resistance map for the roadway or associated roadway network.

[0035] In another aspect, a system is disclosed comprising a processor; and a memory having instructions stored thereon, wherein execution of the instructions by the processor, causes the processor to perform any of the above-discussed methods.

[0036] In another aspect, a non-transitory computer-readable medium is disclosed having instructions stored thereon, wherein execution of the instructions by the processor causes the processor to perform any of the above-discussed methods.

[0037] In another aspect, a system is disclosed comprising: a non-contact sensor (e.g., laser scanner) configured to acquire non-contact measurement data having pavement macrotexture and microtexture of a pavement surface; and an analysis system comprising a processor and a memory having instructions stored thereon, wherein execution of the instructions by the processor causes the processor to: determine using the non-contact measurement data, one or more pavement associated parameters, or associated values, selected from the group consisting of pavement amplitude parameters, pavement statistic parameters, hybrid pavement parameters, and/or pavement spectral parameters; determine a pavement type via at least one of (i) a first classifier using the one or more pavement associated parameters or associated values or (ii) a second classifier configured to determine pavement type based on a second non-contact measurement data; and determine a value for the friction or skid resistance of the roadway using a model (e.g., a trained machine learning model) defined for the pavement type, wherein the value for the friction or skid resistance is employed to (i) determine a maintenance or repair schedule or event for the roadway or associated roadway network or (ii) generate a friction or skid resistance map for the roadway or associated roadway network.

[0038] In another aspect, a system is disclosed comprising: a non-contact sensor (e.g., laser scanner) configured to acquire non-contact measurement data having pavement macrotexture and microtexture of a pavement surface; and network interface configured to transmit the non-contact measurement data to an analysis system, wherein the analysis system comprises a processor and a memory having instructions stored thereon, wherein execution of the instructions by the processor causes the processor to: determine using the non-contact measurement data, one or more pavement associated parameters, or associated values, selected from the group consisting of pavement amplitude parameters, pavement statistic parameters, hybrid pavement parameters, and/or pavement spectral parameters; determine a pavement type via at least one of (i) a first classifier using the one or more pavement associated parameters or associated values or (ii) a second classifier configured to determine pavement type based on a second non-contact measurement data; and determine a value for the friction or skid resistance of the roadway using a model (e.g., a trained machine learning model) defined for the pavement type using at least one of outputs of the first classifier or outputs of the second classifier, wherein the value for the friction or skid resistance is employed to (i) determine a maintenance or repair schedule or event for the roadway or associated roadway network or (ii) generate a friction or skid resistance map for the roadway or associated roadway network.

[0039] In another aspect, a system is disclosed comprising: a non-contact sensor configured to acquire non-contact measurement data having pavement macrotexture of a pavement surface; and an analysis system comprising a processor and a memory having instructions stored thereon, wherein execution of the instructions by the processor causes the processor to: obtain the noncontact measurement data having a set of images of pavement macrotexture of the roadway, wherein the non-contact measurement data is continuously acquired at high speed via one or more non-contact high-speed sensors by a vehicle housing the non-contact high-speed sensor; determine a pavement type via a classifier using the non-contact measurement data, wherein the determined pavement type is employed to (i) determine a maintenance or repair schedule or event for the roadway or associated roadway network, (ii) generate a pavement type map for the roadway or associated roadway network and/or (iii) perform subsequent analysis to determine a friction or skid resistance map for the roadway or associated roadway network.

[0040] In some embodiments, the non-contact sensor is mounted to a trailer.

[0041] In some embodiments, the non-contact sensor is mounted to a vehicle.

[0042] In some embodiments, the non-contact sensor is a line-laser scanner.

[0043] In some embodiments, the non-contact sensor is a high-speed camera.

[0044] In another aspect, any one of the above-discussed systems is configured with a vehicle controller, the vehicle having a processor and memory, the memory having instructions stored thereon, wherein execution of the instructions by the processor causes the processor to: receive a friction or skid resistance map for the roadway or associated roadway network from an analysis system; and update a control operation of a vehicle system (e.g., braking system) using the received friction or skid resistance map.

[0045] In some embodiments, execution of the instructions by the processor further causes the processor to: receive a friction or skid resistance map for the roadway or associated roadway network from an analysis system; and generate, at a display of the vehicle, a notification or indicator to a user of the vehicle associated with the received friction or skid resistance map. [0046] In some embodiments, the analysis system further includes a network interface, the network interface being configured to transmit the determined value for the friction or skid resistance and a corresponding positioning data to a global analysis system, wherein the global analysis system is configured to aggregate the determined value for the friction or skid resistance for the corresponding positioning data along with determined values for the friction or skid resistance of other positioning data to generate a friction or skid resistance map for a given geographic area.

BRIEF DESCRIPTION OF THE DRAWINGS

[0047] Embodiments of the present invention may be better understood from the following detailed description when read in conjunction with the accompanying drawings. Such embodiments, which are for illustrative purposes only, depict novel and non-obvious aspects of the invention. The drawings include the following figures:

[0048] Figs. 1A and IB each shows an example system configured to determine friction or skid resistance of a roadway in accordance with several illustrative embodiments. [0049] Fig. 1C shows an example system configured to determine the pavement type of a roadway in accordance with an illustrative embodiment.

[0050] Figs. 2A-2D each shows aspects of data processing operations to remove noise from an acquired signal data to be used in predicting or estimating friction or skid resistance in accordance with an illustrative embodiment.

[0051] Fig. 3 shows an example set of features that can be used in a machine learning model to classify pavement type in accordance with an illustrative embodiment.

[0052] Figs. 4A and 4B show an example machine-learning operation to generate a trained machine-learning model and regression model and associated results in accordance with an illustrative embodiment.

[0053] Figs. 5A-5E each shows aspects of a study conducted to develop machine learning models to predict or estimate friction or skid resistance in accordance with an illustrative embodiment.

[0054] Figs. 6A-6H show a validation study described in relation to validate the developed machine learning models of Figs. 5A - 5E in accordance with an illustrative embodiment.

[0055] Figs. 7A-7F show a study conducted to develop machine learning models to predict or estimate pavement type using high-speed camera images in accordance with an illustrative embodiment.

DETAILED SPECIFICATION

[0056] Some references, which may include various patents, patent applications, and publications, are cited in a reference list and discussed in the disclosure provided herein. The citation and/or discussion of such references is provided merely to clarify the description of the present disclosure and is not an admission that any such reference is “prior art” to any aspects of the present disclosure described herein. In terms of notation, “[n]” corresponds to the n^th reference in the list. All references cited and discussed in this specification are incorporated herein by reference in their entireties and to the same extent as if each reference was individually incorporated by reference.

[0057] Some references, which may include various patents, patent applications, and publications, are cited in a reference list and discussed in the disclosure provided herein. The citation and/or discussion of such references is provided merely to clarify the description of the disclosed technology and is not an admission that any such reference is “prior art” to any aspects of the disclosed technology described herein. In terms of notation, “[n]” corresponds to the nth reference in the list.

[0058] Example System #1

[0059] Fig. 1A shows an example system 100 (shown as 100a) configured to determine friction or skid resistance of a roadway 102 in accordance with an illustrative embodiment. The system 100a includes a skid device 104 configured with an instrumentation assembly 106 (shown as 106a) to measure the macrotexture and microtexture of the pavement 108. In the example shown in Fig. 1A, the skid device 104 is configured as a trailer that is attached to a vehicle 110.

[0060] In the example shown in Fig. 1 A, the instrumentation assembly 106a includes a noncontact sensor 112 (shown as “line laser” 112), a sensor interface or controller 114 (shown as “controller” 114), energy storage or converter 116, and a data store 118. The non-contact sensor 112, in some embodiments, is configured to generate a laser beam 111 (e.g., a 2D laser beam). In other embodiments, the non-contact sensor 112 is configured to generate a non-contact beam or signals as described herein to acquire the macrotexture and microtexture of the pavement 108, e.g., at macrotexture wavelengths (50 - 0.5 mm) and/or a portion of microtexture wavelengths (less than 0.5 mm) at highway speeds (e.g., at least 80 km/h (50 mph)). In the example shown in Fig. 1A, the laser beam is oriented in the transversal x-axis 113. In other embodiments, the noncontact sensor 112 is configured to generate a laser beam 111 that is diagonal, i.e., having both longitudinal 115 and transversal axes 113. In addition, while the example of Fig. 1 A shows the non-contact sensor 112 is configured to generate the laser beam 111 in the direction of travel, in other embodiments, the non-contact sensor 112 can be configured to generate the non-contact sensor beam 111 in the opposite direction of travel, in different orientations of the laser light, or in other directions to the direction of travel. In some embodiments, multiple laser lines can be used in the same alignment with one another. In other embodiments, the multiple laser lines can be used in the different alignments. The laser can be in the visible range or an infrared range.

[0061] In some embodiments, the sensor interface or controller 114 is configured (as a sensor controller or as a sensor interface) to connect to a mobile computer or laptop that is configured to function as the data store 118. In other embodiments, the controller 114 is configured to connect to integrated or ruggedized data acquisition hardware (not shown). In some embodiments, the sensor interface or controller 114 includes optoelectric isolation components and integrating circuitries to facilitate connection to a computer. The data store 118 may provide the acquired macrotexture and microtexture data 120 to an analysis system 122 configured to generate a friction or skid resistance value. In some embodiments, the friction or skid resistance is a skid number. The analysis system 122 may be a part of the instrumentation assembly 106a. In other embodiments, the analysis system 122 may be remote to the instrumentation assembly 106a, e.g., a server or a cloud infrastructure that is configured to receive the acquired macrotexture and microtexture data 120 to analyze for the friction or skid resistance value.

[0062] In Fig. 1 A, the analysis system 122 is configured to perform data processing operation 124, a pavement classification operation 126, and a pavement friction determination operation 128 (shown as “Pavement friction models” 128) to generate pavement friction output 130. The pavement friction output 130 can be used to generate a roadway pavement friction map 132 (shown as “Road Friction Map” 132), to which maintenance and repair schedules 135 for the roadways and other roadways in the network can be performed. As noted above, the analysis system 122 may be a part of the instrumentation assembly 106a or may be a server or a cloud infrastructure that is located remote to the instrumentation assembly 106a.

[0063] Example System #2

[0064] Fig. IB shows an example system 100 (shown as 100b) configured to determine friction or skid resistance of a roadway 102 in accordance with an illustrative embodiment. The system 100b includes the instrumentation assembly 106 (shown as 106b) integrated into a vehicle 110 (shown as 110b’ and 110b”) to measure the macrotexture and microtexture of the pavement 108.

[0065] In the example shown in Fig. IB, the instrumentation assembly 106b includes a noncontact sensor 112 (shown as “line laser” 112), a sensor interface or controller 114, energy storage or converter 116, and a data store 118. The non-contact sensor 112 is configured to generate a non-contact sensor beam 111 (e.g., a 2D laser beam) to acquire the macrotexture and microtexture of the pavement 108, e.g., at macrotexture wavelengths (50 - 0.5 mm) and/or a portion of microtexture wavelengths (less than 0.5 mm) at highway speeds (e.g., at least 80 km/h (50 mph)).

[0066] In the example shown in Fig. IB, instrumentation assembly 106b is integrated with the vehicle to include a network interface 137 that is configured to communicate through a network 139 to the analysis system 122 (shown as 122b) (also referred to herein as a global analysis system). The analysis system 122b may be a cloud infrastructure that is configured to perform data processing operation 124, a pavement classification operation 126, and a pavement friction determination operation 128 to generate a pavement friction output 130, as described in relation to Fig. 1A.

[0067] The analysis system 122b may obtain the friction data 130 (shown as 130’) from multiple vehicles and further aggregate, via an aggregation operation 132, the pavement friction data 130 from multiple vehicles for a given geographic location to generate a friction or skid resistance map 134 for the given geographic area. The analysis system 122b may provide the friction or skid resistance map 134 (shown as 134 and 134’) to a vehicle control or map system (shown as “Vehicle Control” 136) that can be accessed by vehicles (shown as 110b) to receive friction or skid resistance for the roadways in the area.

[0068] The vehicle control 136 of the vehicle 110b may use the friction or skid resistance data in combination with the vehicle’s position information 138 (shown as “MPS/GPS Data” 138) to provide an indication or alert to the driver of low friction areas of the roadway. In some embodiments, the vehicle control 136 may further provide input to the cruise control operation or autonomous self-driving or semi-self-driving technology modules present in the vehicle 110b to adjust the speed of the vehicle 110b to a safe speed for the given friction or skid resistance. In some embodiments, the analysis system 122b may transmit the friction or skid resistance or maps to client applications running in vehicles for the given area to be presented to the user of the vehicle. In some embodiments, the friction or skid resistance map 134 may be integrated with navigation maps used by the vehicle through an onboard GPS map system or by the user through a map or navigation application, e.g., executing on a smartphone.

[0069] Indeed, in addition to being used for maintenance and repair schedules for the roadways and other roadways in the network, the exemplary system and methods may be performed to provide real-time control of vehicles based on pavement friction data.

[0070] Example System #3

[0071] Fig. 1C shows an example system 100 (shown as 100c) configured to determine the pavement type of a roadway 102 in accordance with an illustrative embodiment. The system 100c includes an instrumentation assembly 106 (shown as 106b) to measure the macrotexture and microtexture of the pavement 108. In the example shown in Fig. 1C, the instrumentation assembly 106b is fixably attached to the vehicle 110. In other examples, the instrumentation assembly 106b may be mounted to a trailer, e.g., similar to that shown in Fig. 1 A. In yet other embodiments, the instrumentation assembly 106b (pavement classification) may be mounted to the trailer 104 of Fig. 1A to operate in connection with the instrumentation assembly 106 (friction or skid resistance classification).

[0072] In the example shown in Fig. 1C, the instrumentation assembly 106b includes a noncontact high-speed sensor 117 (shown as “high-speed camera” 117), a sensor interface or controller 114 (shown as “controller” 114b), energy storage or converter 116, and a data store 118. The non-contact high-speed sensor 117, in some embodiments, is configured to generate a set of high-speed images or video frames of the pavement 108 at highway speeds (e.g., at least 80 km/h (50 mph)). While the example of Fig. 1C shows the non-contact sensor 117 configured to generate a high-speed image in the direction opposite of travel; in other embodiments, the non-contact sensor 117 can be configured to generate the high-speed image in the direction of travel (e.g., the sensor being disposed at the front of the vehicle). The instrumentation assembly 106b may include light projection system 119 (shown as “Illumination System” 119) to illuminate the portion 108a of the roadway 108 being captured by the sensor 117.

[0073] In some embodiments, the sensor interface or controller 114b is configured (as a sensor controller or as a sensor interface) to connect to a mobile computer or laptop that is configured to function as the data store 118. In other embodiments, the controller 114 is configured to connect to integrated or ruggedized data acquisition hardware (not shown). The data store 118 may provide the acquired data 120a to an analysis system 122 (shown as 122b) configured to generate a value, score, or label for a pavement type. The analysis system 122b may be a part of the instrumentation assembly 106b. In other embodiments, the analysis system 122b may be remote to the instrumentation assembly 106b, e.g., a server or a cloud infrastructure that is configured to receive the acquired data 120a to analyze for the friction or skid resistance value.

[0074] In Fig. 1C, the analysis system 122b is configured to perform data processing operation 124a and a pavement classification operation 126 (shown as 126a).

[0075] The pavement type output 131 can be used to generate a roadway pavement type map 133 (shown as “Road Type Map” 133), to which maintenance and repair schedules 135 (shown as 135a) for the roadways and other roadways in the network can be performed. The pavement type output 131 may be combined with models for friction or skid resistance analysis (shown as “Pavement Friction Analysis”). In one example, the pavement type output 131 may be employed in a decision tree model, while the friction or skid resistance values may be derived from equations specific to the pavement types that employs the non-contact measurement data or the one or more determined pavement associated parameters, or associated values, of those described in relation to Fig. 1A.

[0076] As noted above, the analysis system 122b may be a part of the instrumentation assembly 106b or may be a server or a cloud infrastructure that is located remote to the instrumentation assembly 106b.

[0077] Example Data Processing

[0078] Fig. 2A shows an example 2D laser beam (e.g., 106) that can acquire the macrotexture and microtexture of the pavement (e.g., 108), e.g., at macrotexture wavelengths (50 - 0.5 mm) and/or a portion of microtexture wavelengths (less than 0.5 mm) at highway speeds (e.g., at least 80 km/h (50 mph)). In some embodiments, the laser sensor is a 660 nm, class 3B laser configured to output >130 mW. The laser sensor can be configured to provide a resolution of <5 pm and a lateral resolution of 0.1 - 0.3 mm. In some embodiments, the laser-sensor can measure a pre-defined width (e.g., betweenl80 mm and 1000 mm) in the transverse direction to capture the friction equipment’s testing tire’s contact area with the pavement (>50 mm [>2 in.] wide). The width, e.g., can be about 180 mm, 190 mm, 200 mm, 210 mm, 220 mm, 230 mm,

240 mm, 250 mm, 260 mm, 270 mm, 280 mm, 290 mm, 300 mm, 310 mm, 320 mm 330 mm,

340 mm, 350 mm, 360 mm, 370 mm, 380 mm, 390 mm, 400 mm, 410 mm, 420 mm, 430 mm,

440 mm, 450 mm, 460 mm, 470 mm, 480 mm, 490 mm, 500 mm. In some embodiments, the width can be greater than 500 mm.

[0079] The line laser (e.g., 112) and other non-contact sensors described herein can be configured, in alternative embodiments, to generate the laser beam in the opposite direction of travel, in different orientations of the laser light, or in other directions to the direction of travel. In some embodiments, multiple laser lines can be used in the same alignment with one another. In other embodiments, the multiple laser lines can be used in the different alignments.

[0080] In one example, when the testing tires are inflated to their specified pressure (e.g., 20 psi) and for a specified diameter, the laser-sensor can be configured with a pre-defined profile spacing, e.g., of approximately 40 mm. [0081] In the example shown in Fig. 2A, the laser sensor 106 is configured to generate a 2D laser beam 111 that can produce a pre-defined number of points of data (e.g., between 800 and 4000 points) for each instance of the measurement.

[0082] Fig. 2B shows an example processing 200 of the acquired signal from the instrumentation assembly (e.g., 106 ). The processing 200 may be implemented by a number of modules, including an invalid point removal module 202, noise detection and removal 204, data imputation module 206, profile detrending module 208, and data transform 210.

[0083] The modules may operate solely or in combination to remove spikes shown in plots 212, saturations and dropout (shown in plot 214), e.g., due to dark or darker pavement surfaces, and white noise (shown in plot 216).

[0084] Invalid Point Removal module (202). Module 202 is configured to remove dead pixels (invalid data from dropouts) from the actual measurements. Dropouts as invalid readings can appear at the edges of the profile as artifacts generated by a width correction algorithm in the sensor to keep the distance between points at a constant interval due to the sensor’s elevation changes. The camera may capture less information across the x-axis the closer the sensor is to the scanning surface; thus, dropouts can be removed.

[0085] Fig. 2A shows that on the near side (220) (i.e., the height where the sensor is closest to the surface), the number of dropouts (218) at the edges of the profile is generated but decreases as the laser sensor is moved closer to the far side (222). When collecting data, the sensor must be placed at the reference distance (224), so the measurement can oscillate between the near side 220 and far side 222 as it collects data. Fig. 2A, plot 224, shows the dropout values 226 as negative elevations with a high magnitude. The valid measurements typically occur in the middle section of the profile.

[0086] Noise Detection and Removal module (204). Module 204 is configured to detect and remove noise from laser 2D profiles, including white noise, spikes, and flat signal noise, e.g., that result from thermal noise, electrical noise, electromagnetic noise associated with the physics, sensors, data acquisition, or transmission. Module 204 can remove such noise, e.g., to prevent outliers from skewing or biasing the texture statistics. For the application of pavement surface scanning, often, field conditions are not optimal when scanning a pavement surface, which can lead to the introduction of noise within the signal. The most common instances of noise found in a 2D laser profile may include white noise, spikes, and flat signals. [0087] White noise is a random signal having equal intensity at different frequencies, giving it a constant PSD. This type of noise can be misinterpreted as microtexture within the 2D laser profile. In some embodiments, to mitigate the influence of induced microtexture due to sensor’s white noise, a filter (low-pass or band-pass) may be used to remove wavelengths smaller than the vertical resolution of the sensor.

[0088] Spikes can be defined as any data point that shows a short-duration, drastic elevation change. These points can break the trend of the pavement profile and can be identified visually. Module 204 can be configured to remove extreme outliers and mild outliers. As a non-limiting example, because certain cracks can be in the range of 1/16” (1.5 mm) to 1/2” (13 mm), Module 204 can classify spikes (e.g., having 1 or 2 data points) as an artifact for a given sensor.

[0089] Flat signals can be the result of the combination of a low exposure time for the camera and a very dark pavement surface. In this situation, the sensor can omit the measured information and output a flat line where the elevation at multiple locations is the same as the last “good” point measured by the sensor.

[0090] In some embodiments, a Sabillon-Orellana Filtering algorithm (Sabillon-Orellana, 2020) may be used that employs boxplots and the difference between consecutive points to detect and remove all instances of spikes and flatlines along with the profile data, including flat signals removal, spiked flatlines, mid-flatline spike removal, boxplot outlier, difference signal, fine-tuned spikes. Other algorithms may be employed, or some or part of the algorithm may be performed for any line-laser sensor and other contact sensors, e.g., those as described herein. [0091] Examples Process. Figs. 2C and 2D each shows aspects of an example process of operation for the set of profile data (e.g., 2000 profiles) collected with the laser sensor, including a pre-processing operation 228, a boxplot removal operation 230 (shown as “Boxplot Outlier Removal” operation 230), a flat signal removal operation 232, a difference removal operation 234, a spike flatline removal operation 236, a fine-tuned spike removal operation 238, a midflatline spike removal operation 240, and an end-point pre-imputation operation 242.

[0092] The preprocessing operation 228 includes pre-imputation, offset suppression, and difference computation operation that can be performed to determine five initial metrics per Table 1, which can be later used as the threshold inputs in the subsequent filtering stages (e.g., eight filtering stages). Table 1

[0093] At the pre-imputation step 246 of the pre-processing operation 228, operation 228 imputes the value of -97.4 in the profile data that may have missing data 248 to ensure that the dataset is complete. The offset suppression operation 250 is configured to center the profile data around a 0-mm value (252) in the vertical direction by computing the median and subtracting it from the 2D profile data. The difference computation operation (not shown) may compute a Backward Difference (BD) vector and a Forward Difference (FD) vector per Equation Set 1. Forward Difference (FD).

(Eq. Set 1) [0094] In Equation Set 1, pt is a point within the profile and p_L-± is the next following point, and p_i+1 is the immediate preceding point.

[0095] The boxplot removal operation 230 employs boxplots to remove outliers from the profile. Using the fourth (QI) and fifth (Q3) metrics from the pre-processing operation, the average interquartile range (IQR) of profiles is computed as the difference between Q3 and QI . For example, if QI = —3.74 mm, Q3 = 3.42 mm, then IQR = 0.94 mm. The boxplot filter is then defined as any value that is more extreme than either of these two quartiles by more than three times the interquartile range per Equation Set 2.

(Eq. Set 2) [0096] In Equation Set 2, X is the elevation at any point along with the profile. Plot 254 shows an example detected outliers, and plot 256 shows an example output of the boxplot filter. Using the values stated above, the threshold values for the boxplot filter can then be applied. In the example, X < -6.58 mm and X > 6.26 mm are used as the thresholds.

[0097] The flat signal removal operation 232 is configured to remove flat signals 260 from the profile data. Flat signals 260 can blur the measured information and can be observed as a flat line in which the observed elevation at multiple locations is the same at the last valid point measured by the sensor. Plot 262 shows an example output of the flat signal removal operation 232. Operation 232 can remove from the profile data any point that an entry in the BD vector where BD = 0. Plot 258 shows the profile data with the flat signals 260. Plot 260 shows the same profile data with the flat signals 262 removed.

[0098] The difference removal operation 234 is configured to remove mild spikes from the profile data. Operation 234 may employ the determined T1 and T2 from the pre-processing operation. In the above example shown in plot 264, T1 and T2 are -0.21 and 0.21 mm, respectively. Operation 234 may set the threshold value as 0.21 mm to differentiate a mild outlier from the rest of the profile data. The detection and removal criteria used by operation 232 is defined per Equation Set 3.

If (BDi > T2 AND FDi > T2), then remove the point If (BDi < T1 AND FDi < Tl), then remove the point

(Eq. Set 3) [0099] Plot 266 shows an example profile data after the difference removal operation 234. In some embodiments, the boxplot removal operation 230, the flat signal removal operation 232, and the difference removal operation 234 were observed to remove a substantial portion of the noisy data (e.g., greater than 85% for certain datasets).

[0100] The spike flatline removal operation 236 is configured to remove the first point in a flatline whenever that first point was a spike. The threshold used by the operation for the difference in elevation can be rounded up to the nearest integer. Plot 268 shows an example of the spike in the flatline data point, and plot 270 shows the spike removed.

[0101] The fine-tuned spike removal operation 238 is configured to capture mild spikes that were not removed in the prior spike removal stages. The threshold used at this stage is metric #3 (T,) computed in the pre-processing operation. In the example, the threshold value for metric #3 is determined to be 0.90 mm. Operation 238 can be performed per Equation Set 4.

If (BDi+i > T3 AND FDi < - T3), then remove the point If (BDi < - T3 AND FDi+i > T3), then remove the point If (BDi > T3 AND FDi+i < -T3), then remove the point

(Eq. Set 4) [0102] Plot 272 shows an example profile data with the mild spikes 276, and plot 274 shows the profile data after the mild spikes 276 are removed.

[0103] The mid-flatline spike removal operation 240 is configured to remove limited instances in which a single spike occurs in the middle of two flatlines, per Equation set 5.

If pi = NULL, then skip this value

If (pi 4 NULL) AND (pi+i = NULL) AND (pz-i = NULL), then remove the point

(Eq. set 5) [0104] Plot 276 shows an example profile data with the single spike 280, and plot 278 shows the profile data after the single spike 280 is removed.

[0105] The end-point pre-imputation operation 242 is configured to add endpoints of the profile data that are a part of a flatline to address missing data at the endpoints. The operation 242 can verify whether the endpoints and the points adjacent to the endpoints have been removed at previous stages. If the endpoints have been removed, then operation 242 can impute the first or last point with the median value of the profile height in a mean/median imputation operation. Plot 282 shows an example profile data with an endpoint that was removed from the flatline, and plot 284 shows the profile data with the endpoint inserted.

[0106] Data Imputation module (206). Subsequent to the noise removal, e.g., per Module 204, Module 206 is configured to fill in the missing data that may have been removed from the filtering. Imputation has been shown to produce better results than simply using whatever data is complete and deleting those cases that are not complete, which had been observed to lead to biased results (Rubin, 1976; Sabillon-Orellana, 2020).

[0107] In some embodiments, imputation may be employed using linear interpolation, seasonally decomposed missing value, simple moving average, exponential moving average, autoregressive integrated moving average, Stineman interpolation, stochastic regression, mean imputation, deterministic regression, spline interpolation, or a combination thereof. Plot 286 shows the profile data after the imputation operation via module 206 was performed using linear interpolation.

[0108] Profile Detrending module (208). Module 208 is configured to remove polynomial trends and offsets from a profile. After processing the profile data for noise and imputing all missing data points, further detrending may be performed to center all pavement profiles with respect to a flat horizontal plane at the origin prior to transforming the data from the spatial domain to the spectral domain. In some embodiments, integration or regression detrending may be performed to remove a linear or polynomial trend within time series data.

[0109] Plot 288 shows the output of a regression detrending operation that performs a linear or polynomial regression and then subtracts the regression line from the data to achieve an approximately stationary time series per Equation Set 6.

z(t) = y(t) - y(t)

(Eq. Set 6) [0110] In Equation Set 6, y(t) is the regression line that fits the profile, are regression coefficients, x is the transverse coordinate, z(t), is the detrended profile data, and y(t) is the original profile data. Plot 288 shows a time series with a linearly increasing trend 292. The best- fit line is estimated using linear regression and then subtracted from the data to provide the stationary time series in plot 290. While removing trends and offsets, Module 208 still preserves the seasonality of the pavement profile due to the aggregate gradation of the mix.

[0111] Example Machine Learning Operation

[0112] The analysis system 122 is configured to use the acquired measurement data to determine one or more pavement-associated parameters or associated values via a machine learning analysis 302.

[0113] Fig. 3 shows an example set of features 304, such as pavement amplitude parameters or features 306, pavement spatial parameters or features 310, hybrid pavement parameters or features 308, and/or pavement spectral parameters or features 312, that may be determined by the analysis system 122 in the machine learning analysis 302.

[0114] Pavement amplitude parameters or features 306 are amplitude-based or associated statistics that can characterize the pavement surface topography. Table 2 shows examples of pavement amplitude parameters or features 306 that may be employed, including Maximum Height (Rz), Absolute Height Average (R_a), Height Variance (Av), Root Mean Square (RMS), Skewness (Rs), Kurtosis (Rk), Ten Point Mean Roughness (Rt), Mean Profile Depth (MPD), and Solidity Factor (Rr).

Table 2

[0115] In Table 2, h_L is the elevation at point i; h is the mean elevation; n is the number of datapoints; h_pj is the j^th highest peak in the profile; h_Vj is the j^th lowest valley in the profile; h_m/₂ is the elevation value midway through segment; h_m is the elevation value at the end of segment. From experimental results, it was observed that the Kurtosis and Solidarity Factor has potential relevance in predicting and/or estimating the friction/skid resistance. Further descriptions of these features are provided in Sabillon, Christian, et al. Efficient Model for Predicting Friction on Texas Highway Network. No. FHWA/TX-22/0-7031-1. University of Texas at Austin. Center for Transportation Research, 2023, which is incorporated by reference herein in its entirety. Other amplitude-based or associated parameters, including those described herein, can be employed in predicting and/or estimating the friction/skid resistance by the analysis system 122.

[0116] Kurtosis is a measure of the combined size of the tails relative to the whole distribution. When Rk is positive, it indicates the presence of extremely high peaks or deep valleys. When it is negative, it indicates a lack of extreme peaks or values. Lastly, if the value of kurtosis is close to zero, it means that the distribution of height is about normal, with very few high peaks or deep valleys. The below figure shows a comparison between profiles with a positive, neutral, and negative Rk.

[0117] The solidity factor is the ratio between the maximum depth of valleys and the maximum height of the profile. Profiles with a negative R_r and high in magnitude are similar to those with a low R_t, whereas profiles with negative R_r and small in magnitude look similar to profiles with a high R_t.

[0118] Pavement spatial parameters or features 308 can measure the horizontal characteristics of the surface deviations. Table 3 shows examples of pavement statistic parameters or features 308 that may be employed, including Mean Cross Width (C_m), Cross Width Variance (Cv), and Cross Width Skewness (G).

Table 3

[0119] The equations in Table 3 are similar to those in Table 2, though instead of using elevations ( q), these equations use horizontal spacing (x , measured in mm. From experimental results, it was observed that the Mean Cross Width (C_m) has potential relevance in predicting and/or estimating the friction/skid resistance. Further descriptions of these features are provided in Sabillon, Christian, et al. Efficient Model for Predicting Friction on Texas Highway Network. No. FHWA/TX-22/0-7031-1. University of Texas at Austin. Center for Transportation Research, 2023.. Other spatial-based or associated parameters, including those described herein, can be employed in predicting and/or estimating the friction/skid resistance by the analysis system 122. [0120] The cross width measures the horizontal distance between inflection points along the profile data. The spacing parameters are always strictly positive.

[0121] Hybrid pavement parameters or features 310 are a combination of amplitude and spacing evaluations. Any changes that occur in either amplitude or spacing may have effects on the hybrid property. In tribology analysis, surface slope, surface curvature, and developed interfacial area are important factors that influence the tribological properties of surfaces. The tribology refers to all of the characteristics relating to interacting surfaces in relative motion, including friction, wear, and lubrication (Gadelmawla et al. 2002). Table 4 shows examples of pavement hybrid parameters or features 308 that may be employed, including a Two-Points Slope Variance (SV2) and a Six-Points Slope Variance (SV6).

Table 4

[0122] In Table 4, Ax is the spacing between two adjacent points. From experimental results, it was observed that the Two Points Slope Variance (SV2) has potential relevance in predicting and/or estimating the friction/skid resistance. Further descriptions of these features are provided in Sabillon, Christian, et al. Efficient Model for Predicting Friction on Texas Highway Network. No. FHWA/TX-22/0-7031-1. University of Texas at Austin. Center for Transportation Research, 2023.. Other spacing and amplitude hybrid-based or associated parameters, including those described herein, can be employed in predicting and/or estimating the friction/skid resistance by the analysis system 122.

[0123] The two-point slope variance (SV2) measures the slopes between two consecutive points as the difference in height between two consecutive coordinates, divided by the horizontal distance between them. The below figure shows a comparison between a profile with a high SV2 and a low SV2.

[0124] Pavement spectral parameters or features 312 are calculated in the frequency domain and are considered to be scale-independent, given that they are estimated along with a wide range of texture wavelengths covering multiple texture components (Serigos et al., 2016). Obtaining spectral parameters requires the use of Fourier analysis to examine the surface texture profile. In this project, two spectral statistics were used to characterize the PSD of the pavement profiles.

[0125] Pavement spectral parameters may be determined via data transform module 210.

The PSD is a description of how the energy of a pavement texture profile is distributed over the different frequencies. The PSD of a roadway is obtained by applying a DFT to the linear profile of a pavement surface to decompose it into a series of sinusoidal functions with discrete frequencies. Because so many sinusoids must be added together to build complex road profiles, individual amplitudes are almost always small. Hence, the Fourier Transform is adjusted to show how the variance of the profiles is distributed over a set of sinusoids. This adjustment is known as the PSD (Sayers and Karamihas, 1998). Serigos et al. (2016) used the slope and intercept of the linearized PSD curve to characterize the surface macro and microtexture. The method of least squares is used to compute the regression line that best fits the PSD curve and extract its slope and intercept. Furthermore, the logarithm is computed given that the values can be orders of magnitude different from one another. The output may be viewed as a logarithm.

[0126] In an example, Module 210 is configured to allow for the assessment of the same texture data via the application of a Fourier transform to the spectral/frequency domain for quantification of metrics such as the power spectral density (PSD) of the signal. It is noted that other transformations that can be used include Laplace transform, the Hilbert-Huang transform, and the wavelet transform.

[0127] To transform the texture data to the spectral domain, Module 210 may first employ a window operation to reduce the amplitude of the signal to zero at the edges of the profile and avoid leakage. Spectral leakage can occur when a signal is transformed into the spectral domain but with the wrong frequencies being amplified. Spectral leakage can be minimized by increasing the window size of the sampling period: the longer the signal is sampled, the less impactful the errors associated with leakage. In the pavement profile data, this sampling period can refer to the length of the scan. As the length of the scan increases, the bandwidth of the spectrum of the window can become narrower, and hence the effects of spectral leakage are minimized.

[0128] Module 210 may employ a Split Cosine Bell Window (SCBW) (292) because the length of the measured profile data is less than one meter. Equation 7 shows the Split Cosine

Bell Window function:

(Eq. 7A)

[0129] In Equation 7A, w_{i c} is the window coefficient, N is the number of data points, and i is the sample number. The window coefficient may be multiplied by the signal and later normalized by the integral of the window to prevent attenuation of the signal as per Equation 7B.

(Eq. 7B) [0130] In Equation 7B,

is the windowed profile height at point i (mm).

[0131] Module 210 may apply a DFT to the windowed profile per Equation 7C to transform the texture data from the spatial time domain into the spatial frequency domain.

(Eq. 7C) [0132] In Equation 7C, Z_k is the DFT of the windowed profile, and j is the imaginary unit ( j² = —1). The result of the DFT is a constant bandwidth narrow band spectrum with complex values. The bandwidth is a function of the evaluation length defined per Equation 7D.

(Eq. 7D) [0133] In Equation 7D, f_sp is the frequency interval (cycle/meter), and I is the evaluation length (m). The sensor should be capable of sampling two points with a spacing that should be less than two times the shortest wavelength of interest per Shannon theory.

[0134] The PSD can be determined ZPSD per Equation 7E.

(Eq. 7E) [0135] Training Operation

[0136] Fig. 4 shows an example machine- learning operation 400 to generate a trained machine-learning model. The training data for the training may include measured laser scans from chip seals with high macrotexture (e.g., HM-CS), dense fine mixes (e.g., DFM), chip seals with low macrotexture (e.g., LM-CS), open mixes (e.g., OM) or PFCs, dense coarse mixes (e.g., DCM), stone matrix asphalt (e.g., SMA), finish-graded concrete, and a combination thereof. [0137] The training may involve the evaluation of supervised and unsupervised training operations, as well as clustering operations to identify the pavement type. In some embodiments, k-means clustering or agglomerative hierarchical clustering may be employed, among others. A cluster refers to a collection of data points aggregated together because of certain similarities.

The clustering implementations are said to have an unsupervised learning phase because the dataset does not contain the true surface type of the pavement. In machine learning, this variable would be known as the label. Thus, the algorithm analyzes the data in such a way that it finds naturally occurring clusters by determining the similarities and differences between the texture statistics. These statistics are referred to as features in the machine learning context.

[0138] K-means clustering can cluster data by separating samples into k groups of equal variances by minimizing a criterion known as the inertia or the within-cluster sum of squares per Equation 8A.

(Eq. 8A) [0139] In Equation 8A, x_L is the feature vector for the i^tfl observation, /r₇ is the centroid of cluster j, C is the cluster space, and the operator || || is the norm. Inertia is a measure of the variability of the observations within each cluster. In general, a cluster that has small inertia is more compact than a cluster that has large inertia. K is a quantity that must be defined by the user before running the algorithm and refers to the number of centroids needed to create the clusters. Each cluster is characterized by the mean fij of the samples in the cluster. The means are commonly called the cluster “centroids,” and they can be the imaginary or real locations representing the center of each cluster.

[0140] The algorithm can first choose the initial centroids, e.g., choose k samples from the dataset. After initialization, the algorithm loops between two other steps. The second step assigns each sample to its nearest centroid. In this step, typically, “nearest” means the shortest Euclidean distance to the centroid. The third step creates new centroids by taking the mean value of all samples assigned to each previous centroid. The difference between the old and the new centroids is computed, and the algorithm repeats these last two steps until this value is less than a threshold. In other words, it repeats until the centroids do not move significantly (Choromanska and Monteleoni, 2012).

[0141] Given enough time, k-means will converge; however, this may be to a local minimum. This is highly dependent on the initialization of the centroids. As a result, the computation is often done several times, with different initializations of the centroids. One method to help address this issue is the k-mean initialization scheme. This initializes the centroids to be generally distant from each other, leading to generally better results than random initialization (Choromanska and Monteleoni, 2012).

[0142] To find the optimal initialization points for the centroids in implementing k-means, the elbow method may be employed. The elbow method is a heuristic approach that determines an optimal value of k by plotting different values for k against their corresponding distortion value to determine the average inertia of all clusters. As k increases, the average distortions will tend to decrease since each cluster will have fewer constituent instances, and the instances will be closer to their respective centroids. However, the improvements in average distortion will decline as k increases. The value of k at which improvement in distortion declines the most is the elbow and is the point of diminishing returns for the number of clusters in the data. [0143] Hierarchical clustering is a clustering algorithm that builds nested clusters by merging or splitting them successively. There are two main variations of this clustering technique: the agglomerative and the divisive approach. In the agglomerative approach, each observation starts in its own cluster, and pairs of clusters are merged as one moves up the hierarchy. This hierarchy of clusters is represented as a tree (or dendrogram). The root of the tree is the unique cluster that gathers all the samples, the leaves being the clusters with only one sample (Rokach et al., 2005). Upon visual inspection, one can have an estimate of the “optimal” number of naturally occurring clusters in the data, but it is recommended to use the elbow method to corroborate this number. [0144] To decide which clusters should be combined, a measure of dissimilarity between sets of observations can be determined. In most methods of hierarchical clustering, this is achieved using an appropriate metric, a measure of distance between pairs of observations, and a linkage criterion that specifies the dissimilarity of sets as a function of the pairwise distances of observations in the sets. The Euclidean distance is generally the most widely used metric to measure distances between two points when the number of features in the data is not too high. In terms of the linkage criterion, there are multiple options, but the one used for this analysis is known as the Ward linkage criterion. Ward minimizes the sum of squared differences within all clusters. It is a variance-minimizing approach and, in this sense, is like the k-means objective function (Rokach et al., 2005).

[0145] Labels may be assigned to the pavement surfaces based on the results from the unsupervised learning analysis. The labels may be used to train a supervised learning model to classify different pavement surfaces. Supervised machine learning algorithms are designed to learn by example. In supervised training, the model is given the correct label for each observation when learning the patterns of the data. During its training phase, the algorithm searches for patterns in the data that correlate with the desired outputs. After training, a supervised learning algorithm can take in new unseen inputs and determine which label to assign the new inputs based on the prior training data. For this project, a decision tree classifier was utilized.

[0146] Decision trees (404) are a non-parametric supervised learning method used for classification and regression. A classifier can use a decision tree as a predictive model to go from observations about an item, represented by the branches, to conclusions about the item’s target value, represented by the leaves. Tree models where the target variable can take a discrete set of values are called classification trees. In these tree structures, leaves represent class labels, and branches represent conjunctions of features that lead to those class labels. Decision trees where the target variable can take continuous values (typically real numbers) are called regression trees. The goal is to create a model that predicts the value of a target variable by learning simple decision rules inferred from the data features. Decision trees are among the most popular machine learning algorithms, given their intelligibility and simplicity (Wu et al., 2007, Piryonesi and El-Diraby, 2020).

[0147] Equation 9A shows a regression equation to predict skid using the output of the decision tree and other variables.

SN = /3₀ + ^(R_s) + [3₂(RMS) + p₃ LM CS + /?₄( MS) + /?₃(0MS)

(Eq. 9A) [0148] Equation 9 is a combination of variables and continuous random variables. This means the intercept /?₀ represents the pavement surface type that will be used as a baseline. /?₀ refers to the point at which the regression line for chip seals with high macrotexture crosses the y-axis when extrapolated backward. The coefficient

represents the effect that a unit increment in skewness has on the skid of the pavement. In certain experiments, it was observed that for every unit increment in skewness, the skid is reduced by 0.282. This means that regardless of the surface type, the more negative texture present at the pavement surface, the more skid the roadway will provide. The coefficient /?₂ represents the effect that a unit increment in RMS has on the skid of the pavement. The data indicates that for every unit increment in RMS, the skid is increased by 0.196. This means that pavements with lots of macrotexture (high deviations from the horizontal plane) will have, on average more skid, than pavements with smaller deviations. The coefficient /?₃ represents the differential effect between an LM CS and HM CS. The data indicates LM CS has, on average, 0.130 fewer skids than HM CS. The coefficient /?₄ represents the differential effect between a DMS and HM CS. The data indicates dense mixes have, on average, 0.346 less skid than HM CS. Lastly, the coefficient /?₅ represents the differential effect between OMS and HM CS. The data indicate that OMS have, on average, 0.478 less skids than HM CS. A summary of the regression model and the goodness of fit statistics is provided in Table 5. Table 5

[0149] Equation 9B shows a final regression model to predict skid using field texture data.

SN=0.238-0.282(R_s )+0.196(RMS)-0.130(LMCS)-0.346(DMS)-0.478(OMS)

(Eq. 9B) [0150] Fig. 4B shows an example plot of the regression model of Equation 9B. Other types of machine learning algorithms, as described herein, may be employed.

[0151] Machine Learning. In addition to the machine learning features described above, the analysis system can be implemented using one or more artificial intelligence and machine learning operations. The term “artificial intelligence” can include any technique that enables one or more computing devices or comping systems (i.e., a machine) to mimic human intelligence. Artificial intelligence (Al) includes but is not limited to knowledge bases, machine learning, representation learning, and deep learning. The term “machine learning” is defined herein to be a subset of Al that enables a machine to acquire knowledge by extracting patterns from raw data. Machine learning techniques include, but are not limited to, logistic regression, support vector machines (SVMs), decision trees, Naive Bayes classifiers, and artificial neural networks. The term “representation learning” is defined herein to be a subset of machine learning that enables a machine to automatically discover representations needed for feature detection, prediction, or classification from raw data. Representation learning techniques include, but are not limited to, autoencoders and embeddings. The term “deep learning” is defined herein to be a subset of machine learning that enables a machine to automatically discover representations needed for feature detection, prediction, classification, etc., using layers of processing. Deep learning techniques include but are not limited to artificial neural networks or multilayer perceptron (MLP).

[0152] Machine learning models include supervised, semi-supervised, and unsupervised learning models. In a supervised learning model, the model learns a function that maps an input (also known as feature or features) to an output (also known as target) during training with a labeled data set (or dataset). In an unsupervised learning model, the algorithm discovers patterns among data. In a semi-supervised model, the model learns a function that maps an input (also known as feature or features) to an output (also known as a target) during training with both labeled and unlabeled data.

[0153] Neural Networks. An artificial neural network (ANN) is a computing system including a plurality of interconnected neurons (e.g., also referred to as “nodes”). This disclosure contemplates that the nodes can be implemented using a computing device (e.g., a processing unit and memory as described herein). The nodes can be arranged in a plurality of layers such as input layer, an output layer, and optionally one or more hidden layers with different activation functions. An ANN having several hidden layers can be referred to as a deep neural network or multilayer perceptron (MLP). Each node is connected to one or more other nodes in the ANN. For example, each layer is made of a plurality of nodes, where each node is connected to all nodes in the previous layer. The nodes in a given layer are not interconnected with one another, i.e., the nodes in a given layer function independently of one another. As used herein, nodes in the input layer receive data from outside of the ANN, nodes in the hidden layer(s) modify the data between the input and output layers, and nodes in the output layer provide the results. Each node is configured to receive an input, implement an activation function (e.g., binary step, linear, sigmoid, tanh, or rectified linear unit (ReLU) function), and provide an output in accordance with the activation function. Additionally, each node is associated with a respective weight. ANNs are trained with a dataset to maximize or minimize an objective function. In some implementations, the objective function is a cost function, which is a measure of the ANN’s performance (e.g., error such as LI or L2 loss) during training, and the training algorithm tunes the node weights and/or bias to minimize the cost function. This disclosure contemplates that any algorithm that finds the maximum or minimum of the objective function can be used for training the ANN. Training algorithms for ANNs include but are not limited to backpropagation. It should be understood that an artificial neural network is provided only as an example machine learning model. This disclosure contemplates that the machine learning model can be any supervised learning model, semi-supervised learning model, or unsupervised learning model. Optionally, the machine learning model is a deep learning model. Machine learning models are known in the art and are therefore not described in further detail herein.

[0154] A convolutional neural network (CNN) is a type of deep neural network that has been applied, for example, to image analysis applications. Unlike traditional neural networks, each layer in a CNN has a plurality of nodes arranged in three dimensions (width, height, depth). CNNs can include different types of layers, e.g., convolutional, pooling, and fully-connected (also referred to herein as “dense”) layers. A convolutional layer includes a set of filters and performs the bulk of the computations. A pooling layer is optionally inserted between convolutional layers to reduce the computational power and/or control overfitting (e.g., by downsampling). A fully-connected layer includes neurons, where each neuron is connected to all of the neurons in the previous layer. The layers are stacked similar to traditional neural networks. GCNNs are CNNs that have been adapted to work on structured datasets such as graphs.

[0155] Other Supervised Learning Models. A logistic regression (LR) classifier is a supervised classification model that uses the logistic function to predict the probability of a target, which can be used for classification. LR classifiers are trained with a data set (also referred to herein as a “dataset”) to maximize or minimize an objective function, for example, a measure of the LR classifier’s performance (e.g., error such as LI or L2 loss), during training. This disclosure contemplates that any algorithm that finds the minimum of the cost function can be used. LR classifiers are known in the art and are therefore not described in further detail herein.

[0156] An Naive Bayes’ (NB) classifier is a supervised classification model that is based on Bayes’ Theorem, which assumes independence among features (i.e., the presence of one feature in a class is unrelated to the presence of any other features). NB classifiers are trained with a data set by computing the conditional probability distribution of each feature given a label and applying Bayes’ Theorem to compute the conditional probability distribution of a label given an observation. NB classifiers are known in the art and are therefore not described in further detail herein.

[0157] A k-NN classifier is an unsupervised classification model that classifies new data points based on similarity measures (e.g., distance functions). The k-NN classifiers are trained with a data set (also referred to herein as a “dataset”) to maximize or minimize a measure of the k-NN classifier’s performance during training. This disclosure contemplates any algorithm that finds the maximum or minimum. The k-NN classifiers are known in the art and are therefore not described in further detail herein.

[0158] A majority voting ensemble is a meta-classifier that combines a plurality of machine learning classifiers for classification via majority voting. In other words, the majority voting ensemble’s final prediction (e.g., class label) is the one predicted most frequently by the member classification models. The majority voting ensembles are known in the art and are therefore not described in further detail herein.

[0159] Example Evaluable Pavement Surfaces

[0160] The exemplary system and method can be used to evaluate flexible pavement, concrete, and stone mastic asphalt, among other pavement materials described herein.

[0161] Flexible pavements can include dense-graded mixes, surface treatments, or open- graded mixes, among others. These mixes and their variations use different types of binders and aggregates in different proportions and have different purposes. For instance, open-graded mixes are designed to allow the flow of water through the pores and prevent water accumulation at the surface. However, in terms of their surface macro and microtexture characteristics, most of the flexible pavement mixes can be grouped into three broad categories: surface treatments, dense- graded mixes, and open-graded mixes.

[0162] Surface treatments encompass multiple pavement-surface treatments, including crack seal, fog seal, slurry seal, chip seal, geotextile seal, microsurfacing, and others. One of the most common treatments, the traditional chip seal, consists of an asphaltic material sprayed over a pavement surface, followed by a uniformly graded aggregate cover (12). Chip seals are employed to seal small cracks, waterproof surfaces, and improve friction, providing a long- wearing course on low-volume roads at a relatively low initial cost compared to other treatments [12, 13], They could be single seals or multiple seals. In general, chip seals are expected to have high levels of surface texture that, in turn, improve the skid resistance of the road but also tend to increase the tire/pavement noise.

[0163] Dense-graded mixes are pavements produced by combining hot asphalt binder with well-graded aggregates. These mixes are intended for general use and can accommodate a wide range of traffic volumes. They are generally referred to by their nominal maximum aggregate size and are relatively impermeable when properly designed and constructed. Dense mixes can be further classified as either fine-graded or coarse-graded, where fine-graded mixes include more fines and sand-sized particles than coarse-graded mixes [14],

[0164] Open-graded mixes consist of crushed stones and a small percentage of manufactured sands. Unlike dense-graded mixes, open-graded hot mix asphalts are designed to be water permeable. Because of their open structure, precautions are taken to minimize asphalt draindown using fibers or modified binders. Two typical open-graded mixes are open-graded friction course (OGFC) and asphalt-treated permeable base (ATPB). OGFC, also referred to as permeable friction course or PFC, is used for surface courses only and typically includes more than 15 percent air voids with no maximum air voids specified. The high air voids reduce tireroad noise by up to 50 percent. ATPBs are used as a drainage layer below dense-graded hot mix asphalt and therefore have less stringent specifications than OGFC [15],

[0165] Three major types of asphalt surfacing are characterized by a mixture of bitumen and stone aggregate and include Dense Graded asphalt (DGA); Stone Mastic Asphalt (SMA), and Open Graded Asphalt (OGA). Asphalt surfacing differs by the proportion of different sizes of aggregate, the amount of bitumen added, and the presence of other additives and materials.

[0166] Experimental Results and Examples

[0167] First Study (friction and/or skid resistance estimation via non-contact measurements)

[0168] A study was conducted sponsored by the Department of Transportation of Texas that developed texture data collection prototype equipment to measure roadway texture at highway speeds. The study used the collected data along with sophisticated signal processing techniques and data analytics to create a pavement surface classification model using machine learning techniques. The study developed a model to classify asphalt pavement surfaces to be later used in the prediction of skid resistance based on texture statistics. Cluster analysis demonstrated that, by using a combination of amplitude, spatial, and hybrid texture statistics, clear distinctions could be made about six asphalt pavement surface types found in the areas evaluated in the study: two distinctive chip seals, two types of dense-graded mixes, and two types of open-graded mixes.

[0169] While the strong correlations between pavement texture and other surface characteristics, such as friction, drainage, and noise, are widely known, those relationships are not straightforward and have a strong dependency on the type and properties of the pavement surface. Incorporating surface information into a prediction model for skid resistance has the potential to drastically increase its predictive power. In practice, pavement surface information is often unknown at the network level and must be inferred based on local expertise. The study addressed this technical issue by developing an objective classification model capable of identifying different asphalt pavement surfaces with a high degree of accuracy using only field texture data.

[0170] High-resolution texture data were collected on 21 highway sections with varying surface types within 60 miles from the city of Austin using a prototype developed in-house. Sophisticated data-processing algorithms were created to ensure the data used in developing the classification model were of the highest quality possible. Based on the most recent literature on the topic, more than twenty texture summary statistics were assessed during this study to find the best combination to predict the surface characteristics of asphalt pavements.

[0171] The results of this study are robust and indicate that texture summary statistics alone, when used in the right combination, have enough information to develop a pavement surface classification model with a predictive power as high as 89% based on the Fl score.

[0172] The study determined that mean cross-width Cm, 2-point slope variance SV2, solidity factor Rr, and kurtosis Rk are strong predictors in classifying pavement surface type, providing a predictive power of 89% in terms of Fl score using a decision tree analysis. Indeed, statistical models already in place for skid resistance can now use the surface information of the pavement in disclosed analysis. In addition, agencies can use the classifier at a network level to identify the type of pavement surface present at different locations along the highway system. While 2D pavement data were used, it is expected that pavement images can be employed via a prediction model and a neural network classifier to further enhance the capabilities of the model, e.g., to between very specific mix types, such as a dense-graded mix with different aggregate sizes. The analysis can be extended to rigid pavements, e.g., to determine the type of pavement finish or for use as a quality control measure of the type of surface finishing.

[0173] Methodology. Fig. 5A shows the five stages undertaken in the study: data collection, data processing, feature engineering, pavement surface prediction, and skid prediction modeling. In the data collection stage (502), the study developed a prototype data measuring and collection equipment to simultaneously collect friction and textual data on the same wheel path and at high speeds. In the data processing step (504), an analysis system was used to clean the texture and friction data. At the feature engineering step (506), the study computed texture summary statistics and manipulated them to enhance the efficiency of the prediction models. At the pavement surface prediction stage (508), the texture data that had already been analyzed were fed into multiple machine learning models to create the most accurate prediction for the pavement surface present at each of the surveyed sites. The overall model accuracy was 94%. The study then created a friction prediction model (510) that best predicts friction for a field texture data. The result from the friction prediction model is greater than 80% in adjusted R² and is expected to increase with additional surface types employed in the training. R software was used for quality control and feature engineering (data processing of the laser distance measurement data). Python software is used for modeling (machine learning and regression). [0174] The same friction data would not be required after the model has been trained.

[0175] During this study, eight distinct highways around Austin were selected, and texture data were collected on at least two locations along each highway. Texture data for each highway section were collected at highway speeds using a prototype system that includes a GripTester trailer used for validation of predicted friction and a line laser sensor to simultaneously collect skid resistance and laser scan profiles. The sensor can measure the full spectrum of pavement macrotexture (0.5 to 50 mm) and a portion of the first decade of microtexture (up to a wavelength of 0.35 mm) at highway speeds. The specifications of the laser sensor are shown in Table 6.

Table 6

[0176] The laser-sensor was synchronized with the GripTester and centered in front of the measuring tire and the water outlet. The laser-sensor can measure 330 mm in width on the transverse direction, enabling the researchers to capture the Griptester’s testing tire’s contact area with the pavement (50 mm [2 in.] wide). The two systems work together as they are both simultaneously triggered to collect data. The trigger in the laser-sensor is used to capture a single cross-sectional profile of height values. When the Griptester’s tires are inflated to 20 psi, the diameter is 254 mm, resulting in profile spacing of approximately 40 mm, subject to vary as the tire expands and contracts.

[0177] Data Collection Protocol Using the Prototype. The study started the test when the survey vehicle reached the test section that had been previously marked and pre-programmed in a portable GPS. Near the test site, the data collection system was connected to the test vehicle via two cables for data communication with the test acquisition computer from inside the vehicle. Two individuals conducted the test: the first was driving the vehicle and the second operating the software. During the data collection, two sets of software are used: SkidTexture, an in-house software to measure texture, and Roads, proprietary software for the GripTester. The vehicle was stationary when initializing both pieces of software. Once initialized, the driver accelerated the vehicle to the data collection speed of 50 mph. The second user monitored the real-time data from the laser-sensor (cross-sectional pavement profiles, MPD, camera exposure, and the laser’s brightness threshold) and real-time data from the GripTester (grip number, target water flow, actual water flow, and vehicle speed). As data was collected, the software operator can make notes that are saved with the data and stamped with the corresponding location. This feature is important when there is a specific start and stop in the test section or when the operator needs to denote changes in the pavement surface. As data was collected, the operator checked the real-time data and confirmed the proper operation of the system visually.

[0178] Data Processing. Before performing any statistical analysis, the cross-sectional pavement profiles are subjected to a series of signal-processing operations. The study trimmed the acquired texture profiles to 100 mm to include only the texture that is within and in the path of the Griptester’s test path. The width was kept larger than the actual contact area to allow the estimation of the MPD. The study applied a robust filter to identify and remove noisy data points such as spikes and flat signals. Spikes were defined as any data point portraying a rapid and drastic elevation change. Flat signals were the result of the combination of a low exposure time for the camera and a very dark pavement surface. The study used a denoising algorithm that centered the profile around an elevation of zero and subtracted the median elevation from the profile. The denoising algorithm then detected and removed, using a boxplot, the profile’s interquartile range in elevation. Mild spikes and flatlines were removed by taking the difference between adjacent points and removing outliers and zeros. The three steps alone removed, on average, 85% of all data points that were deemed as noise after visual inspection.

[0179] The study also performed linear interpolation [6] to impute all the noise data points that were removed by the denoising algorithm. Linear interpolation provides strong imputation in terms of accuracy, computation time, and computing unbiased statistics such as MPD or kurtosis [17], The study also executed a regression detrending algorithm to remove polynomial trends from the profile and center it around the flat horizontal plane at zero elevation. Regression detrending entails fitting a line through the profile and subtracting the regression line from the data to achieve an approximately stationary time series.

[0180] Quality Control Filtering. The study performed a noise removal analysis to remove any spikes that would skew the data. The threshold at which a profile was considered invalid was set at 33%. Any profile with a percentage of noise strictly higher than the threshold was discarded. In terms of data points in a profile, 33% noise corresponds to more than 205 noisy data points out of a total of 622. This threshold is high compared to other standards, but it was observed that chip seals with high macrotexture or dark fresh-asphalt pavements yielded higher noise percentages.

[0181] Texture Summary Statistics. The study characterized pavement texture using summary statistics of the surface profile, including spatial and spectral parameters.

[0182] Spatial parameters were calculated in the spatial domain and are scale-dependent, i.e., the same parameters can be defined separately for different components of texture. Spatial statistics were further broken down into four sub-categories: amplitude, spatial, hybrid, and functional. Amplitude parameters were used to measure the vertical characteristics of the surface deviations. Spacing parameters measured the horizontal characteristics of the surface deviations. The hybrid property was a combination of amplitude and spacing. The functional parameters gave information about the surface structure based on the material-bearing ratio curve.

[0183] Spectral parameters were calculated in the frequency domain and were considered to be scale independent, given that they were estimated along with a wide range of texture wavelengths covering multiple texture components [1], The study determined kurtosis, profile solidity factor, mean cross width, cross width variance, and 2-pt slope variance (Table 7) to be the strong predictors of pavement surface types.

Table 7

[0184] Statistic Normalization. The study employed machine learning algorithms that attempt to find trends by analyzing and comparing every observation in the dataset. To address different scales, the study employed Z-score normalization to remove biases that were measured at different scales. Z-score normalization adjusts the values of each variable in the data to have a mean of zero and a unit variance by applying the transformation Z = x-x!S_n, where Z is the standard normal score, Sn is the sample standard deviation, and x is the sample mean of the variable. Fig. 5B shows texture statistics used in this experiment. The top plots of Fig. 5B illustrate the dissimilarity in the scale of four statistics used in this analysis. As the figure demonstrates, some of the statistics have multiple outliers. Once normalization was completed, all statistics were transformed into standard normal variables with similar scales, as shown in the bottom plots of Fig. 5B.

[0185] Data Mining. The study used data mining techniques, including k-clustering and agglomerative hierarchical clustering, to determine whether the texture summary statistics contained sufficient information for the classification operation. The study analyzes the data to identify naturally occurring clusters by determining the similarities and differences between the texture statistics. The elbow method was used, via k-clustering, to determine the optimal k = 6 number of clusters. The elbow method employs a heuristic to determine an optimal value of clusters by plotting different values for k against their corresponding distortion (average variability of the observations within each cluster across all the clusters). The value of k at which improvement in distortion declines the most is the elbow. Both clustering algorithms provided almost identical clusters for the texture data.

[0186] The six pavement surface types identified in the study included: (i) chip seals with high macrotexture and rounded aggregates (HM - CS); (ii) chip seals with lower macrotexture and angular aggregates (LM - CS); (iii) open-graded mixes (OM); (iv) open mixes where the asphalt binder exudate to the near-surface (BOM); (v) dense coarse mixes (DCM); and dense fine mixes and micro-surfacing (DFM).

[0187] Fig. 5C shows the distribution of two statistics used during the cluster analysis. In Fig. 5C, it can be observed that the factor of solidity demonstrates a clear distinction between the two chip seals and every other mix. The mean cross-width analysis shows that once the chip seals have been classified, they can be readily isolated for the open mixes. The clustering results were validated by personnel from the Maintenance Division at the Texas Department of Transportation (TxDOT).

[0188] Pavement Surface Classifier. The study trained a decision tree classification model [22] to classify different pavement surfaces using the results from the cluster analysis. Decision tree is a non-parametric supervised learning method used for classification and regression. This classifier uses a decision tree as a predictive model to go from observations about an item to conclusions about the item’s target value. Tree models where the target variable can take a discrete set of values are called classification trees. For the tree structure developed in this study, leaves represent the pavement surface prediction, and branches represent conjunctions of texture statistics that lead to those predictions.

[0189] The study analyzed each data point in this model corresponds to a single profile. In total the study used 2,612 data points for CS #1; 2,356 for CS #2; 2,418 for OM; 2,911 for DCM; 2,418 for BOM; and 1,181 for DFM. The study used 80% of the data to train the model and the remaining 20% to test it. A statistic Fl score was used to determine the accuracy of a multi class classification model. The Fl score is the harmonic mean between the model precision and recall. It ranges from zero to one, where zero is the worst, and one is perfect accuracy. Precision is determined by the proportion of items correctly classified within each of the groups. Recall is the proportion of correct predictions made by the algorithm. Classification models favor the Fl score as opposed to using the ratio of the true positives to the total number of observations (model accuracy) because the Fl score is the most robust measure of accuracy for imbalanced datasets — i.e., if one of the groups being classified has significantly more observations than the other groups.

[0190] Fig. 5D shows the confusion matrix (both normalized and non-normalized) of the model. The confusion matrix visually shows an assessment of the accuracy of the model. A reliable model should have the main diagonal of the confusion matrix populated with a high percentage of the observations. The observations on the main diagonal indicate that the predictions made by the model are correct. Observations off the main diagonal are false positives or false negatives. The overall model accuracy was determined to be 0.89.

[0191] Fig. 5E shows the full generated decision tree diagram. Table 8 shows the classification summary for the testing of the decision tree classifier.

Table 8

[0192] Per Table 8, it can be observed that the model has an average accuracy of 86% when classifying both types of chip seals. In terms of classifying the open-graded mixes, the classifier has an almost perfect classification with 97% and 98% in the Fl score for BOM and OM, respectively. Finally, DCM and DFM have the relatively lowest Fl scores, with 80% each. In the case of the DCM, most of the misclassifications arose because the classifier confused some of the DCM sections with chip seals or DFM. In the case of DFM, most misclassifications happened because the section was considered a DCM. Nonetheless, an 80% in Fl score is still a high accuracy from a practical standpoint. The overall accuracy of the model measured in terms of Fl score is 89% (Table 6). This value is extremely high and proves that clearly there is enough information in cross-sectional pavement profiles to make a practical model for predicting the pavement surface at a network level with a high degree of accuracy and reliability.

[0193] Validation Study (friction and/or skid resistance estimation via non-contact measurements) [0194] After developing a methodology to process and analyze the texture and friction data, a second study (or a second part of the above-discussed study) was performed that selected a total of twenty-nine pavement sections close to the city of Austin, Texas that cover a wide range of different textures and skid numbers encountered in Texas. The study demonstrated that texture statistics computed using high-definition texture profile instrumentation have a statistically significant influence on pavement skid resistance and can be used, at least, to predict the pavement surface and to predict friction with a high degree of accuracy.

[0195] Fig. 6A shows the pavement sections used in the validation study. Fig. 6A also shows a table with a summary of the highway and the measured surface present at each roadway section.

[0196] Out of the twenty-nine sections, fifteen were surveyed using only the exemplary noncontact measurement system, and these were tested during the development stage. The other fourteen sections with varying pavement surfaces were tested simultaneously with the exemplary and the locked wheel tester (LWT) as the ground truth. The exemplary non-contact measurement system and the LTW characterized the skid of the road in terms of grip number (GN) and skid number (SN), respectively. Both measurements have a scale from 0.0 to 1.0, where zero represents a frictionless surface, and one is a surface with maximum friction. These data were used to validate the friction measurements obtained with the exemplary non-contact measurement system.

[0197] The texture measurements collected with the in-motion laser sensor were crossverified with readings from stationary tests, including those of CTM (ASTM E 2157), LLS, and SPT (ASTM E 965). The skid measurements collected with the exemplary non-contact measurement system were also correlated with the DFT (ASTM E 1911).

[0198] Correlation between Non-Contact Measurement and LWT. Fig. 6B shows the average skid reading of the exemplary non-contact measurement system and the LWT for the fourteen sections that were surveyed using both pieces of equipment. It can be observed in Fig. 6B that the correlation is high even though both pieces of equipment measure skid resistance at different transverse positions. The few sections with higher discrepancies found were those where the surface texture on the inner wheel path and along the centerline were significantly different. Although the friction prediction models were developed to generate grip numbers by the exemplary system study, it is contemplated that the exemplary non-contact measurement system can be employed to predict/estimate skid number (SN).

[0199] Speed Impact Evaluation. The study also quantified the impact of speed on the measured grip number by the exemplary non-contact measurement system. The study collected data on a straight roadway at varying speeds while measuring the skid resistance at increments of 0.5 miles at speeds of 40, 50, and 60 mph. The data indicated that for every additional mph in the speed of the vehicle (starting at 40 mph and ending at 60 mph), the GN was reduced by 0.003. Thus, even though the data shows with 95% confidence that this reduction in skid due to increasing speeds is statistically significant, the value of 0.003 is too low to cause any dramatic drop in the skid of the road for the range of speeds being measured. To this end, even if the vehicle went from 40 to 60 mph, the predicted reduction in skid would be equal to 0.06, which is not a significant difference.

[0200] To validate the measurements collected with prototype, a series of tests of hypothesis were conducted to check if the average texture or skid measurement is significantly different from the average measurements taken with the stationary equipment. The null hypothesis for this test is that the measurement from the prototype is identical to the measurement from the stationary equipment.

[0201] Fig. 6C shows the results of the hypothesis test for all the skid measurements, and Fig. 6D shows all the relevant texture measurements. From the hypothesis tests, it can be observed that the measurements from both the exemplary non-contact measurement system closely resemble the F60 skid measurements for the DFT. Surprisingly, the DFT measurements at 80 km/h are drastically different from the SN and GN measurements. This result indicates that there may be benefits in converting the measurements of GN to SN using linear regression.

[0202] In terms of texture, it can be observed that the measurements of the exemplary noncontact measurement system are consistent with the LLS for most of the statistics. The statistics where these metrics differ the most are the cross-width variance, where the measurements of the LLS are an order of magnitude larger.

[0203] Cluster analysis. The study performed cluster analysis of the pavement surfaces. IT was observed that there appeared to be six clusters that correspond to six different flexible pavement surfaces, including chip seals with medium to low macrotexture (raveling, flushing or aggregate polishing), chip seals with high macrotexture (good condition), dense coarse mixes (Types D and C), dense fine mixes (Type F and TOM), open Friction Coarse surfaces (PFC), and stone Matrix Asphalts (SMA).

[0204] This information was then corroborated with the expert opinion of the transportation department, further demonstrating that the texture data alone (as btexture statistics computed from 2D pavement profiles) appears to have sufficient information to characterize distinct asphalt pavement surfaces accurately. Indeed, the results of the clustering analysis indicated that the texture information could be used to predict the type of asphalt surface present in a highway. [0205] Decision Tree Classifier Evaluation. The study then evaluated the accuracy of a classification decision tree model using Fl score. The Fl score is the harmonic mean between the precision and the recall of the model. Precision is the proportion of correctly classified groups over the total number of observations for each group. Recall is the proportion of correctly classified groups over the total number of predictions for each group. Four sections out of the six that could be detected were used for the pavement surface classifier. In terms of their surface skid properties SMA, dense fine mixes and dense coarse mixes behave in similar ways, thus they can be categorized together into a single group. Thus, the four pavement surfaces the classifier could predict were high macrotexture chip seals (HM CS), low to medium macrotexture chip seals (LM CS), dense mix surfaces and SMA (DMS), and open-graded mix surfaces (OMS) [0206] The study used the set of all statistics to train the first decision tree. However, only the most relevant statistics were kept for the final decision tree in a process of elimination by trial and error. Furthermore, 80% of the data was used to train the model, and the remaining 20% was used to test the model. Fig. 6E shows the final tree and the results of the testing.

[0207] The decision tree (Fig. 6E) was trained using over 10,000 pavement profiles coming from different types of flexible pavement surfaces and tested with over 4,000 profiles. The study conducted pre-pruning and post-pruning on the decision tree model to avoid overfitting and have a robust classifier with a high degree of accuracy. The classifier employed the texture statistics: cross-width variance (Cv), ten-point mean roughness ( /? t), solidity factor (Rr and two-point slope variance (SV 2)

[0208] The decision tree is configured to output one out of four flexible pavement surfaces: “HM CS” for a chip seal with high macrotexture, “LM CS” for a chip seal with low macrotexture, “DMS” for stone matrix asphalts or dense mixes (both coarse or fine), and “OMS” for the open graded mixes or PFCs.. [0209] Fig. 6E shows that at an individual level, the smallest accuracy the decision tree model has is 82%, in Fl Score when classifying HM CS. This percentage is relatively low compared to the other Fl scores because as soon as chip seals with high macrotexture start experiencing aggregate polishing after years of wear and tear, they start behaving more like a LM CS. In the few instances where HM CS is misclassified, it is misclassified as an LM CS, and this can be confirmed by looking at the confusion matrix shown in Fig. 6F.

[0210] On the aggregate, the overall accuracy of the model measured in terms of Fl score is 94%. This value is extremely high and proves that there is enough information obtained from the texture profile alone to make a model that is applicable and reliable for predicting the pavement surface at a network level.

[0211] Skid number estimation. The study then evaluated the estimation of skid in terms of

SN using texture data. A linear regression suggests that the relation between GN and SN is linear and can be estimated by Equation 10.

[0212] In Equation 10, the parameter yo represents characteristics of the skid number that are not captured in the regression model. For example, the difference in skid between the inner wheel path and the aggregates in between wheel paths. The parameter yi can capture the effect that a unit increment in the measured GN has on the SN of the road. The data indicated that for every unit increment in GN, the SN increases by 0.634. The goodness of fit of this relation is 0.69 in R² , which is a good statistical fit to the data. Fig. 6G shows a summary of the regression analysis, and Fig. 6H shows a visualization of the average grip number versus the skid number for the sections for the exemplary system (referenced as the “Griptester”) and the ground truth system (referenced as “LWT”).

[0213] Additional experimental results may be found in Sabillon, Christian, et al. Efficient Model for Predicting Friction on Texas Highway Network. No. FHWA/TX-22/0-7031 -1. University of Texas at Austin. Center for Transportation Research, 2023, which is incorporated by reference herein.

[0214] Third Study (pavement type estimation via non-contact measurements)

[0215] A third study, or a third portion of the above-discussed study, was conducted that developed texture data collection prototype equipment to measure roadway texture at highway speeds. The study further captured high-speed image data to be used in a classifier to provide an estimation of pavement type to be used in conjunction with the textual data for friction and/or skid resistance estimation. It is contemplated that the high-speed image data and derived pavement type analysis may be used independent of the friction and/or skid resistance estimation, e.g., as its own application domain.

[0216] The study considered neural networks as the classifier. In an example, the neural network used in the study is based on ResNet 50 and is configured with 50 layers. The neural network was trained with 1000 object categories using 224 x 224 image input. Table 9 shows the object categories.

Table 9

[0217] Figs. 7A and 7B show two test setups for instrumentation mounted onto a truck and an automobile. In Fig. 7A, a high-speed camera is mounted to the back of the truck. In Fig. 7B, a high-speed camera is mounted to the side door of an automobile along with light illumination system comprising two spot lights.

[0218] Fig. 7C shows example images acquired by the high-speed camera system at high speed. Fig. 7D shows images of example pavement types over which the high-speed camera system had captured images. Fig. 7E shows the results of the classifier for a set of pavement types. It can be observed that the developed system could determine pavement type from highspeed images.

[0219] In the study, a high-speed camera manufactured by ArduCam was employed. The camera was configured to capture high speed at 10-ps shutter speed (for a vehicle travel speed of 60 miles per hour). Other cameras were evaluated, including those that captured the images at 125 ps, which was observed to be blurry. A 6-ps shutter speed was also evaluated and observed to provide a sufficient image or video frame for the classification. Fig. 7F, subpanels A and B, shows an example captured image for the 10- ps shutter speed camera system; subpanel A shows a static captured image, i.e., 0 miles per hour; subpanel B shows high-speed captured images at 60 miles per hour. The study collected high-speed images in combination with in-line laser measurements while traveling at 65+ miles per hour over various pavement types.

[0220] The study observed that the above-discussed neural network could correctly identify pavement types from high-speed camera images acquired at a vehicle traveling at high-speed (up to 65+ MPH).

[0221] Discussion

[0222] Background Highway Pavement. Traditionally, highway pavements have been designed to resist deformation and cracking under the combined action of traffic and the environment. Current demands on highways go beyond providing solely structural integrity. Modern freeways must provide adequate skid and rolling resistance, low noise, reduced splash and spray, a high level of rider comfort, and structural soundness. To accommodate all those demands in a cost-effective manner, highway transportation agencies must provide proactive pavement maintenance to their network. However, achieving that goal requires pavement management systems to contain complete information about their network’s condition on a yearly basis. This can be especially challenging for surface characteristics such as skid resistance. Skid resistance is a surface characteristic that changes seasonally and over the years. Measuring friction with stationary equipment like the dynamic friction tester or the British pendulum at a network level would be too time-consuming and require constant traffic control to avoid exposing field personnel to incoming traffic. Likewise, collecting friction measurements dynamically using equipment like the locked wheel tester or the sideway-force coefficient routine investigation machine (also known as SCRIM) is a resource- intensive and inefficient process that requires large volumes of water to test a few miles of highway. As highway networks continue to grow, performing full coverage measurements on a yearly basis with the available budgets becomes impossible. However, highway agencies can circumvent this issue by using pavement texture data to estimate pavement skid with a high degree of accuracy and reliability.

[0223] Studies have shown a strong correlation between pavement texture and skid as well as between texture and noise; however, those relationships are not straightforward and greatly depend on the type of pavement surface used in the roadway (1, 2). Zuniga demonstrated that incorporating surface information into a regression model to predict skid resistance can drastically improve the predictive power of the model from 50 to 70 percent (1). However, in practice, the surface information for the pavement is not known at a network level, and even at a project level, this information may still not be easy to obtain. For this reason, the authors of this study have investigated ways of predicting the pavement surface using texture data that is readily available. This study proposes an innovative way of using field texture data to estimate the pavement surface type of a roadway at a network level with the highest degree of accuracy. [0224] Pavement texture is defined by the irregularities on a pavement surface that deviate from a perfectly horizontal surface. Texture has been deemed a crucial characteristic of a road surface because it determines most tire/pavement interactions, such as noise, friction, and rolling resistance (3). However, pavement texture is complex, and characterization typically requires specialized equipment and mathematical tools. A linear profile is the simplest representation of pavement texture. The profile is a two-dimensional (2D) representation of the surface texture based on its distance and height and is measured using a sensor device. Profiles are considered stationary random functions of a given distance along the surface (1, 4).

[0225] In 1987, the Permanent International Association of Road Congresses (PIARC) proposed characterizing pavement surfaces using four categories: unevenness (or roughness), megatexture, macrotexture, and microtexture [5], Each category is a function of the texture amplitude and wavelengths (A) or spatial frequency (fS), given that they are related by fS=l/A [6, 7], Each of the four components influences tire/pavement interactions to varying degrees. However, two components — microtexture and macrotexture — seemingly have significant interactions with multiple other surface characteristics. A clear example of this significant interaction occurs with skid resistance. Macro and microtexture play a major role in the wetweather friction characteristics of pavement surfaces. High macrotexture provides drainage channels for the water, which allows for better contact between the pavement and the vehicle tire and thus can ultimately prevent hydroplaning [6, 8], Therefore, pavements constructed to accommodate vehicles traveling at speeds of 50 mph (80 km/h) or faster require good macrotexture. High microtexture provides tire-aggregate contact through the thin water film, resulting in higher skid resistance. This interaction makes it critical for smooth, fine mix pavements to have a high level of microtexture to provide adequate skid during wet-weather conditions [7, 9],

[0226] Macrotexture refers to the large-scale texture of the pavement surface created by the aggregate particle arrangement. Macrotexture in flexible pavements is controlled by mixture properties, such as aggregate shape, size, and gradation; in rigid pavements, macrotexture is controlled by the finishing method, i.e., tining, grooving width and spacing, and direction of the texturing. Even the state-of-the-art practice methodologies used for measuring pavement texture at highway speed typically account only for macrotexture [1, 10, 11], as the wavelengths associated with this type of texture are similar in size to tire tread elements in the tire/road interface [5], Microtexture alludes to the sub- visible or microscopic asperities of the aggregate surface, which control the contact between the tire rubber and the pavement surface [11], Microtexture is a function of the individual aggregate particle mineralogy and petrology as well as the aggregate source (natural or manufactured) and is affected by the environmental effects and the action of traffic [1], This component is responsible for making the surface feel harsh, but it is too small to be assessed by the naked eye [5],

[0227] Commercial Methods. Prior commercial methods of obtaining skid resistance have used the mechanism of dragging a material on a surface and calculating the drag force over the load to calculate the coefficient of friction. With advances in laser technology and data acquisition capabilities, the texture on the surface of a road can be captured with high accuracy at highway speeds, as disclosed herein. By collecting data with macrotexture wavelengths (50 - 0.5 mm), the robust prediction model has proven to be sufficient to estimate friction without the need for physical contact.

[0228] The exemplary system and method can be utilized for network-level studies for skid resistance allowing transportation agencies to effectively reduce expenses for monitoring the friction on the roadway network and/or expand the frequency and extent of the monitoring.

[0229] To predict friction reliably and accurately, the algorithm (e.g., decision tree classifier or other classifier described herein) can classify the surface type based on the texture information collected (e.g., open mixes, dense mixes, grades of chip seals, mix such as stone mastic asphalt (SMA), finishes of concrete, among other surfaces described herein).

[0230] British Pendulum Test (BPT), Dynamic Friction Tester (DFT), Mu-Meter, grip testers and micro-grip testers, Locked- Wheel Skid Tester (LWT), Sideway-Force Coefficient Routine Investigation Machine (SCRIM), are commercially available tests to evaluate pavement friction. BPT, DFT, and micro-grip testers are typically used for low-speed evaluation or spot-checking.

[0231] Griptesters use fixed slip mode for measuring friction experienced by vehicles with ABS braking system. LWT is the most common test and tests the frictional properties of the surface under emergency braking conditions for vehicles without anti-lock braking systems (ABS) by testing under a locked wheel mode. SCRIM is a surface friction tester commonly used in Europe to measure wet-road skidding resistance. The machine operates by applying a freely rotating fifth wheel at an angle of 20° to the direction of travel on the road surface under a known load. Controlled waterjets within the machine wet the pavement surface directly in front of the test wheel to emulate wet weather conditions.

[0232] For transportation agencies worldwide, it is of utmost importance that most highway networks have an adequate skid resistance to reduce the probability of wet-weather crashes. Nevertheless, there are three limitations that affect nearly all currently available skid measuring devices: 1) They are static devices that require traffic control to be used in the field, 2) technology used in the devices is improving, but measurement repeatability is not, and the biggest limitation of them all, 3) the in-motion equipment is highly inefficient as it requires significant volumes of water to test a few miles of highway. For instance, to use the Griptester a 200-gallon water tank must first be filled with clean water. Once the tank is filled, surveyors can measure skid resistance in terms of or the coefficient of friction as measured by the Griptester, for approximately twenty miles before having to refill the tank. As the mileage in a highway network increases, it becomes clear that continuous measurements of skid resistance for all roadways are not a sustainable operation; hence agencies often prioritize their major highway links when collecting skid measurements. This is the reason why agencies and research institutions have developed alternative techniques such as mathematical models to estimate/predict skid resistance on pavement surfaces.

[0233] Statistical Analysis. Using statistical analysis to comprehend the relationship between skid resistance and pavement characteristics is an approach to estimate pavement surface friction. Serigos et al. (2014) surveyed a total of 28 in-service flexible pavement sections within the state of Texas with multiple configurations of fine and coarse aggregate macrotexture and smooth and rough microtexture. They measured both pavement macrotexture and microtexture using the Laser Texture Scanner (LTS) device and used a wide array of texture summary statistics as predictors to estimate low-speed skid resistance measured in terms of British Pendulum Number (BPN), a unit of friction as measured by the British Pendulum.. Their results showed that using spectral parameters that incorporate microtexture can increase the predictive power of the model from 0.67 to 0.76 in R². Furthermore, they showed that using spatial parameters, such as MPD, can yield prediction values for BPN that are as good as those obtained using spectral parameters, so long as both microtexture and the type of pavement surface are accounted for.

[0234] Zuniga-Garcia (2017) surveyed a total of 36 pavements across Texas, including different surface types such as hot-mix asphalt, surface treatments, and concrete sidewalk. The friction data were collected using the BPT, the DFT, and the Micro- Griptester, whereas macro and microtexture measurements were collected using the CTM and the Line Laser Scanner (LLS). Furthermore, when characterizing the microtexture, only the active area of the profile was considered for computing the summary statistics. In this study, Zuniga-Garcia tested three different skid prediction models using panel data analysis. One model included only macrotexture information, the second model included only microtexture data, and the final model incorporated both macro and microtexture information. The main findings of this study were 1) including the surface type of the pavement when modeling friction is important as there is no unique relationship between texture and friction. The relationship between friction and texture is strong, but it varies depending on the type of surface. 2) MPD was the most significant parameter at the macro and microtexture levels out of the eight summary statistics that were tested to explain the variance in friction measurements, and lastly, 3) all skid prediction measurements obtained using the BPT and the DFT were significantly improved once both macro and microtexture data were incorporated into the model, increasing the predictive power in R² from 0.70 up to 0.80.

[0235] Rado and Kane (2014) used the Hilbert-Huang Transform to analyze texture and friction relationships and found a set of parameters calculated from basic functions of the texture profile to be highly correlated with pavement friction. More recently, Rado, Kane, and Timmons (2015) expanded on the study performed in 2014 and surveyed a total of eleven pavement surfaces in The French Institute of Science and Technology for Transport, Development and Networks test track to determine how the wavelengths, number, and shape of pavement surface asperities affect pavement friction using the empirical mode decomposition of the Hilbert-Huang Transform to decompose the texture. All their texture measurements were collected using the CTM, whereas the friction measurements were collected using the DFT at test speeds of 20, 40, and 60 kph. Once decomposed, the texture’s “sharpness” and “density” were quantified and correlated to friction. The study concluded that using the texture density solely in the models can account for up to 77% of the variability in the friction measurements of the DFT, and texture sharpness alone can account for up to 66%, but the product of density and sharpness was able to account up to 85% of the friction measurements. This study further confirmed that accounting for pavement microtexture when predicting friction can increase the predictive power of the model significantly.

[0236] Subedi, Wu and Abadie (2016) tested a total of 22 asphalt pavement sections across 15 parishes in the state of Louisiana. They collected friction measurements using the LWT and the DFT and collected texture data on the pavements using the CTM. The pavements they tested consisted of 12.5-mm Superpave, 19-mm Superpave, stone mastic asphalt and open graded friction course, the most common mixes present in Louisiana. Furthermore, once the researchers knew what aggregates were used in each of the test sections, they obtained a blend PSV based on the PSV of the aggregates used in the mix. This information, plus the design lane ADT, ADT growth factor, and other gradation parameters, were used to estimate the MPD of the road, SN at 40 kph using either a smooth (SN40S) or ribbed (SN40R) test tire on the LWT, and the low- speed skid resistance from the DFT20, which was used as a surrogate for microtexture. The results of their research are as follows: 1) while the prediction values obtained for MPD showed some discrepancies, it is thought that the predictions were only able to capture most of the field macrotexture component but none of the microtexture, 2) one can estimate DFT20 with an R² of 0.88 using solely the PSV of the aggregates and traffic information, 3) incorporating macrotexture (predicted MPD) and microtexture (DFT20) into a non-linear prediction model for skid resistance using smooth tires yielded a relationship that can account for 73% of the variability in the SN40S measurements, and lastly 4) by combining the predicted MPD and predicted SN40S, one can properly predict SN40R with an R² of 0.75. This study showed that incorporating mixed design parameters and traffic information into the prediction model can be used as an alternative to avoid taking texture measurements of the network and still provide a good enough prediction of pavement skid resistance.

[0237] Example Computing System

[0238] It should be appreciated that the logical operations described above for the analysis system and in the appendix can be implemented (1) as a sequence of computer-implemented acts or program modules running on a computing system and/or (2) as interconnected machine logic circuits or circuit modules within the computing system. The implementation is a matter of choice dependent on the performance and other requirements of the computing system. Accordingly, the logical operations described herein are referred to variously as state operations, acts, or modules. These operations, acts, and/or modules can be implemented in software, in firmware, in special purpose digital logic, in hardware, and any combination thereof. It should also be appreciated that more or fewer operations can be performed than shown in the figures and described herein. These operations can also be performed in a different order than those described herein.

[0239] The computer system is capable of executing the software components described herein for the exemplary method or systems. In an embodiment, the computing device may comprise two or more computers in communication with each other that collaborate to perform a task. For example, but not by way of limitation, an application may be partitioned in such a way as to permit concurrent and/or parallel processing of the instructions of the application.

Alternatively, the data processed by the application may be partitioned in such a way as to permit concurrent and/or parallel processing of different portions of a data set by the two or more computers. In an embodiment, virtualization software may be employed by the computing device to provide the functionality of a number of servers that are not directly bound to the number of computers in the computing device. For example, virtualization software may provide twenty virtual servers on four physical computers. In an embodiment, the functionality disclosed above may be provided by executing the application and/or applications in a cloud computing environment. Cloud computing may comprise providing computing services via a network connection using dynamically scalable computing resources. Cloud computing may be supported, at least in part, by virtualization software. A cloud computing environment may be established by an enterprise and/or can be hired on an as-needed basis from a third-party provider. Some cloud computing environments may comprise cloud computing resources owned and operated by the enterprise as well as cloud computing resources hired and/or leased from a third-party provider.

[0240] In its most basic configuration, a computing device includes at least one processing unit and system memory. Depending on the exact configuration and type of computing device, system memory may be volatile (such as random-access memory (RAM)), non-volatile (such as read-only memory (ROM), flash memory, etc.), or some combination of the two.

[0241] The processing unit may be a standard programmable processor that performs arithmetic and logic operations necessary for the operation of the computing device. While only one processing unit is shown, multiple processors may be present. As used herein, processing unit and processor refers to a physical hardware device that executes encoded instructions for performing functions on inputs and creating outputs, including, for example, but not limited to, microprocessors (MCUs), microcontrollers, graphical processing units (GPUs), and applicationspecific circuits (ASICs). Thus, while instructions may be discussed as executed by a processor, the instructions may be executed simultaneously, serially, or otherwise executed by one or multiple processors. The computing device may also include a bus or other communication mechanism for communicating information among various components of the computing device. [0242] Computing devices may have additional features/functionality. For example, the computing device may include additional storage such as removable storage and non-removable storage including, but not limited to, magnetic or optical disks or tapes. Computing devices may also contain network connection(s) that allow the device to communicate with other devices, such as over the communication pathways described herein. The network connection(s) may take the form of modems, modem banks, Ethernet cards, universal serial bus (USB) interface cards, serial interfaces, token ring cards, fiber distributed data interface (FDDI) cards, wireless local area network (WLAN) cards, radio transceiver cards such as code division multiple access (CDMA), global system for mobile communications (GSM), long-term evolution (LIE), worldwide interoperability for microwave access (WiMAX), and/or other air interface protocol radio transceiver cards, and other well-known network devices. Computing devices may also have input device(s) such as keyboards, keypads, switches, dials, mice, trackballs, touch screens, voice recognizers, card readers, paper tape readers, or other well-known input devices. Output device(s) such as printers, video monitors, liquid crystal displays (LCDs), touch screen displays, displays, speakers, etc., may also be included. The additional devices may be connected to the bus in order to facilitate the communication of data among the components of the computing device. All these devices are well known in the art and need not be discussed at length here.

[0243] The processing unit may be configured to execute program code encoded in tangible, computer-readable media. Tangible, computer-readable media refers to any media that is capable of providing data that causes the computing device (i.e., a machine) to operate in a particular fashion. Various computer-readable media may be utilized to provide instructions to the processing unit for execution. Example tangible, computer-readable media may include but is not limited to volatile media, non-volatile media, removable media, and non-removable media implemented in any method or technology for storage of information such as computer-readable instructions, data structures, program modules, or other data. System memory, removable storage, and non-removable storage are all examples of tangible computer storage media. Example tangible, computer-readable recording media include, but are not limited to, an integrated circuit (e.g., field-programmable gate array or application-specific IC), a hard disk, an optical disk, a magneto-optical disk, a floppy disk, a magnetic tape, a holographic storage medium, a solid-state device, RAM, ROM, electrically erasable program read-only memory (EEPROM), flash memory or other memory technology, CD-ROM, digital versatile disks (DVD) or other optical storage, magnetic cassettes, magnetic tape, magnetic disk storage or other magnetic storage devices.

[0244] In light of the above, it should be appreciated that many types of physical transformations take place in the computer architecture in order to store and execute the software components presented herein. It also should be appreciated that the computer architecture may include other types of computing devices, including hand-held computers, embedded computer systems, personal digital assistants, and other types of computing devices known to those skilled in the art.

[0245] In an example implementation, the processing unit may execute program code stored in the system memory. For example, the bus may carry data to the system memory, from which the processing unit receives and executes instructions. The data received by the system memory may optionally be stored on the removable storage or the non-removable storage before or after execution by the processing unit.

[0246] It should be understood that the various techniques described herein may be implemented in connection with hardware or software or, where appropriate, with a combination thereof. Thus, the methods and apparatuses of the presently disclosed subject matter, or certain aspects or portions thereof, may take the form of program code (i.e., instructions) embodied in tangible media, such as floppy diskettes, CD-ROMs, hard drives, or any other machine-readable storage medium wherein, when the program code is loaded into and executed by a machine, such as a computing device, the machine becomes an apparatus for practicing the presently disclosed subject matter. In the case of program code execution on programmable computers, the computing device generally includes a processor, a storage medium readable by the processor (including volatile and non-volatile memory and/or storage elements), at least one input device, and at least one output device. One or more programs may implement or utilize the processes described in connection with the presently disclosed subject matter, e.g., through the use of an application programming interface (API), reusable controls, or the like. Such programs may be implemented in a high-level procedural or object-oriented programming language to communicate with a computer system. However, the program(s) can be implemented in assembly or machine language, if desired. In any case, the language may be a compiled or interpreted language, and it may be combined with hardware implementations. [0247] Although example embodiments of the present disclosure are explained in some instances in detail herein, it is to be understood that other embodiments are contemplated. Accordingly, it is not intended that the present disclosure be limited in its scope to the details of construction and arrangement of components set forth in the following description or illustrated in the drawings. The present disclosure is capable of other embodiments and of being practiced or carried out in various ways.

[0248] It must also be noted that, as used in the specification and the appended claims, the singular forms “a,” “an,” and “the” include plural referents unless the context clearly dictates otherwise. Ranges may be expressed herein as from “about” or “approximately” one particular value and/or to “about” or “approximately” another particular value. When such a range is expressed, other exemplary embodiments include from the one particular value and/or to the other particular value.

[0249] By “comprising” or “containing” or “including” is meant that at least the name compound, element, particle, or method step is present in the composition or article or method, but does not exclude the presence of other compounds, materials, particles, method steps, even if the other such compounds, material, particles, method steps have the same function as what is named.

[0250] In describing example embodiments, terminology will be resorted to for the sake of clarity. It is intended that each term contemplates its broadest meaning as understood by those skilled in the art and includes all technical equivalents that operate in a similar manner to accomplish a similar purpose. It is also to be understood that the mention of one or more steps of a method does not preclude the presence of additional method steps or intervening method steps between those steps expressly identified. Steps of a method may be performed in a different order than those described herein without departing from the scope of the present disclosure. Similarly, it is also to be understood that the mention of one or more components in a device or system does not preclude the presence of additional components or intervening components between those components expressly identified.

[0251] The term “about,” as used herein, means approximately, in the region of, roughly, or around. When the term “about” is used in conjunction with a numerical range, it modifies that range by extending the boundaries above and below the numerical values set forth. In general, the term “about” is used herein to modify a numerical value above and below the stated value by a variance of 10%. In one aspect, the term “about” means plus or minus 10% of the numerical value of the number with which it is being used. Therefore, about 50% means in the range of 45%-55%. Numerical ranges recited herein by endpoints include all numbers and fractions subsumed within that range (e.g., 1 to 5 includes 1, 1.5, 2, 2.75, 3, 3.90, 4, 4.24, and 5).

[0252] Similarly, numerical ranges recited herein by endpoints include subranges subsumed within that range (e.g., 1 to 5 includes 1-1.5, 1.5-2, 2-2.75, 2.75-3, 3-3.90, 3.90-4, 4-4.24, 4.24-5, 2-5, 3-5, 1-4, and 2-4). It is also to be understood that all numbers and fractions thereof are presumed to be modified by the term “about.”

[0253] The following patents, applications, and publications, as listed below and throughout this document, are hereby incorporated by reference in their entirety herein.

[1] Zuniga-Garcia, N. Predicting Friction with Improved Texture Characterization [Master’s Thesis], The University of Texas at Austin; 2017

[2] Smit, A., Trevino, M., Zuniga-Garcia, N., Buddhavarapu, P., Prozzi, J. Selection and Design of Quiet Pavement Surfaces. Center for Highway Pavement Preservation (CHPP). Technical Report No: 0-6819-1. 2016.

[3] Maguire, D., Carme, C. LOUDNESS tells you nothing. Ishn. 2015; 49(8):34.

[4] Sandberg, U. Road traffic noise, The influence of the road surface and its characterization. Applied Acoustics. 1987; 21(2):97-118.

[5] Permanent International Association of Road Congresses (PIARC). Permanent International Association of Road Congresses Technical Committee on Surface Characteristics: Technical Committee Report No 1 to the XVIII World Road Congress. Brussels: PIARC; 1987.

[6] Hoerner, T. E.; Smith, K. D. High Performance Concrete Pavement: Pavement Texturing and Tire-pavement Noise. Federal Highway Administration; Applied Pavement Technology, Incorporated, 2002, 40 p.

[7] Serigos, P. Bayesian Estimation of Finite Mixture Roughness Model [Ph.D.]. University of Texas 16 at Austin; 2016.

[8] Fontes, L., Pereira, P., Pais, J., Teiches, G. Improvement of the Functional Pavement Quality with Asphalt Rubber Mixtures. 2006.

[9] Hall, J., Smith, K., Wambold, J., Yager, T., Rado, Z. Guide for Pavement Friction. Washington DC: National Cooperative Highway Research Program (NCHRP); 2009. [10] AASHTO. Guide for pavement friction. 1st ed. Washington D.C.: American Association of State Highway Transportation Officials; 2008.

[11] Serigos, P., Buddhavarapu, P., Gorman, G, Hong, F., Prozzi, J. The Contribution of Micro and Macro-Texture to the Skid Resistance of Flexible Pavement. Center for Transportation Research; 2016.

[12] FHWA. Evaluation of Pavement Safety Performance. U. S. Department of Transportation, FHWA-HRT- 14-065, 2015.

[13] Chang, C., Saenz, D., Abdallah, I., Wimsatt, A., Freeman, T., and Fernando, E. TxDOT Guidelines to Assign Treatment Levels. Report: FHWA/TX- 14/0-6673 -Pl. Center for Transportation Infrastructure Systems, The University of Texas at El Paso, 2014.

[14] Asphalt Concrete, Asphalt Pavement Association of New Mexico. https://apanm.org/asphalt-concrete/. Accessed Jun. 21, 2021.

[15] National Asphalt Pavement Association (NAPA). HMA Pavement Mix Type Selection Guide. Washington DC: U.S. Department of Transportation; 2001.

[16] International Organization for Standardization (ISO). Characterization of Pavement Texture by Use of Surface Profiles - Part 4: Spectral Analysis of Surface Profiles. Geneve; 2008.

[17] Sabillon-Orellana, C. Evaluation of data imputation techniques in pavement texture processing [Master’s Thesis], University of Texas at Austin; 2020.

[18] Jakobsen, J., Gluud, C., Wetterslev, J., Winkel, P. When and how should multiple imputation be used for handling missing data in randomised clinical trials, a practical guide with flowcharts. BMC Medical Research Methodology. 2017; 17(1).

[19] Kang, H. The prevention and handling of the missing data. Korean Journal of Anesthesiology. 43 2013; 64(5):402.

[20] International Organisation for Standardisation (ISO). Geometrical Product Specifications, Surface texture: Profile method, Terms, definitions, surface texture parameters. Geneve; 2008.

[21] Gadelmawla, E., Koura, M., Maksoud, T., Elewa, I., Soliman, H. Roughness parameters. Journal 47 of Materials Processing Technology. 2002; 123(1): 133-145.

[22] Piryonesi, S., El-Diraby, T. Data Analytics in Asset Management: Cost-Effective Prediction of the Pavement Condition Index. Journal of Infrastructure Systems. 2020; 26(l):04019036.

Second Reference List

[1] Acheson, D.J. (2009) Elementary Fluid Dynamics. Oxford: Clarendon Press. [2] Addison, P.S. (2017) The Illustrated Wavelet Transform Handbook: Introductory Theory and

[3] Applications in Science, Engineering, Medicine, and Finance. CRC Press, Boca Raton.

[4] Afshar, T. et al. (2017) “Impact of particle shape on breakage of recycled construction and demolition aggregates,” Powder Technology, 308, pp. 1-12.

[5] Aggarwal, Charu C. (2018) Neural Networks and Deep Learning: a Textbook / by Charu C.

[6] Aggarwal. Ist ed. Cham: Springer International Publishing, Web.

[7] Ahmed, M., Haas, C. and Haas, R. (2011) “Toward Low-Cost 3D Automatic Pavement Distress Surveying: The Close Range Photogrammetry Approach,” Canadian Journal of Civil Engineering, 38, pp. 1301-1313.

[8] Agresti, A. (2015). Foundations of Linear and Generalized Linear Models. New York: Wiley, pp. 183-186.

[9] American Association of State Highway and Transportation Officials (AASHTO). (2008). AASHTO Guide for Pavement Friction, first ed., Washington DC.

[10] American Concrete Pavement Association (ACPA), (n.d.). “Surface Texture.” American Concrete Pavement Association, http://metiebar.acpa.org/Concrete_Pavement/Technical/Fundamentals/Surface_Texture.asp.

[11] American Society for Testing and Materials (ASTM). (2021). ASTM E178-21 Standard Practice for Dealing with Outlying Observations. West Conshohocken, PA.

[12] American Society of Mechanical Engineers (ASME). (2020). ASME B46.1 Surface texture: Surface roughness, waviness, and lay. New York City: New York.

[13] American Society for Testing and Materials (ASTM). (2015). ASTM E1845-15 Standard Practice for Calculating Pavement Macrotexture Mean Profile Depth. West Conshohocken, PA. American Society for Testing and Materials (ASTM). (2019). ASTM E2380/E2380M-15 Standard Test Method for Measuring Pavement Texture Drainage Using an Outflow Meter. West Conshohocken, PA.

[14] American Society for Testing and Materials (ASTM). (2019). ASTM E2157-15 Standard Test Method for Measuring Pavement Macrotexture Properties Using the Circular Track Meter. West Conshohocken, PA.

[15] Bagheri, GH. et al. (2015) “On the characterization of size and shape of irregular particles,” Powder Technology, 270, pp. 141-153. Available at: https://doi.Org/10.1016/j.powtec.2014.10.015. [16] Bateman, D. (2016). Design Guide for Road Surface Dressing - Road Note 39 (Issue 7).

TRL Limited.

[17] Benjamini, Y. and Hochberg, Y. (1995) “Controlling the false discovery rate: A practical and powerful approach to multiple testing,” Journal of the Royal Statistical Society: Series B (Methodological), 57(1), pp. 289-300.

[18] Bitelli, G. et al. (2012) “Laser scanning on road pavements: A new approach for characterizing surface texture,” Sensors, 12(7), pp. 9110-9128.

[19] Breiman, L. (1996) “Bagging predictors,” Machine Learning, 24(2), pp. 123-140. Available at: https : //do i . org/ 10.1007/bf00058655.

[20] Brown, E. R. (1993). Experience with stone matrix asphalt in the United States (No. Report No. 9). National Center for Asphalt Technology (US).

[21] Brown, E. R., Haddock, J. E., Mallick, R. B., & Lynn, T. A. (1997). Development of a mixture design procedure for stone matrix asphalt.

[22] Chamberlin, W.P. and Amsler, D.E. (1982). "Measuring Surface Texture by the Sand-Patch Method," Pavement Surface Characteristics and Materials, ASTM STP 763. C. M. Hayden, Ed., American Society for Testing and Materials, pp. 3-15.

[23] Chen, S. et al. (2022) “A State-of-the-Art Review of Asphalt Pavement Surface Texture and its Measurement Techniques,” Journal of Road Engineering, 2(2), pp. 156-180.

[24] Chicco D., Toetsch N., and Jurman G. (2021). "The Matthews correlation coefficient (MCC) is more reliable than balanced accuracy, bookmaker informedness, and markedness in two-class confusion matrix evaluation". BioData Mining. 14 (13): 1-22.

[25] Choromanska, A. and Monteleoni, C., (2012). Online Clustering with Experts. Journal of Machine Learning Research, 26.

[26] Colebrook, C.F. (1939) “Turbulent flow in pipes, with particular reference to the transition region between the smooth and rough pipe laws.,” Journal of the Institution of Civil Engineers, 11(4), pp. 133-156.

[27] Cooley, L., Ray, E., & Watson, D. E. (2000). Evaluation of open-graded friction course mixtures containing cellulose fibers. Transportation Research Record, 1723(1), 19-25.

Cortes, C. and Vapnik, V. (1995) “Support-Vector Networks,” Machine Learning, 20(3), pp. 273-297. [28] Cover, T.M.; Hart, P. E. (1967). "Nearest Neighbor Pattern Classification". IEEE Transactions on Information Theory. 13 (1): 21-27. CiteSeerX 10.1.1.68.2616.

[29] Dagnail, H. (1980) Exploring surface texture. Leicester: Rank Taylor Hobson.

[30] Darcy, H. (1865) Recherches experimentales relatives au Mouvement de 1'eau dans les tuyaux [Experimental Research Relating to the Movement of Water in Pipes], Paris: Imprimerie nationale.

[31] Deng, H. et al. (2021) “Ensemble learning for the early prediction of neonatal jaundice with genetic features,” BMC Medical Informatics and Decision Making, 21(1). Available at: https: //doi. org/10.1186/s 12911 -021 -01701 -9.

[32] Devore, J.L. (2009) “Chapter 1 / Boxplots that Show Outliers,” in Probability and statistics for engineering and the Sciences. Belmont, CA: Brooks/Cole, pp. 40-41.

[33] Deutsches Institut fur Normung (DIN) [German Institute for Standardization], (1982). DIN 4760: 1982-06 Gestaltabweichungen; Begriffe, Ordnungssystem [Form Deviations; Concepts; Classification System], Berlin, Germany.

[34] Diaspro, A. (2005) “Review of ‘Fractals and Chaos: The Mandelbrot Set and Beyond’, by B. Mandelbrot,” BioMedical Engineering OnLine, 4(1).

[35] Donders, A.R. et al. (2006) “Review: A gentle introduction to imputation of missing values,” Journal of Clinical Epidemiology, 59(10), pp. 1087-1091.

[36] Donoho, D.L. and Johnstone, I.M. (1994) “Ideal spatial adaptation by Wavelet Shrinkage,” Biometrika, 81(3), pp. 425-455.

[37] Doty, R. N. (1975). "Study of the Sand Patch and Outflow Meter Methods of Pavement Surface Texture Measurement," Surface Texture Versus Skidding: Measurements, Frictional Aspects, and Safety Features of Tire-Pavement Interactions, ASTM STP 583, American Society for Testing and Materials, pp. 42-61.

[38] Diirig, T. et al. (2011) “Comparative analyses of glass fragments from brittle fracture experiments and volcanic ash particles,” Bulletin of Volcanology, 74(3), pp. 691-704.

Duris, M. et al. (2016) “Optimal pixel resolution for sand particles size and shape analysis,” Powder Technology, 302, pp. 177-186.

[39] El Gendy, A. et al. (2011) “Stereo- vision applications to reconstruct the 3D texture of pavement surface,” International Journal of Pavement Engineering, 12(3), pp. 263-273. El Hachem, Y. and Prozzi, J. A. (2019) The Utexas seal coat design method using 3D Laser Technology, thesis. University of Texas at Austin.

[40] Epps, J. A., Gallaway, B. M., & Hughes, C. H. (1981). Field Manual on Design and Construction of Seal Coats. Research Report No. 214-25. College Station: Texas A&M Transportation Institute.

[41] Fanning, J.T. (1887) A practical treatise on hydraulic and water-supply engineering: Relating to the hydrology, hydrodynamics, and practical construction of water-works, in North America. New York: Van Nostrand.

[42] Fawcett, Tom (2006). "An Introduction to ROC Analysis" (PDF). Pattern Recognition Letters. 27 (8): 861-874.

[43] Federal Aviation Administration (FAA). (1997). Measurement, Construction, and Maintenance of Skid-Resistant Airport Pavement Surfaces. Washington DC, District of Columbia.

[44] Federal Highway Administration (FHWA). (2005). Surface Texture for Asphalt and Concrete Pavements. TA 5040.36. Federal Highway Administration, Washington, DC.

[45] Federal Highway Administration (FHWA). (2022). High Friction Surface Treatments (HFST). Federal Highway Administration, Washington, DC

[46] Fernando, E.G., Hu, S. and Crockford, W. (2022) “Comparative evaluation of pavement macrotexture measurements from different devices,” Transportation Research Record: Journal of the Transportation Research Board.

[47] Fix, E.; Hodges, J.L. (1951). “Discriminatory Analysis. Nonparametric Discrimination: Consistency Properties”. USAF School of Aviation Medicine, Randolph Field, Texas.

[48] Flintsch, G., Huang, M. and McGhee, K. (2005) “Harmonization of macrotexture measuring devices,” Journal of ASTM International, 2(9), p. 13043.

[49] Flintsch, G., Viner, H. and Dunford, A. (2014) Splash and Spray and its Impact on Drivers, rep. Safer Roads International Conference.

[50] Freund, Y. and Schapire, R.E. (1995) “A desicion-theoretic generalization of on-line learning and an application to boosting,” Lecture Notes in Computer Science, pp. 23-37.

[51] Gadelmawla, E., Koura, M., Maksoud, T., Elewa, I. and Soliman, H., (2002). Roughness parameters. Journal of Materials Processing Technology, 123(1), pp.133-145. [52] Gunaratne, M. et al. (2000) “Correlation of Tire Wear and Friction to Texture of Concrete Pavements,” Journal of Materials in Civil Engineering, 12(1), pp. 46-54.

[53] Guo, R. et al. (2020) “Degradation state recognition of piston pump based on ICEEMDAN and XGBoost,” Applied Sciences, 10(18), p. 6593. Available at: https://doi.org/10.3390/appl0186593.

[54] Habig, K.H. (1993) Tribologisches Verhalten Keramischer Werkstoffe [Tribological Behavior of Ceramic Materials], Ehningen, Baden-Wurttemberg: Expert Verlag.

[55] Hand, D.J. and Yu, K. (2001) “Idiot's bayes: Not so stupid after all?,” International Statistical Review / Revue Internationale de Statist! que, 69(3), p. 385. Available at: https://doi.org/10.2307/1403452.

[56] Hanson, D. and Waller, B. (2005) Evaluation of the Noise Characteristics of Minnesota Pavements, rep. Minnesota Department of Transportation.

[57] Hanson, F. (1935). Bituminous Surface Treatment of Rural Highways. Wellington: New Zealand Society of Civil Engineers.

[58] Harrison, O. (2019) Machine learning basics with the K-nearest neighbors algorithm, Medium. Towards Data Science. Available at: https://towardsdatascience.com/machine-learning- basics-with-the-k-nearest-neighbors-algorithm-6a6e71d01761 (Accessed: February 13, 2023).

[59] Hastie, T. et al. (2009) “Multi-class AdaBoost,” Statistics and Its Interface, 2(3), pp. 349- 360. Available at: https://doi.org/10.4310/sii.2009.v2.n3.a8.

[60] Hastie, T.; Tibshirani, R.; Friedman, J.H. (2009). The Elements of Statistical Learning: Data Mining, Inference, and Prediction (2nd ed.).

[61] Henry, J. J. (2000). Evaluation of Pavement Friction Characteristics. NCHRP Synthesis 291. National Cooperative Highway Research Program, Washington, D.C.

[62] Hibbs, B. and Larson, R. (1996) Tire Pavement Noise and Safety Performance, PCC Surface Texture Technical Working Group, rep. Washington DC, District of Columbia: Federal Highway Administration (FHWA).

[63] Ho, T.K. (1998) “The random subspace method for constructing decision forests,” IEEE Transactions on Pattern Analysis and Machine Intelligence, 20(8), pp. 832-844.

[64] Ho, T.K. (1995). Random Decision Forests. Proceedings of the 3rd International Conference on Document Analysis and Recognition, Montreal, QC. Pp. 278-282. [65] Holzschuher, C. (n.d.) High Friction Surface Treatment Guidelines: Project Selection, Materials, and Construction, rep. Gainsville, Florida: Florida Department of Transportation (FDOT).

[66] Hornik, K., Stinchcombe, M. and White, H. (1989) “Multilayer feedforward networks are universal approximators,” Neural Networks, 2(5), pp. 359-366.

[67] Huang, H. and Pan, J. (2006) “Speech Pitch Determination based on Hilbert-Huang Transform,” Signal Processing, 86(4), pp. 792-803.

[68] Hutchings, I.M. and Shipway, P. (2017) “Chapters 2: Surface Topography and Surfaces in Contact,” in Tribology: Friction and wear of Engineering Materials. Amsterdam: Butterworth- Heinemann.

[69] International Organisation for Standardisation (ISO). (2008). ISO 13473-4 Characterization of Pavement Texture by Use of Surface Profiles - Part 4: Spectral Analysis of Surface Profiles. Geneve, Switzerland.

[70] International Organisation for Standardisation (ISO). (2009). ISO 13473-5 Characterization of Pavement Texture by Use of Surface Profiles - Part 5: Determination of Megatexture. Geneve, Switzerland.

[71] International Organisation for Standardisation (ISO). (2021). ISO 21920-2 Geometrical Product Specifications (GPS) — Surface Texture: Profile — Part 2: Terms, Definitions and Surface Texture Parameters. Geneve, Switzerland.

[72] Inzerillo, L., Di Mino, G. and Roberts, R. (2018) “Image-Based 3D Reconstruction Using Traditional and UAV Datasets for Analysis of Road Pavement Distress,” Automation in Construction, 96, pp. 457-469.

[73] Jorgensen, B. (1997). The Theory of Dispersion Models. Chapman & Hall.

Kandlikar, S.G et al. (2005) “Characterization of Surface Roughness Effects on Pressure Drop in Single-Phase Flow in Minichannels,” Physics of Fluids, 17(10), p. 100606.

[74] Kane, M., Rado, Z. and Timmons, A. (2014) “Exploring the Texture-friction Relationship: From Texture Empirical Decomposition to Pavement Friction,” International Journal of Pavement Engineering, 16(10), pp. 919-928.

[75] Katicha, S., D. Mogrovejo, G. Flintsch, and E. de Leon Izeppi. (2014). Latest Development in the Processing of Pavement Macrotexture Measurements of High-Speed Laser Devices. Proc., Ninth International Conference on Managing Pavement Assets, Alexandria, Va. [76] Katicha, S.W. et al. (2015) “Adaptive Spike Removal Method for High-Speed Pavement Macrotexture Measurements by Controlling the False Discovery Rate,” Transportation Research Record: Journal of the Transportation Research Board, 2525(1), pp. 100-110.

[77] Larhmam. (2022) Support Vector Machine, Wikipedia. Wikimedia Foundation. Available at: https://en.wikipedia.org/wiki/Support_vector_machine (Accessed: February 13, 2023).

[78] Leung, K. (2022) Micro, Macro & Weighted Averages of Fl Score, Clearly Explained, Medium. Towards Data Science. Available at: https://towardsdatascience.com/micro-macro- weighted-averages-of-fl-score-clearly-explained-b603420b292f (Accessed: February 13, 2023).

[79] Liu, E.J., Cashman, K.V. and Rust, A.C. (2015) “Optimising Shape Analysis to Quantify Volcanic Ash Morphology,” GeoResJ, 8, pp. 14-30. Available at: https://doi.Org/10.1016/j.grj.2015.09.001.

[80] Luo, W., Liu, L. and Li, L. (2020) “Measuring Rutting Dimension and Lateral Position using 3D Line Scanning Laser and Inertial Measuring Unit,” Automation in Construction, 111, p. 103056.

[81] Mandelbrot, B.B. and Wheeler, J.A. (1983) “The Fractal Geometry of Nature,” American Journal of Physics, 51(3), pp. 286-287.

[82] Mang, T., Bobzin, K. and Bartels, T. (2011) Industrial Tribology: Tribosystems, Friction, Wear and Surface Engineering, Lubrication. Weinheim: Wiley-VCH.

[83] Martin, T. (2023) Receiver Operating Characteristic, Wikipedia. Wikimedia Foundation. Available at: https://en.wikipedia.org/wiki/Receiver_operating_characteristic (Accessed: February 13, 2023).

[84] Masad, E. et al. (2007) “Test methods for characterizing aggregate shape, texture, and angularity,” National Cooperative Highway Research Program.

[85] McLeod, N. W. (1987). Design of dense graded asphalt concrete pavements. In Proceedings, Canadian Technical Asphalt Association (Vol. 32, p. 352).

[86] Miller, T., Swiertz, D., Tashman, L., Tabatabaee, N. and Bahia, H. (2011). Characterization of Asphalt Pavement Surface Texture. Madison: Transportation Research Board.

[87] Mitchell, T. M. (1997). Machine learning. MacGraw-Hill.

Moody, L.F. (1944) “Friction factors for pipe flow,” Journal of Fluids Engineering, 66(8), pp. 671-678. [88] Mora, C.F. and Kwan, A.K.H. (2000) “Sphericity, shape factor, and convexity measurement of coarse aggregate for concrete using digital image processing,” Cement and Concrete Research, 30(3), pp. 351-358.

[89] Mouaket, I., Sinha, K., & White, T. (1992). Guidelines for Management of Chip and Sand Seal Coating Activity in Indiana. Washington D.C.: Transportation Research Record: Journal of the Transportation Engineering Board, No. 1507, pp. 81-90, National Research Council.

[90] Mulenga, F.K. (2017) “Sensitivity analysis of Austin's scale-up model for tumbling ball mills — part 2. effects of full-scale milling parameters,” Powder Technology, 317, pp. 6-12.

[91] Nason, G.P. (2006) “Stationary and Non-stationary Time Series,” in Statistics in Volcanology. London: Geological Society, pp. 129-142.

[92] Nielsen, F., (2016). Introduction to HPC with MPI for Data Science. Bern: Springer International Publishing.

[93] Nikuradse, J. (1937) Laws of flow in rough pipes. Washington: National Advisory Committee for Aeronautics.

[94] Ongel, A. et al. (2008) Investigation of Noise, Durability, Permeability, and Friction Performance Trends for Asphaltic Pavement Surface Types: First- and Second-Year Results, rep. Caltrans Division of Research and Innovation.

[95] O'Shea, Keiron, and Ryan Nash. (2015) An Introduction to Convolutional Neural Networks. Ithaca: Cornell University Library, arXiv.org, ProQuest. Web. 10 Nov. 2022.

[96] Paterson, W.D. (1986) “International Roughness Index: Relationship to Other Measures of Roughness and Riding Quality,” Transportation Research Record, 1084.

[97] Perera, R.W. and Kohn, S.D. (2002) Issues in Pavement Smoothness: A Summary Report, rep. Plymouth, Michigan: National Cooperative Highway Research Program (NCHRP).

[98] Perera R.W., M; Orthmeyer, R. De Leon Izeppi, E. (2013) AFD 90, Group III Texture Measurements and Use.

[99] PIARC, (1987). PIARC Technical Committee on Surface Characteristics: Technical Committee Report No 1 to the XVIII World Road Congress, Brussels, Belgium.

[100] PIARC, (1995) International PIARC Experiment to Compare and Harmonize Texture and Skid Resistance Measurements. PIARC World Road Association, Report of PIARC Technical Committee on Surface Characteristics, Paris and Madrid. [101] Piryonesi M.; El-Diraby T. (2020). "Data Analytics in Asset Management: Cost-Effective Prediction of the Pavement Condition Index". Journal of Infrastructure Systems. 26 (1).

[102] Powers, D. (2011). "Evaluation: From Precision, Recall and F-Measure to ROC, Informedness, Markedness & Correlation". Journal of Machine Learning Technologies. 2 (1): 37-63.

[103] Rajaei, S., Chatti, K., and Dargazany, R. (2017). A Review: Pavement Surface Microtexture and its Contribution to Surface Friction. Transportation Research Board 96th Annual Meeting, Paper 17-06773.

[104] Rao, S. et al. (1999) “Longevity of diamond-ground concrete pavements,” Transportation Research Record: Journal of the Transportation Research Board, 1684(1), pp. 128-136.

[105] Rasmussen, R. et al. (2004). Concrete Pavement Solutions for Reducing Tire-Pavement Noise. National Concrete Pavement Technology Center, Iowa State University, Ames, IA.

[106] Rasmussen, R. et al. (2012) How to Reduce Tire-Pavement Noise: Better Practices for Constructing and Texturing Concrete Pavement Surfaces, rep. National Concrete Pavement Technology Center.

[107] Robbins, M. and Tran, N. (2015) Literature Review: The Impact of Pavement Roughness on Vehicle Operating Costs. Auburn, Alabama: National Center for Asphalt Technology. NCAT Report 15-02

[108] Rokach, L., and Oded M..(2005.) "Clustering methods." Data mining and knowledge discovery handbook. Springer US, 321-352.

[109] Rubin, D.B. (1976) “Inference and Missing Data ,” Biometrika, 63(3), pp. 581-592.

Rungruangvirojn, P. and Kanitpong, K. (2010) “Measurement of visibility loss due to splash and spray: Porous, SMA and conventional asphalt pavements,” International Journal of Pavement Engineering, 11 (6), pp. 499-510.

[110] Russell, S.J. and Norvig, P. (2009) Artificial Intelligence: A Modern Approach. Upper Saddle River, NJ: Pearson Education.

[111] Sabillon, C.A. et al. (2022) “Classification of flexible pavements based on texture data,” Transportation Research Record: Journal of the Transportation Research Board, 2676(12), pp. 140-151. [112] Sabillon, C. et al. (2023) Development of an Efficient Prediction Model of Highway Friction on Texas Highway Network, rep. Austin, Texas: Texas Department of Transportation (TxDOT).

[113] Sabillon-Orellana, C. (2020). Evaluation of data imputation techniques in pavement texture processing [Master Thesis], University of Texas at Austin.

[114] Sabillon-Orellana, C. et al. (2021) Evaluation of the Relationship between Friction, Texture, and Noise Properties of Preservation Treatments, rep. Austin, Texas: Center for Highway Pavement Preservation (CHPP).

[115] Sagi, O. and Rokach, L. (2021) “Approximating XGBoost with an interpretable decision tree,” Information Sciences, 572, pp. 522-542. Available at: https://doi.Org/10.1016/j.ins.2021.05.055.

[116] Sandberg, U. (1987), "Road traffic noise — The influence of the road surface and its characterization", Applied Acoustics, vol. 21, no. 2, pp. 97-118.

[117] Sandberg, U., and Descornet, G. (1980). Road Surface Influence on Tyre/Road Noise - Parts I and II. Proceedings of Inter-Noise 80, Miami, Florida, pages 259-272.

[118] Sandberg, U. and Ejsmont, J. A. (2002) Tyre/Road Noise: Reference Book. Kisa, Sweden: Informex Ejsmont & Sandberg Handelsbolag.

[119] Sayers, M.W. and Karamihas, S.M. (1996) Interpretation of Road Roughness Profile Data, rep. Ann Arbor, Michigan: Federal Highway Administration (FHWA).

[120] Sayers, M. and Karamihas, S. (1998). The Little Book of Profiling. New York, NY: Regent of the University of Michigan.

[121] Schmitt, D.J. and Kandlikar, S.G (2005) “Effects of repeating microstructures on pressure drop in rectangular minichannels,” ASME 3rd International Conference on Microchannels and Minichannels, Parts A and B.

[122] Scofield, L. (2012) Development and Implementation of the Next Generation Concrete Surface, rep. American Concrete Pavement Association (ACPA), Skokie, IL

[123] Serigos, P., De Fortier Smit, A. and Prozzi, J. (2014). Incorporating Surface Microtexture in the Prediction of Skid Resistance of Flexible Pavements. Transportation Research Record: Journal of the Transportation Research Board, 2457(1), pp.105-113. [124] Serigos, P., Buddhavarapu, P., Gorman, G., Hong, F. and Prozzi, J. (2016). The contribution of Micro and Macro-Texture to the Skid Resistance of Flexible Pavement. Austin: Center for Transportation Research, Report No. SWUTC/16/600451-00085-1.

[125] Sheehan, Timothy (2015). High Friction Surface Treatment (HFST). Web. Accessed: 24 Feb. 2015, https://www.theconf.com/files/2021/06/High-Friction-Surface-installed-at-the-I-57_I- 74-Interchange.pdf.

[126] Sheng Y, et al. (2017). Effect of Fibers on Mixture Design of Stone Matrix Asphalt. Applied Sciences. 7(3):297.

[127] Sibson, R., 1973. SLINK: An optimally efficient algorithm for the single-link cluster method. The Computer Journal, 16(1), pp.30-34.

[128] Smith, H. A. (1992). Performance Characteristics of Open-graded Friction Courses. United States: Transportation Research Board.

[129] Snyder, M. (2006). Pavement Surface Characteristics: A Synthesis and Guide. American Concrete Pavement Association (ACPA), Skokie, IL

[130] Snyder, M. (2019) Concrete Pavement Texturing, rep. Federal Highway Administration

[131] (FHWA). Report Number: FHWA-HIF- 17-011.

Spencer, W. T. (1961). Dense-Graded Aggregate Base Construction-ISHC Specifications and Practice.

[132] Stachowiak, G.P., Stachowiak, GW. and Podsiadlo, P. (2008) “Automated classification of wear particles based on their surface texture and shape features,” Tribology International, 41(1), pp. 34-43.

[133] Stehman, S. V. (1997). "Selecting and interpreting measures of thematic classification accuracy". Remote Sensing of Environment. 62 (1): 77-89.

[134] Taylor, G. J. (2014). Open graded friction Courses (OGFC), Course No: C02-058. Continuing Education and Development, Inc, 22.

[135] Taylor, J.B., Carrano, A.L. and Kandlikar, S.G (2006) “Characterization of the Effect of Surface Roughness and Texture on Fluid Flow — Past, Present, and Future,” International Journal of Thermal Sciences, 45(10), pp. 962-968.

[136] Texas Department of Transportation (TxDOT). (2014). Standard Specifications for Construction and Maintenance of Highways, Streets, and Bridges. Austin, Texas.

[137] Tolles, J. and Meurer, W.J. (2016) “Logistic regression,” JAMA, 316(5), p. 533. Underwood, S. et al. (2022) Evolution of Pavement Friction and Macrotexture After Asphalt [138] Overlay, rep. Raleigh, North Carolina: North Carolina Department of Transportation. Report No. FHWA/NC/2020-11

[139 ] Villani, M., Scarpas, A., Bondt, A., Khedoe, R. and Artamendi, I. (2014). Application of Fractal Analysis for Measuring the Effects of Rubber Polishing on the Friction of Asphalt Concrete Mixtures. Wear, pp.179-188.

[140] Wang, H. et al. (2015) “Development of morphological properties of road surfacing aggregates during the polishing process,” International Journal of Pavement Engineering, 18(4), pp. 367-380.

[141] Wang, K.C. (2011) “Elements of automated survey of pavements and a 3D methodology,” Journal of Modern Transportation, 19(1), pp. 51-57.

[142] Watson, D., A. Johnson, and D. Jared. (1998). Georgia Department of Transportation’s Progress in Open-Graded Friction Course Development. In Transportation Research 1616, TRB, National Research Council, Washington, D.C., pp. 30-33.

[143] Wei, L., Fwa, T.F. and Zhe, Z. (2005) “Wavelet analysis and interpretation of Road Roughness,” Journal of Transportation Engineering, 131(2), pp. 120-130.

[144] Whitehouse, D.J. (1994) Handbook of Surface Metrology. Bristol: Institute of Physics Pub.

[145] Wu, C-L. and Nagi, M.A. (1995) Optimizing surface texture of concrete pavement. Skokie, Ill: PCA Portland Cement Association.

[146] Wu, X. et al. (2007) “Top 10 algorithms in Data Mining,” Knowledge and Information Systems, 14(1), pp. 1-37.

[147] Xie, H., & Watson, D. E. (2004). Determining air voids content of compacted stone matrix asphalt mixtures. Transportation research record, 1891(1), 203-211.

[148] Xing, B. et al. (2021) “Influence of Particle Mineralogy and Size on the Morphological Characteristics of Mineral Fillers,” Journal of Materials Research and Technology, 15, pp. 3995- 4009.

[149] Xu, Y. (2018) “Fractal dimension of demolition waste fragmentation and its implication of compactness,” Powder Technology, 339, pp. 922-929. Available at: https://d0i.0rg/l 0.1016/j .powtec.2018.08.071.

[150] Yang, Y., Wang, J.F. and Cheng, Y.M. (2016) “Quantified evaluation of particle shape effects from micro-to-macro scales for non-convex grains,” Particuology, 25, pp. 23-35. [151] Zelelew, H., Khasawneh, M. and Abbas, A. (2014) “Wavelet- based characterisation of asphalt pavement surface macro-texture,” Road Materials and Pavement Design, 15(3), pp. 622- 641.

[152] Zhang, D. et al. (2019) “Automatic Road-Marking Detection and Measurement from Laser-Scanning 3D Profile Data,” Automation in Construction, 108, p. 102957.

[153] Zuniga-Garcia, N. and Prozzi, J. A. (2019) “High-definition field texture measurements for predicting pavement friction,” Transportation Research Record: Journal of the Transportation Research Board, 2673(1), pp. 246-260.

Claims

What is claimed is:

1. A method to determine friction or skid resistance of a roadway, the method comprising: obtaining, by a processor, non-contact measurement data having pavement macrotexture and microtexture of the roadway, wherein the non-contact measurement data is continuously acquired via one or more non-contact sensors by a vehicle housing the non-contact sensor ; determining, by the processor, using the non-contact measurement data, one or more pavement-associated parameters, or associated values, selected from the group consisting of pavement amplitude parameters, pavement statistic parameters, hybrid pavement parameters, and/or pavement spectral parameters; determining, by the processor, a pavement type via at least one of (i) a first classifier using the one or more pavement-associated parameters or associated values or (ii) a second classifier configured to determine pavement type based on a second non-contact measurement data; and determining, by the processor, a value for the friction or skid resistance of the roadway using a model defined for the pavement type using at least one of outputs of the first classifier or outputs of the second classifier, wherein the value for the friction or skid resistance is employed to (i) determine a maintenance or repair schedule or event for the roadway or associated roadway network or (ii) generate a friction or skid resistance map for the roadway or associated roadway network.

2. A method to determine friction or skid resistance of a roadway, the method comprising: obtaining, by a processor, non-contact measurement data having pavement macrotexture and microtexture of the roadway, wherein the non-contact measurement data is continuously acquired via one or more non-contact sensors by a vehicle housing the non-contact sensor measuring the roadway; and transmitting, by the processor, the non-contact measurement data to an analysis system, wherein the analysis system is configured to: determine using the non-contact measurement data, one or more pavement- associated parameters, or associated values, selected from the group consisting of pavement amplitude parameters, pavement statistic parameters, hybrid pavement parameters, and/or pavement spectral parameters; determine a pavement type via at least one of (i) a first classifier using the one or more pavement-associated parameters or associated values or (ii) a second classifier configured to determine pavement type based on a second non-contact measurement data; and determine a value for the friction or skid resistance of the roadway using a model defined for the pavement type using at least one of outputs of the first classifier or outputs of the second classifier, wherein the value for the friction or skid resistance is employed to (i) determine a maintenance or repair schedule or event for the roadway or associated roadway network or (ii) generate a friction or skid resistance map for the roadway or associated roadway network.

3. A method of claim 2 further comprising: receiving, by the processor, friction or skid resistance map for the roadway or associated roadway network from an analysis system; and updating, by the processor, a control operation of a vehicle system (e.g., braking system) using the received friction or skid resistance map.

4. A method of claim 2 further comprising: receiving, by the processor, friction or skid resistance map for the roadway or associated roadway network from an analysis system; and generating, by the processor, at a display of the vehicle, a notification or indicator to a user of the vehicle associated with the received friction or skid resistance map.

5. The method of any one of claims 1-4, wherein the non-contact measurement data is acquired from a line-laser scanner.

6. The method of any one of claims 1-5, wherein the one or more pavement-associated parameters or associated values include a kurtosis parameter that indicates a presence of extremely high peaks or deep valleys.

7. The method of any one of claims 1 -6, wherein the one or more pavement-associated parameters or associated values include a profile solidity factor parameter that is a ratio between a maximum depth of identified valleys and a maximum height of an acquired 2D scan associated with a non-contact sensor of the non-contact sensors.

8. The method of any one of claims 1-7, wherein the one or more pavement-associated parameters or associated values include a mean cross- width parameter that is a measure of an average distance between points where an acquired 2D scan associated with a non-contact sensor of the non-contact sensors crosses from above to below a baseline horizontal plane at a zero elevation.

9. The method of any one of claims 1-8, wherein the one or more pavement-associated parameters or associated values include a cross-width variance that measures a variance of a distance between points where an acquired 2D scan associated with a non-contact optical sensor of the non-contact optical sensors crosses a determined mean of the acquired 2D scan.

10. The method of any one of claims 1-9, wherein the one or more pavement-associated parameters or associated values include a 2-Pt slope variance measure of a slope between two consecutive points.

11. The method of any one of claims 1-10, wherein the one or more pavement associated parameters or associated values include at least one of: (i) a kurtosis parameter that a presence of extremely high peaks or deep valleys, (ii) a profile solidity factor parameter that is a ratio between a maximum depth of identified valleys and a maximum height of an acquired 2D scan associated with a non-contact sensor of the non-contact sensors, (iii) a mean cross width parameter that is a measure of an average distance between points where the acquired 2D scan, (iv) a cross width variance that measures a variance of a distance between points where the acquired 2D scan crosses a determined mean of the acquired 2D scan, (v) a 2-Pt slope variance measure of a slope between two consecutive points, or (v) a combination thereof.

12. The method of any one of claims 1-11, wherein the value for the friction or skid resistance of the roadway is determined using a trained machine learning model, as the first classifier, comprising a decision tree classifier.

13. The method of claim 12, wherein the trained machine learning model was trained using data selected from the group consisting of: chip seals with high macrotexture; dense fine mixes; chip seals with low macrotexture; open mixes or PFCs; dense coarse mixes; stone matrix asphalt; finish-graded concrete; or a combination thereof.

14. The method of any one of claims 1-13 further comprising: transmitting, by the processor, the determined value for the friction or skid resistance and a corresponding positioning data to a global analysis system, wherein the global analysis system is configured to aggregate the determined value for the friction or skid resistance for the corresponding positioning data along with determined values for the friction or skid resistance of other positioning data to generate a friction or skid resistance map for a given geographic area.

15. The method of any one of claims 1-14 further comprising: transmitting, by the processor, the determined value for the friction or skid resistance and a corresponding positioning data to a global analysis system, wherein the global analysis system is configured to aggregate the determined value for the friction or skid resistance for the corresponding positioning data along with determined values for the friction or skid resistance of other positioning data to generate a road risk map for a given geographic area.

16. The method of claims 13 or 14, wherein the friction or skid resistance map or the road risk map generated for the given geographic area are subsequently transmitted, in whole or in part, to vehicles traveling through the given geographic area.

17. The method of any one of claims 1-15, wherein the global analysis system is a cloudbased system.

18. A method of any one of claims 1-17, wherein the step of determining pavement type based on the second non-contact measurement data comprises: obtaining, by a processor, second non-contact measurement data, wherein the second non-contact measurement data is continuously acquired via one or more second non-contact sensors by the vehicle housing the one or more second non-contact sensors; and determining, by the processor, using the second non-contact measurement data, a pavement type via the second classifier, wherein the second classifier was trained for different pavement types.

19. The method of claim 18, wherein both (i) the value for the friction or skid resistance and (ii) the determined pavement type is employed to (iii) determine the maintenance or repair schedule or event for the roadway or associated roadway network and/or (iv) generate a pavement type map for the roadway or associated roadway network.

20. The method of claim 19, wherein the pavement type map includes indicators of patched roadway repairs.

21. The method of any one of claims 1 -4, wherein the second non-contact measurement data is acquired from one or more high-speed cameras.

22. A method to determine a pavement type of a roadway, the method comprising: obtaining, by a processor, non-contact measurement data having a set of images of pavement of the roadway, wherein the measurement data is continuously acquired at high speed via one or more non-contact high-speed sensors by a vehicle housing the non-contact high-speed sensor; determining, by the processor, a pavement type via a classifier using the non-contact measurement data; wherein the determined pavement type is employed to (i) determine a maintenance or repair schedule or event for the roadway or associated roadway network, (ii) generate a pavement type map for the roadway or associated roadway network and/or (iii) perform subsequent analysis to determine a friction or skid resistance map for the roadway or associated roadway network.

23. A system comprising: a processor; and a memory having instructions stored thereon, wherein execution of the instructions by the processor, causes the processor to perform any of the methods of claims 1-22.

24. A non-transitory computer-readable medium having instructions stored thereon, wherein execution of the instructions by the processor causes the processor to perform any of the methods of claims 1-23.

25. A system comprising: a non-contact sensor configured to acquire non-contact measurement data having pavement macrotexture and microtexture of a pavement surface; and an analysis system comprising a processor and a memory having instructions stored thereon, wherein execution of the instructions by the processor causes the processor to: determine using the non-contact measurement data, one or more pavement- associated parameters, or associated values, selected from the group consisting of pavement amplitude parameters, pavement statistic parameters, hybrid pavement parameters, and/or pavement spectral parameters; determine a pavement type via at least one of (i) a first classifier using the one or more pavement-associated parameters or associated values or (ii) a second classifier configured to determine pavement type based on a second non-contact measurement data; and determine a value for the friction or skid resistance of the roadway using a model defined for the pavement type, wherein the value for the friction or skid resistance is employed to (i) determine a maintenance or repair schedule or event for the roadway or associated roadway network or (ii) generate a friction or skid resistance map for the roadway or associated roadway network.

26. A system comprising: a non-contact sensor configured to acquire non-contact measurement data having pavement macrotexture and microtexture of a pavement surface; and a network interface configured to transmit the non-contact measurement data to an analysis system, wherein the analysis system comprises a processor and a memory having instructions stored thereon, wherein execution of the instructions by the processor causes the processor to: determine using the non-contact measurement data, one or more pavement- associated parameters, or associated values, selected from the group consisting of pavement amplitude parameters, pavement statistic parameters, hybrid pavement parameters, and/or pavement spectral parameters; determine a pavement type via at least one of (i) a first classifier using the one or more pavement-associated parameters or associated values or (ii) a second classifier configured to determine pavement type based on a second non-contact measurement data; and determine a value for the friction or skid resistance of the roadway using a model defined for the pavement type using at least one of outputs of the first classifier or outputs of the second classifier, wherein the value for the friction or skid resistance is employed to (i) determine a maintenance or repair schedule or event for the roadway or associated roadway network or (ii) generate a friction or skid resistance map for the roadway or associated roadway network.

27. A system comprising: a non-contact sensor configured to acquire non-contact measurement data of a pavement surface; and an analysis system comprising a processor and a memory having instructions stored thereon, wherein execution of the instructions by the processor causes the processor to: obtain the non-contact measurement data having a set of images of pavement of the roadway, wherein the non-contact measurement data is continuously acquired at high speed via one or more non-contact high-speed sensors by a vehicle housing the non-contact high-speed sensor; determine a pavement type via a classifier using the non-contact measurement data; wherein the determined pavement type is employed to (i) determine a maintenance or repair schedule or event for the roadway or associated roadway network, (ii) generate a pavement type map for the roadway or associated roadway network, and/or (iii) perform subsequent analysis to determine a friction or skid resistance map for the roadway or associated roadway network.

28. The system of any one of claims 25-27, wherein the non-contact sensor is mounted to a trailer.

29. The system of any one of claims 25-27, wherein the non-contact sensor is mounted to a vehicle.

30. The system of any one of claims 25-26 or 28-29, wherein the non-contact sensor is a linelaser scanner.

30. The system of any one of claims 27-29, wherein the non-contact sensor is a high-speed camera.

31. A system of any one of claims 25-30 further comprising: a vehicle controller, the vehicle having a processor and memory, the memory having instructions stored thereon, wherein execution of the instructions by the processor causes the processor to: receive a friction or skid resistance map for the roadway or associated roadway network from an analysis system; and update a control operation of a vehicle system using the received friction or skid resistance map.

32. A system of claim 31, wherein execution of the instructions by the processor further causes the processor to: receive a friction or skid resistance map for the roadway or associated roadway network from an analysis system; and generate at a display of the vehicle, a notification or indicator to a user of the vehicle associated with the received friction or skid resistance map.

33. The system of any one of claims 25-32, wherein the analysis system further comprises: a network interface, the network interface being configured to transmit the determined value for the friction or skid resistance and a corresponding positioning data to a global analysis system, wherein the global analysis system is configured to aggregate the determined value for the friction or skid resistance for the corresponding positioning data along with determined values for the friction or skid resistance of other positioning data to generate a friction or skid resistance map for a given geographic area.