EP4673761A1

EP4673761A1 - Camera-to-lidar calibration and validation model

Info

Publication number: EP4673761A1
Application number: EP24717391.7A
Authority: EP
Inventors: Aaron Weng Soon LOW; Mathieu Daniel Jean-Paul COCHETEUX; Marius Lukas Brühlmeier
Original assignee: Motional AD LLC
Current assignee: Motional AD LLC
Priority date: 2023-03-01
Filing date: 2024-03-01
Publication date: 2026-01-07
Also published as: CN121039526A; WO2024182787A1; KR20250156137A

Abstract

The present disclosure describes Camera-to-LiDAR (C2L) extrinsic calibration that leverages self-attention mechanisms through a Transformer-based backbone network to infer the 6-degree of freedom (DoF) relative transformation between sensors.

Description

CAMERA-TO-LIDAR CALIBRATION AND VALIDATION MODEL

CROSS-REFERENCE TO RELATED APPLICATIONS

This application claims the benefit of priority to U.S. Provisional Patent Application No. 63/449,170, filed on March 1, 2023, the entire contents of which are incorporated by reference herein.

BRIEF DESCRIPTION OF THE FIGURES

[0001] FIG. 1 is an example environment in which a vehicle including one or more components of an autonomous system can be implemented;

[0002] FIG. 2 is a diagram of one or more systems of a vehicle including an autonomous system; [0003] FIG. 3 is a diagram of components of one or more devices and/or one or more systems of FIGS. 1 and 2;

[0004] FIG. 4A is a diagram of certain components of an autonomous system;

[0005] FIG. 4B is a diagram of an implementation of a neural network;

[0006] FIG. 4C and 4D are a diagram illustrating example operation of a CNN; and

[0007] FIGS. 5A-5D show implementations 500A-500D of a camera to LiDAR calibration and/or validation model.

[0008] FIG. 6 shows the projection of LiDAR points onto a corresponding camera image. [0009] FIG. 7 shows an architecture of a camera-to-LiDAR calibration and/or validation model. [0010] FIG. 8 shows heatmaps generated by the camera-to-LiDAR calibration and validation model.

[0011] FIG. 9 shows an intensity map, depth map, and feature map.

[0012] FIG. 10 is a process flow diagram of a process 1000 for a camera-to-LiDAR calibration and validation model.

DETAILED DESCRIPTION

[0013] In the following description numerous specific details are set forth in order to provide a thorough understanding of the present disclosure for the purposes of explanation. It will be apparent, however, that the embodiments described by the present disclosure can be practiced without these specific details. In some instances, well-known structures and devices are illustrated in block diagram form in order to avoid unnecessarily obscuring aspects of the present disclosure. [0014] Specific arrangements or orderings of schematic elements, such as those representing systems, devices, modules, instruction blocks, data elements, and/or the like are illustrated in the drawings for ease of description. However, it will be understood by those skilled in the art that the specific ordering or arrangement of the schematic elements in the drawings is not meant to imply that a particular order or sequence of processing, or separation of processes, is required unless explicitly described as such. Further, the inclusion of a schematic element in a drawing is not meant to imply that such element is required in all embodiments or that the features represented by such element may not be included in or combined with other elements in some embodiments unless explicitly described as such.

[0015] Further, where connecting elements such as solid or dashed lines or arrows are used in the drawings to illustrate a connection, relationship, or association between or among two or more other schematic elements, the absence of any such connecting elements is not meant to imply that no connection, relationship, or association can exist. In other words, some connections, relationships, or associations between elements are not illustrated in the drawings so as not to obscure the disclosure. In addition, for ease of illustration, a single connecting element can be used to represent multiple connections, relationships or associations between elements. For example, where a connecting element represents communication of signals, data, or instructions (e.g., “software instructions”), it should be understood by those skilled in the art that such element can represent one or multiple signal paths (e.g., a bus), as may be needed, to affect the communication. [0016] Although the terms first, second, third, and/or the like are used to describe various elements, these elements should not be limited by these terms. The terms first, second, third, and/or the like are used only to distinguish one element from another. For example, a first contact could be termed a second contact and, similarly, a second contact could be termed a first contact without departing from the scope of the described embodiments. The first contact and the second contact are both contacts, but they are not the same contact.

[0017] The terminology used in the description of the various described embodiments herein is included for the purpose of describing particular embodiments only and is not intended to be limiting. As used in the description of the various described embodiments and the appended claims, the singular forms “a,” “an” and “the” are intended to include the plural forms as well and can be used interchangeably with “one or more” or “at least one,” unless the context clearly indicates otherwise. It will also be understood that the term “and/or” as used herein refers to and encompasses any and all possible combinations of one or more of the associated listed items. It will be further understood that the terms “includes,” “including,” “comprises,” and/or “comprising,” when used in this description specify the presence of stated features, integers, steps, operations, elements, and/or components, but do not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and/or groups thereof.

[0018] As used herein, the terms “communication” and “communicate” refer to at least one of the reception, receipt, transmission, transfer, provision, and/or the like of information (or information represented by, for example, data, signals, messages, instructions, commands, and/or the like). For one unit (e.g., a device, a system, a component of a device or system, combinations thereof, and/or the like) to be in communication with another unit means that the one unit is able to directly or indirectly receive information from and/or send (e.g., transmit) information to the other unit. This may refer to a direct or indirect connection that is wired and/or wireless in nature. Additionally, two units may be in communication with each other even though the information transmitted may be modified, processed, relayed, and/or routed between the first and second unit. For example, a first unit may be in communication with a second unit even though the first unit passively receives information and does not actively transmit information to the second unit. As another example, a first unit may be in communication with a second unit if at least one intermediary unit (e.g., a third unit located between the first unit and the second unit) processes information received from the first unit and transmits the processed information to the second unit. In some embodiments, a message may refer to a network packet (e.g., a data packet and/or the like) that includes data.

[0019] As used herein, the term “if’ is, optionally, construed to mean “when”, “upon”, “in response to determining,” “in response to detecting,” and/or the like, depending on the context. Similarly, the phrase “if it is determined” or “if [a stated condition or event] is detected” is, optionally, construed to mean “upon determining,” “in response to determining,” “upon detecting [the stated condition or event],” “in response to detecting [the stated condition or event],” and/or the like, depending on the context. Also, as used herein, the terms “has”, “have”, “having”, or the like are intended to be open-ended terms. Further, the phrase “based on” is intended to mean “based at least partially on” unless explicitly stated otherwise. [0020] Reference will now be made in detail to embodiments, examples of which are illustrated in the accompanying drawings. In the following detailed description, numerous specific details are set forth in order to provide a thorough understanding of the various described embodiments. However, it will be apparent to one of ordinary skill in the art that the various described embodiments can be practiced without these specific details. In other instances, well-known methods, procedures, components, circuits, and networks have not been described in detail so as not to unnecessarily obscure aspects of the embodiments.

[0021] General Overview

[0022] In some aspects and/or embodiments, systems, methods, and computer program products described herein include and/or implement a Camera-to-LiDAR Calibration and Validation Model.

[0023] Referring now to FIG. 1, illustrated is example environment 100 in which vehicles that include autonomous systems, as well as vehicles that do not, are operated. As illustrated, environment 100 includes vehicles 102a-102n, objects 104a-104n, routes 106a-106n, area 108, vehicle-to-infrastructure (V2I) device 110, network 112, remote autonomous vehicle (AV) system 114, fleet management system 116, and V2I system 118. Vehicles 102a-102n, vehicle-to- infrastructure (V2I) device 110, network 112, autonomous vehicle (AV) system 114, fleet management system 116, and V2I system 118 interconnect (e.g., establish a connection to communicate and/or the like) via wired connections, wireless connections, or a combination of wired or wireless connections. In some embodiments, objects 104a-104n interconnect with at least one of vehicles 102a-102n, vehicle-to-infrastructure (V2I) device 110, network 112, autonomous vehicle (AV) system 114, fleet management system 116, and V2I system 118 via wired connections, wireless connections, or a combination of wired or wireless connections.

[0024] Vehicles 102a-102n (referred to individually as vehicle 102 and collectively as vehicles 102) include at least one device configured to transport goods and/or people. In some embodiments, vehicles 102 are configured to be in communication with V2I device 110, remote AV system 114, fleet management system 116, and/or V2I system 118 via network 112. In some embodiments, vehicles 102 include cars, buses, trucks, trains, and/or the like. In some embodiments, vehicles 102 are the same as, or similar to, vehicles 200, described herein (see FIG. 2). In some embodiments, a vehicle 200 of a set of vehicles 200 is associated with an autonomous fleet manager. In some embodiments, vehicles 102 travel along respective routes 106a-106n (referred to individually as route 106 and collectively as routes 106), as described herein. In some embodiments, one or more vehicles 102 include an autonomous system (e.g., an autonomous system that is the same as or similar to autonomous system 202).

[0025] Objects 104a-104n (referred to individually as object 104 and collectively as objects 104) include, for example, at least one vehicle, at least one pedestrian, at least one cyclist, at least one structure (e.g., a building, a sign, a fire hydrant, etc.), and/or the like. Each object 104 is stationary (e.g., located at a fixed location for a period of time) or mobile (e.g., having a velocity and associated with at least one trajectory). In some embodiments, objects 104 are associated with corresponding locations in area 108.

[0026] Routes 106a-106n (referred to individually as route 106 and collectively as routes 106) are each associated with (e.g., prescribe) a sequence of actions (also known as a trajectory) connecting states along which an AV can navigate. Each route 106 starts at an initial state (e.g., a state that corresponds to a first spatiotemporal location, velocity, and/or the like) and ends at a final goal state (e.g., a state that corresponds to a second spatiotemporal location that is different from the first spatiotemporal location) or goal region (e.g. a subspace of acceptable states (e.g., terminal states)). In some embodiments, the first state includes a location at which an individual or individuals are to be picked-up by the AV and the second state or region includes a location or locations at which the individual or individuals picked-up by the AV are to be dropped-off. In some embodiments, routes 106 include a plurality of acceptable state sequences (e.g., a plurality of spatiotemporal location sequences), the plurality of state sequences associated with (e.g., defining) a plurality of trajectories. In an example, routes 106 include only high level actions or imprecise state locations, such as a series of connected roads dictating turning directions at roadway intersections. Additionally, or alternatively, routes 106 may include more precise actions or states such as, for example, specific target lanes or precise locations within the lane areas and targeted speed at those positions. In an example, routes 106 include a plurality of precise state sequences along the at least one high level action sequence with a limited lookahead horizon to reach intermediate goals, where the combination of successive iterations of limited horizon state sequences cumulatively correspond to a plurality of trajectories that collectively form the high level route to terminate at the final goal state or region.

[0027] Area 108 includes a physical area (e.g., a geographic region) within which vehicles 102 can navigate. In an example, area 108 includes at least one state (e.g., a country, a province, an individual state of a plurality of states included in a country, etc.), at least one portion of a state, at least one city, at least one portion of a city, etc. In some embodiments, area 108 includes at least one named thoroughfare (referred to herein as a “road”) such as a highway, an interstate highway, a parkway, a city street, etc. Additionally, or alternatively, in some examples area 108 includes at least one unnamed road such as a driveway, a section of a parking lot, a section of a vacant and/or undeveloped lot, a dirt path, etc. In some embodiments, a road includes at least one lane (e.g., a portion of the road that can be traversed by vehicles 102). In an example, a road includes at least one lane associated with (e.g., identified based on) at least one lane marking.

[0028] Vehicle-to-Infrastructure (V2I) device 110 (sometimes referred to as a Vehicle-to- Infrastructure or Vehicle-to-Everything (V2X) device) includes at least one device configured to be in communication with vehicles 102 and/or V2I infrastructure system 118. In some embodiments, V2I device 110 is configured to be in communication with vehicles 102, remote AV system 114, fleet management system 116, and/or V2I system 118 via network 112. In some embodiments, V2I device 110 includes a radio frequency identification (RFID) device, signage, cameras (e.g., two-dimensional (2D) and/or three-dimensional (3D) cameras), lane markers, streetlights, parking meters, etc. In some embodiments, V2I device 110 is configured to communicate directly with vehicles 102. Additionally, or alternatively, in some embodiments V2I device 110 is configured to communicate with vehicles 102, remote AV system 114, and/or fleet management system 116 via V2I system 118. In some embodiments, V2I device 110 is configured to communicate with V2I system 118 via network 112.

[0029] Network 112 includes one or more wired and/or wireless networks. In an example, network 112 includes a cellular network (e.g., a long term evolution (LTE) network, a third generation (3G) network, a fourth generation (4G) network, a fifth generation (5G) network, a code division multiple access (CDMA) network, etc.), a public land mobile network (PLMN), a local area network (LAN), a wide area network (WAN), a metropolitan area network (MAN), a telephone network (e.g., the public switched telephone network (PSTN), a private network, an ad hoc network, an intranet, the Internet, a fiber optic-based network, a cloud computing network, etc., a combination of some or all of these networks, and/or the like.

[0030] Remote AV system 114 includes at least one device configured to be in communication with vehicles 102, V2I device 110, network 112, fleet management system 116, and/or V2I system 118 via network 112. In an example, remote AV system 114 includes a server, a group of servers, and/or other like devices. In some embodiments, remote AV system 114 is co-located with the fleet management system 116. In some embodiments, remote AV system 114 is involved in the installation of some or all of the components of a vehicle, including an autonomous system, an autonomous vehicle compute, software implemented by an autonomous vehicle compute, and/or the like. In some embodiments, remote AV system 114 maintains (e.g., updates and/or replaces) such components and/or software during the lifetime of the vehicle.

[0031] Fleet management system 116 includes at least one device configured to be in communication with vehicles 102, V2I device 110, remote AV system 114, and/or V2I infrastructure system 118. In an example, fleet management system 116 includes a server, a group of servers, and/or other like devices. In some embodiments, fleet management system 116 is associated with a ridesharing company (e.g., an organization that controls operation of multiple vehicles (e.g., vehicles that include autonomous systems and/or vehicles that do not include autonomous systems) and/or the like).

[0032] In some embodiments, V2I system 118 includes at least one device configured to be in communication with vehicles 102, V2I device 110, remote AV system 114, and/or fleet management system 116 via network 112. In some examples, V2I system 118 is configured to be in communication with V2I device 110 via a connection different from network 112. In some embodiments, V2I system 118 includes a server, a group of servers, and/or other like devices. In some embodiments, V2I system 118 is associated with a municipality or a private institution (e.g., a private institution that maintains V2I device 110 and/or the like).

[0033] The number and arrangement of elements illustrated in FIG. 1 are provided as an example. There can be additional elements, fewer elements, different elements, and/or differently arranged elements, than those illustrated in FIG. 1. Additionally, or alternatively, at least one element of environment 100 can perform one or more functions described as being performed by at least one different element of FIG. 1. Additionally, or alternatively, at least one set of elements of environment 100 can perform one or more functions described as being performed by at least one different set of elements of environment 100.

[0034] Referring now to FIG. 2, vehicle 200 (which may be the same as, or similar to vehicles 102 of FIG. 1) includes or is associated with autonomous system 202, powertrain control system 204, steering control system 206, and brake system 208. In some embodiments, vehicle 200 is the same as or similar to vehicle 102 (see FIG. 1). In some embodiments, autonomous system 202 is configured to confer vehicle 200 autonomous driving capability (e.g., implement at least one driving automation or maneuver-based function, feature, device, and/or the like that enable vehicle 200 to be partially or fully operated without human intervention including, without limitation, fully autonomous vehicles (e.g., vehicles that forego reliance on human intervention such as Level 5 ADS-operated vehicles), highly autonomous vehicles (e.g., vehicles that forego reliance on human intervention in certain situations such as Level 4 ADS-operated vehicles), conditional autonomous vehicles (e.g., vehicles that forego reliance on human intervention in limited situations such as Level 3 ADS-operated vehicles) and/or the like. In one embodiment, autonomous system 202 includes operational or tactical functionality required to operate vehicle 200 in on-road traffic and perform part or all of Dynamic Driving Task (DDT) on a sustained basis. In another embodiment, autonomous system 202 includes an Advanced Driver Assistance System (ADAS) that includes driver support features. Autonomous system 202 supports various levels of driving automation, ranging from no driving automation (e.g., Level 0) to full driving automation (e.g., Level 5). For a detailed description of fully autonomous vehicles and highly autonomous vehicles, reference may be made to SAE International's standard J3016: Taxonomy and Definitions for Terms Related to On-Road Motor Vehicle Automated Driving Systems, which is incorporated by reference in its entirety. In some embodiments, vehicle 200 is associated with an autonomous fleet manager and/or a ridesharing company.

[0035] Autonomous system 202 includes a sensor suite that includes one or more devices such as cameras 202a, LiDAR sensors 202b, radar sensors 202c, and microphones 202d. In some embodiments, autonomous system 202 can include more or fewer devices and/or different devices (e.g., ultrasonic sensors, inertial sensors, GPS receivers (discussed below), odometry sensors that generate data associated with an indication of a distance that vehicle 200 has traveled, and/or the like). In some embodiments, autonomous system 202 uses the one or more devices included in autonomous system 202 to generate data associated with environment 100, described herein. The data generated by the one or more devices of autonomous system 202 can be used by one or more systems described herein to observe the environment (e.g., environment 100) in which vehicle 200 is located. In some embodiments, autonomous system 202 includes communication device 202e, autonomous vehicle compute 202f, drive-by-wire (DBW) system 202h, and safety controller 202g. [0036] Cameras 202a include at least one device configured to be in communication with communication device 202e, autonomous vehicle compute 202f, and/or safety controller 202g via a bus (e.g., a bus that is the same as or similar to bus 302 of FIG. 3). Cameras 202a include at least one camera (e.g., a digital camera using a light sensor such as a Charge-Coupled Device (CCD), a thermal camera, an infrared (IR) camera, an event camera, and/or the like) to capture images including physical objects (e.g., cars, buses, curbs, people, and/or the like). In some embodiments, camera 202a generates camera data as output. In some examples, camera 202a generates camera data that includes image data associated with an image. In this example, the image data may specify at least one parameter (e.g., image characteristics such as exposure, brightness, etc., an image timestamp, and/or the like) corresponding to the image. In such an example, the image may be in a format (e.g., RAW, JPEG, PNG, and/or the like). In some embodiments, camera 202a includes a plurality of independent cameras configured on (e.g., positioned on) a vehicle to capture images for the purpose of stereopsis (stereo vision). In some examples, camera 202a includes a plurality of cameras that generate image data and transmit the image data to autonomous vehicle compute 202f and/or a fleet management system (e.g., a fleet management system that is the same as or similar to fleet management system 116 of FIG. 1). In such an example, autonomous vehicle compute 202f determines depth to one or more objects in a field of view of at least two cameras of the plurality of cameras based on the image data from the at least two cameras. In some embodiments, cameras 202a is configured to capture images of objects within a distance from cameras 202a (e.g., up to 100 meters, up to a kilometer, and/or the like). Accordingly, cameras 202a include features such as sensors and lenses that are optimized for perceiving objects that are at one or more distances from cameras 202a.

[0037] In an embodiment, camera 202a includes at least one camera configured to capture one or more images associated with one or more traffic lights, street signs and/or other physical objects that provide visual navigation information. In some embodiments, camera 202a generates traffic light data associated with one or more images. In some examples, camera 202a generates TLD (Traffic Light Detection) data associated with one or more images that include a format (e.g., RAW, JPEG, PNG, and/or the like). In some embodiments, camera 202a that generates TLD data differs from other systems described herein incorporating cameras in that camera 202a can include one or more cameras with a wide field of view (e.g., a wide-angle lens, a fish-eye lens, a lens having a viewing angle of approximately 120 degrees or more, and/or the like) to generate images about as many physical objects as possible. [0038] Light Detection and Ranging (LiDAR) sensors 202b include at least one device configured to be in communication with communication device 202e, autonomous vehicle compute 202f, and/or safety controller 202g via a bus (e.g., a bus that is the same as or similar to bus 302 of FIG. 3). LiDAR sensors 202b include a system configured to transmit light from a light emitter (e.g., a laser transmitter). Light emitted by LiDAR sensors 202b include light (e.g., infrared light and/or the like) that is outside of the visible spectrum. In some embodiments, during operation, light emitted by LiDAR sensors 202b encounters a physical object (e.g., a vehicle) and is reflected back to LiDAR sensors 202b. In some embodiments, the light emitted by LiDAR sensors 202b does not penetrate the physical objects that the light encounters. LiDAR sensors 202b also include at least one light detector which detects the light that was emitted from the light emitter after the light encounters a physical object. In some embodiments, at least one data processing system associated with LiDAR sensors 202b generates an image (e.g., a point cloud, a combined point cloud, and/or the like) representing the objects included in a field of view of LiDAR sensors 202b. In some examples, the at least one data processing system associated with LiDAR sensor 202b generates an image that represents the boundaries of a physical object, the surfaces (e.g., the topology of the surfaces) of the physical object, and/or the like. In such an example, the image is used to determine the boundaries of physical objects in the field of view of LiDAR sensors 202b.

[0039] Radio Detection and Ranging (radar) sensors 202c include at least one device configured to be in communication with communication device 202e, autonomous vehicle compute 202f, and/or safety controller 202g via a bus (e.g., a bus that is the same as or similar to bus 302 of FIG. 3). Radar sensors 202c include a system configured to transmit radio waves (either pulsed or continuously). The radio waves transmitted by radar sensors 202c include radio waves that are within a predetermined spectrum. In some embodiments, during operation, radio waves transmitted by radar sensors 202c encounter a physical object and are reflected back to radar sensors 202c. In some embodiments, the radio waves transmitted by radar sensors 202c are not reflected by some objects. In some embodiments, at least one data processing system associated with radar sensors 202c generates signals representing the objects included in a field of view of radar sensors 202c. For example, the at least one data processing system associated with radar sensor 202c generates an image that represents the boundaries of a physical object, the surfaces (e.g., the topology of the surfaces) of the physical object, and/or the like. In some examples, the image is used to determine the boundaries of physical objects in the field of view of radar sensors 202c.

[0040] Microphones 202d includes at least one device configured to be in communication with communication device 202e, autonomous vehicle compute 202f, and/or safety controller 202g via a bus (e.g., a bus that is the same as or similar to bus 302 of FIG. 3). Microphones 202d include one or more microphones (e.g., array microphones, external microphones, and/or the like) that capture audio signals and generate data associated with (e.g., representing) the audio signals. In some examples, microphones 202d include transducer devices and/or like devices. In some embodiments, one or more systems described herein can receive the data generated by microphones 202d and determine a position of an object relative to vehicle 200 (e.g., a distance and/or the like) based on the audio signals associated with the data.

[0041] Communication device 202e includes at least one device configured to be in communication with cameras 202a, LiDAR sensors 202b, radar sensors 202c, microphones 202d, autonomous vehicle compute 202f, safety controller 202g, and/or DBW (Drive-By-Wire) system 202h. For example, communication device 202e may include a device that is the same as or similar to communication interface 314 of FIG. 3. In some embodiments, communication device 202e includes a vehicle-to-vehicle (V2V) communication device (e.g., a device that enables wireless communication of data between vehicles).

[0042] Autonomous vehicle compute 202f include at least one device configured to be in communication with cameras 202a, LiDAR sensors 202b, radar sensors 202c, microphones 202d, communication device 202e, safety controller 202g, and/or DBW system 202h. In some examples, autonomous vehicle compute 202f includes a device such as a client device, a mobile device (e.g., a cellular telephone, a tablet, and/or the like), a server (e.g., a computing device including one or more central processing units, graphical processing units, and/or the like), and/or the like. In some embodiments, autonomous vehicle compute 202f is the same as or similar to autonomous vehicle compute 400, described herein. Additionally, or alternatively, in some embodiments autonomous vehicle compute 202f is configured to be in communication with an autonomous vehicle system (e.g., an autonomous vehicle system that is the same as or similar to remote AV system 114 of FIG. 1), a fleet management system (e.g., a fleet management system that is the same as or similar to fleet management system 116 of FIG. 1), a V2I device (e.g., a V2I device that is the same as or similar to V2I device 110 of FIG. 1), and/or a V2I system (e.g., a V2I system that is the same as or similar to V2I system 118 of FIG. 1).

[0043] Safety controller 202g includes at least one device configured to be in communication with cameras 202a, LiDAR sensors 202b, radar sensors 202c, microphones 202d, communication device 202e, autonomous vehicle computer 202f, and/or DBW system 202h. In some examples, safety controller 202g includes one or more controllers (electrical controllers, electromechanical controllers, and/or the like) that are configured to generate and/or transmit control signals to operate one or more devices of vehicle 200 (e.g., powertrain control system 204, steering control system 206, brake system 208, and/or the like). In some embodiments, safety controller 202g is configured to generate control signals that take precedence over (e.g., overrides) control signals generated and/or transmitted by autonomous vehicle compute 202f.

[0044] DBW system 202h includes at least one device configured to be in communication with communication device 202e and/or autonomous vehicle compute 202f. In some examples, DBW system 202h includes one or more controllers (e.g., electrical controllers, electromechanical controllers, and/or the like) that are configured to generate and/or transmit control signals to operate one or more devices of vehicle 200 (e.g., powertrain control system 204, steering control system 206, brake system 208, and/or the like). Additionally, or alternatively, the one or more controllers of DBW system 202h are configured to generate and/or transmit control signals to operate at least one different device (e.g., a turn signal, headlights, door locks, windshield wipers, and/or the like) of vehicle 200.

[0045] Powertrain control system 204 includes at least one device configured to be in communication with DBW system 202h. In some examples, powertrain control system 204 includes at least one controller, actuator, and/or the like. In some embodiments, powertrain control system 204 receives control signals from DBW system 202h and powertrain control system 204 causes vehicle 200 to make longitudinal vehicle motion, such as start moving forward, stop moving forward, start moving backward, stop moving backward, accelerate in a direction, decelerate in a direction or to make lateral vehicle motion such as performing a left turn, performing a right turn, and/or the like. In an example, powertrain control system 204 causes the energy (e.g., fuel, electricity, and/or the like) provided to a motor of the vehicle to increase, remain the same, or decrease, thereby causing at least one wheel of vehicle 200 to rotate or not rotate. [0046] Steering control system 206 includes at least one device configured to rotate one or more wheels of vehicle 200. In some examples, steering control system 206 includes at least one controller, actuator, and/or the like. In some embodiments, steering control system 206 causes the front two wheels and/or the rear two wheels of vehicle 200 to rotate to the left or right to cause vehicle 200 to turn to the left or right. In other words, steering control system 206 causes activities necessary for the regulation of the y-axis component of vehicle motion.

[0047] Brake system 208 includes at least one device configured to actuate one or more brakes to cause vehicle 200 to reduce speed and/or remain stationary. In some examples, brake system 208 includes at least one controller and/or actuator that is configured to cause one or more calipers associated with one or more wheels of vehicle 200 to close on a corresponding rotor of vehicle 200. Additionally, or alternatively, in some examples brake system 208 includes an automatic emergency braking (AEB) system, a regenerative braking system, and/or the like.

[0048] In some embodiments, vehicle 200 includes at least one platform sensor (not explicitly illustrated) that measures or infers properties of a state or a condition of vehicle 200. In some examples, vehicle 200 includes platform sensors such as a global positioning system (GPS) receiver, an inertial measurement unit (IMU), a wheel speed sensor, a wheel brake pressure sensor, a wheel torque sensor, an engine torque sensor, a steering angle sensor, and/or the like. Although brake system 208 is illustrated to be located in the near side of vehicle 200 in FIG. 2, brake system 208 may be located anywhere in vehicle 200.

[0049] Referring now to FIG. 3, illustrated is a schematic diagram of a device 300. As illustrated, device 300 includes processor 304, memory 306, storage component 308, input interface 310, output interface 312, communication interface 314, and bus 302. In some embodiments, device 300 corresponds to at least one device of vehicles 102 (e.g., at least one device of a system of vehicles 102), and/or one or more devices of network 112 (e.g., one or more devices of a system of network 112). In some embodiments, one or more devices of vehicles 102 (e.g., one or more devices of a system of vehicles 102) and/or one or more devices of network 112 (e.g., one or more devices of a system of network 112) include at least one device 300 and/or at least one component of device 300. As shown in FIG. 3, device 300 includes bus 302, processor 304, memory 306, storage component 308, input interface 310, output interface 312, and communication interface 314. [0050] Bus 302 includes a component that permits communication among the components of device 300. In some cases, processor 304 includes a processor (e.g., a central processing unit (CPU), a graphics processing unit (GPU), an accelerated processing unit (APU), and/or the like), a microphone, a digital signal processor (DSP), and/or any processing component (e.g., a field- programmable gate array (FPGA), an application specific integrated circuit (ASIC), and/or the like) that can be programmed to perform at least one function. Memory 306 includes random access memory (RAM), read-only memory (ROM), and/or another type of dynamic and/or static storage device (e.g., flash memory, magnetic memory, optical memory, and/or the like) that stores data and/or instructions for use by processor 304.

[0051] Storage component 308 stores data and/or software related to the operation and use of device 300. In some examples, storage component 308 includes a hard disk (e.g., a magnetic disk, an optical disk, a magneto-optic disk, a solid state disk, and/or the like), a compact disc (CD), a digital versatile disc (DVD), a floppy disk, a cartridge, a magnetic tape, a CD-ROM, RAM, PROM, EPROM, FLASH-EPROM, NV-RAM, and/or another type of computer readable medium, along with a corresponding drive.

[0052] Input interface 310 includes a component that permits device 300 to receive information, such as via user input (e.g., a touchscreen display, a keyboard, a keypad, a mouse, a button, a switch, a microphone, a camera, and/or the like). Additionally or alternatively, in some embodiments input interface 310 includes a sensor that senses information (e.g., a global positioning system (GPS) receiver, an accelerometer, a gyroscope, an actuator, and/or the like). Output interface 312 includes a component that provides output information from device 300 (e.g., a display, a speaker, one or more light-emitting diodes (LEDs), and/or the like).

[0053] In some embodiments, communication interface 314 includes a transceiver-like component (e g., a transceiver, a separate receiver and transmitter, and/or the like) that permits device 300 to communicate with other devices via a wired connection, a wireless connection, or a combination of wired and wireless connections. In some examples, communication interface 314 permits device 300 to receive information from another device and/or provide information to another device. In some examples, communication interface 314 includes an Ethernet interface, an optical interface, a coaxial interface, an infrared interface, a radio frequency (RF) interface, a universal serial bus (USB) interface, a Wi-Fi® interface, a cellular network interface, and/or the like. [0054] In some embodiments, device 300 performs one or more processes described herein. Device 300 performs these processes based on processor 304 executing software instructions stored by a computer-readable medium, such as memory 305 and/or storage component 308. A computer-readable medium (e.g., a non-transitory computer readable medium) is defined herein as a non-transitory memory device. A non-transitory memory device includes memory space located inside a single physical storage device or memory space spread across multiple physical storage devices.

[0055] In some embodiments, software instructions are read into memory 306 and/or storage component 308 from another computer-readable medium or from another device via communication interface 314. When executed, software instructions stored in memory 306 and/or storage component 308 cause processor 304 to perform one or more processes described herein. Additionally or alternatively, hardwired circuitry is used in place of or in combination with software instructions to perform one or more processes described herein. Thus, embodiments described herein are not limited to any specific combination of hardware circuitry and software unless explicitly stated otherwise.

[0056] Memory 306 and/or storage component 308 includes data storage or at least one data structure (e.g., a database and/or the like). Device 300 is capable of receiving information from, storing information in, communicating information to, or searching information stored in the data storage or the at least one data structure in memory 306 or storage component 308. In some examples, the information includes network data, input data, output data, or any combination thereof.

[0057] In some embodiments, device 300 is configured to execute software instructions that are either stored in memory 306 and/or in the memory of another device (e.g., another device that is the same as or similar to device 300). As used herein, the term “module” refers to at least one instruction stored in memory 306 and/or in the memory of another device that, when executed by processor 304 and/or by a processor of another device (e.g., another device that is the same as or similar to device 300) cause device 300 (e.g., at least one component of device 300) to perform one or more processes described herein. In some embodiments, a module is implemented in software, firmware, hardware, and/or the like.

[0058] The number and arrangement of components illustrated in FIG. 3 are provided as an example. In some embodiments, device 300 can include additional components, fewer components, different components, or differently arranged components than those illustrated in FIG. 3. Additionally or alternatively, a set of components (e.g., one or more components) of device 300 can perform one or more functions described as being performed by another component or another set of components of device 300.

[0059] Referring now to FIG. 4A, illustrated is an example block diagram of an autonomous vehicle compute 400 (sometimes referred to as an “AV stack”). As illustrated, autonomous vehicle compute 400 includes perception system 402 (sometimes referred to as a perception module), planning system 404 (sometimes referred to as a planning module), localization system 406 (sometimes referred to as a localization module), control system 408 (sometimes referred to as a control module), and database 410. In some embodiments, perception system 402, planning system 404, localization system 406, control system 408, and database 410 are included and/or implemented in an autonomous navigation system of a vehicle (e.g., autonomous vehicle compute 202f of vehicle 200). Additionally, or alternatively, in some embodiments perception system 402, planning system 404, localization system 406, control system 408, and database 410 are included in one or more standalone systems (e.g., one or more systems that are the same as or similar to autonomous vehicle compute 400 and/or the like). In some examples, perception system 402, planning system 404, localization system 406, control system 408, and database 410 are included in one or more standalone systems that are located in a vehicle and/or at least one remote system as described herein. In some embodiments, any and/or all of the systems included in autonomous vehicle compute 400 are implemented in software (e.g., in software instructions stored in memory), computer hardware (e.g., by microprocessors, microcontrollers, application-specific integrated circuits (ASICs), Field Programmable Gate Arrays (FPGAs), and/or the like), or combinations of computer software and computer hardware. It will also be understood that, in some embodiments, autonomous vehicle compute 400 is configured to be in communication with a remote system (e.g., an autonomous vehicle system that is the same as or similar to remote AV system 114, a fleet management system 116 that is the same as or similar to fleet management system 116, a V2I system that is the same as or similar to V2I system 118, and/or the like).

[0060] In some embodiments, perception system 402 receives data associated with at least one physical object (e.g., data that is used by perception system 402 to detect the at least one physical object) in an environment and classifies the at least one physical object. In some examples, perception system 402 receives image data captured by at least one camera (e.g., cameras 202a), the image associated with (e.g., representing) one or more physical objects within a field of view of the at least one camera. In such an example, perception system 402 classifies at least one physical object based on one or more groupings of physical objects (e.g., bicycles, vehicles, traffic signs, pedestrians, and/or the like). In some embodiments, perception system 402 transmits data associated with the classification of the physical objects to planning system 404 based on perception system 402 classifying the physical objects.

[0061] In some embodiments, planning system 404 receives data associated with a destination and generates data associated with at least one route (e.g., routes 106) along which a vehicle (e.g., vehicles 102) can travel along toward a destination. In some embodiments, planning system 404 periodically or continuously receives data from perception system 402 (e.g., data associated with the classification of physical objects, described above) and planning system 404 updates the at least one trajectory or generates at least one different trajectory based on the data generated by perception system 402. In other words, planning system 404 may perform tactical function-related tasks that are required to operate vehicle 102 in on-road traffic. Tactical efforts involve maneuvering the vehicle in traffic during a trip, including but not limited to deciding whether and when to overtake another vehicle, change lanes, or selecting an appropriate speed, acceleration, deacceleration, etc. In some embodiments, planning system 404 receives data associated with an updated position of a vehicle (e.g., vehicles 102) from localization system 406 and planning system 404 updates the at least one trajectory or generates at least one different trajectory based on the data generated by localization system 406.

[0062] In some embodiments, localization system 406 receives data associated with (e.g., representing) a location of a vehicle (e.g., vehicles 102) in an area. In some examples, localization system 406 receives LiDAR data associated with at least one point cloud generated by at least one LiDAR sensor (e.g., LiDAR sensors 202b). In certain examples, localization system 406 receives data associated with at least one point cloud from multiple LiDAR sensors and localization system 406 generates a combined point cloud based on each of the point clouds. In these examples, localization system 406 compares the at least one point cloud or the combined point cloud to two- dimensional (2D) and/or a three-dimensional (3D) map of the area stored in database 410. Localization system 406 then determines the position of the vehicle in the area based on localization system 406 comparing the at least one point cloud or the combined point cloud to the map. In some embodiments, the map includes a combined point cloud of the area generated prior to navigation of the vehicle. In some embodiments, maps include, without limitation, high- precision maps of the roadway geometric properties, maps describing road network connectivity properties, maps describing roadway physical properties (such as traffic speed, traffic volume, the number of vehicular and cyclist traffic lanes, lane width, lane traffic directions, or lane marker types and locations, or combinations thereof), and maps describing the spatial locations of road features such as crosswalks, traffic signs or other travel signals of various types. In some embodiments, the map is generated in real-time based on the data received by the perception system.

[0063] In another example, localization system 406 receives Global Navigation Satellite System (GNSS) data generated by a global positioning system (GPS) receiver. In some examples, localization system 406 receives GNSS data associated with the location of the vehicle in the area and localization system 406 determines a latitude and longitude of the vehicle in the area. In such an example, localization system 406 determines the position of the vehicle in the area based on the latitude and longitude of the vehicle. In some embodiments, localization system 406 generates data associated with the position of the vehicle. In some examples, localization system 406 generates data associated with the position of the vehicle based on localization system 406 determining the position of the vehicle. In such an example, the data associated with the position of the vehicle includes data associated with one or more semantic properties corresponding to the position of the vehicle.

[0064] In some embodiments, control system 408 receives data associated with at least one trajectory from planning system 404 and control system 408 controls operation of the vehicle. In some examples, control system 408 receives data associated with at least one trajectory from planning system 404 and control system 408 controls operation of the vehicle by generating and transmitting control signals to cause a powertrain control system (e.g., DBW system 202h, powertrain control system 204, and/or the like), a steering control system (e g., steering control system 206), and/or a brake system (e.g., brake system 208) to operate. For example, control system 408 is configured to perform operational functions such as a lateral vehicle motion control or a longitudinal vehicle motion control. The lateral vehicle motion control causes activities necessary for the regulation of the y-axis component of vehicle motion. The longitudinal vehicle motion control causes activities necessary for the regulation of the x-axis component of vehicle motion. In an example, where a trajectory includes a left turn, control system 408 transmits a control signal to cause steering control system 206 to adjust a steering angle of vehicle 200, thereby causing vehicle 200 to turn left. Additionally, or alternatively, control system 408 generates and transmits control signals to cause other devices (e.g., headlights, turn signal, door locks, windshield wipers, and/or the like) of vehicle 200 to change states.

[0065] In some embodiments, perception system 402, planning system 404, localization system 406, and/or control system 408 implement at least one machine learning model (e.g., at least one multilayer perceptron (MLP), at least one convolutional neural network (CNN), at least one recurrent neural network (RNN), at least one autoencoder, at least one transformer, and/or the like). In some examples, perception system 402, planning system 404, localization system 406, and/or control system 408 implement at least one machine learning model alone or in combination with one or more of the above-noted systems. In some examples, perception system 402, planning system 404, localization system 406, and/or control system 408 implement at least one machine learning model as part of a pipeline (e.g., a pipeline for identifying one or more objects located in an environment and/or the like). An example of an implementation of a machine learning model is included below with respect to FIGS. 4B-4D.

[0066] Database 410 stores data that is transmitted to, received from, and/or updated by perception system 402, planning system 404, localization system 406 and/or control system 408. In some examples, database 410 includes a storage component (e.g., a storage component that is the same as or similar to storage component 308 of FIG. 3) that stores data and/or software related to the operation and uses at least one system of autonomous vehicle compute 400. In some embodiments, database 410 stores data associated with 2D and/or 3D maps of at least one area. In some examples, database 410 stores data associated with 2D and/or 3D maps of a portion of a city, multiple portions of multiple cities, multiple cities, a county, a state, a State (e.g., a country), and/or the like). In such an example, a vehicle (e.g., a vehicle that is the same as or similar to vehicles 102 and/or vehicle 200) can drive along one or more drivable regions (e.g., single-lane roads, multi-lane roads, highways, back roads, off road trails, and/or the like) and cause at least one LiDAR sensor (e.g., a LiDAR sensor that is the same as or similar to LiDAR sensors 202b) to generate data associated with an image representing the objects included in a field of view of the at least one LiDAR sensor. [0067] In some embodiments, database 410 can be implemented across a plurality of devices. In some examples, database 410 is included in a vehicle (e.g., a vehicle that is the same as or similar to vehicles 102 and/or vehicle 200), an autonomous vehicle system (e.g., an autonomous vehicle system that is the same as or similar to remote AV system 114, a fleet management system (e.g., a fleet management system that is the same as or similar to fleet management system 116 of FIG. 1, a V2I system (e.g., a V2I system that is the same as or similar to V2I system 118 of FIG. 1) and/or the like.

[0068] Referring now to FIG. 4B, illustrated is a diagram of an implementation of a machine learning model. More specifically, illustrated is a diagram of an implementation of a convolutional neural network (CNN) 420. For purposes of illustration, the following description of CNN 420 will be with respect to an implementation of CNN 420 by perception system 402. However, it will be understood that in some examples CNN 420 (e.g., one or more components of CNN 420) is implemented by other systems different from, or in addition to, perception system 402 such as planning system 404, localization system 406, and/or control system 408. While CNN 420 includes certain features as described herein, these features are provided for the purpose of illustration and are not intended to limit the present disclosure.

[0069] CNN 420 includes a plurality of convolution layers including first convolution layer 422, second convolution layer 424, and convolution layer 426. In some embodiments, CNN 420 includes sub-sampling layer 428 (sometimes referred to as a pooling layer). In some embodiments, sub-sampling layer 428 and/or other subsampling layers have a dimension (i.e., an amount of nodes) that is less than a dimension of an upstream system. By virtue of sub-sampling layer 428 having a dimension that is less than a dimension of an upstream layer, CNN 420 consolidates the amount of data associated with the initial input and/or the output of an upstream layer to thereby decrease the amount of computations necessary for CNN 420 to perform downstream convolution operations. Additionally, or alternatively, by virtue of sub-sampling layer 428 being associated with (e.g., configured to perform) at least one subsampling function (as described below with respect to FIGS. 4C and 4D), CNN 420 consolidates the amount of data associated with the initial input.

[0070] Perception system 402 performs convolution operations based on perception system 402 providing respective inputs and/or outputs associated with each of first convolution layer 422, second convolution layer 424, and convolution layer 426 to generate respective outputs. In some examples, perception system 402 implements CNN 420 based on perception system 402 providing data as input to first convolution layer 422, second convolution layer 424, and convolution layer 426. In such an example, perception system 402 provides the data as input to first convolution layer 422, second convolution layer 424, and convolution layer 426 based on perception system 402 receiving data from one or more different systems (e g., one or more systems of a vehicle that is the same as or similar to vehicle 102), a remote AV system that is the same as or similar to remote AV system 114, a fleet management system that is the same as or similar to fleet management system 116, a V2I system that is the same as or similar to V2I system 118, and/or the like). A detailed description of convolution operations is included below with respect to FIG. 4C. [0071] In some embodiments, perception system 402 provides data associated with an input (referred to as an initial input) to first convolution layer 422 and perception system 402 generates data associated with an output using first convolution layer 422. In some embodiments, perception system 402 provides an output generated by a convolution layer as input to a different convolution layer. For example, perception system 402 provides the output of first convolution layer 422 as input to sub-sampling layer 428, second convolution layer 424, and/or convolution layer 426. In such an example, first convolution layer 422 is referred to as an upstream layer and sub-sampling layer 428, second convolution layer 424, and/or convolution layer 426 are referred to as downstream layers. Similarly, in some embodiments perception system 402 provides the output of sub-sampling layer 428 to second convolution layer 424 and/or convolution layer 426 and, in this example, sub-sampling layer 428 would be referred to as an upstream layer and second convolution layer 424 and/or convolution layer 426 would be referred to as downstream layers.

[0072] In some embodiments, perception system 402 processes the data associated with the input provided to CNN 420 before perception system 402 provides the input to CNN 420. For example, perception system 402 processes the data associated with the input provided to CNN 420 based on perception system 402 normalizing sensor data (e.g., image data, LiDAR data, radar data, and/or the like).

[0073] In some embodiments, CNN 420 generates an output based on perception system 402 performing convolution operations associated with each convolution layer. In some examples, CNN 420 generates an output based on perception system 402 performing convolution operations associated with each convolution layer and an initial input. In some embodiments, perception system 402 generates the output and provides the output as fully connected layer 430. In some examples, perception system 402 provides the output of convolution layer 426 as fully connected layer 430, where fully connected layer 430 includes data associated with a plurality of feature values referred to as Fl, F2 . . . FN. In this example, the output of convolution layer 426 includes data associated with a plurality of output feature values that represent a prediction.

[0074] In some embodiments, perception system 402 identifies a prediction from among a plurality of predictions based on perception system 402 identifying a feature value that is associated with the highest likelihood of being the correct prediction from among the plurality of predictions. For example, where fully connected layer 430 includes feature values Fl, F2, . . . FN, and Fl is the greatest feature value, perception system 402 identifies the prediction associated with Fl as being the correct prediction from among the plurality of predictions. In some embodiments, perception system 402 trains CNN 420 to generate the prediction. In some examples, perception system 402 trains CNN 420 to generate the prediction based on perception system 402 providing training data associated with the prediction to CNN 420.

[0075] Referring now to FIGS. 4C and 4D, illustrated is a diagram of example operation of CNN 440 by perception system 402. In some embodiments, CNN 440 (e.g., one or more components of CNN 440) is the same as, or similar to, CNN 420 (e.g., one or more components of CNN 420) (see FIG. 4B).

[0076] At step 450, perception system 402 provides data associated with an image as input to CNN 440 (step 450). For example, as illustrated, perception system 402 provides the data associated with the image to CNN 440, where the image is a greyscale image represented as values stored in a two-dimensional (2D) array. In some embodiments, the data associated with the image may include data associated with a color image, the color image represented as values stored in a three- dimensional (3D) array. Additionally, or alternatively, the data associated with the image may include data associated with an infrared image, a radar image, and/or the like.

[0077] At step 455, CNN 440 performs a first convolution function. For example, CNN 440 performs the first convolution function based on CNN 440 providing the values representing the image as input to one or more neurons (not explicitly illustrated) included in first convolution layer 442. In this example, the values representing the image can correspond to values representing a region of the image (sometimes referred to as a receptive field). In some embodiments, each neuron is associated with a filter (not explicitly illustrated). A filter (sometimes referred to as a kernel) is representable as an array of values that corresponds in size to the values provided as input to the neuron. In one example, a filter may be configured to identify edges (e.g., horizontal lines, vertical lines, straight lines, and/or the like). In successive convolution layers, the filters associated with neurons may be configured to identify successively more complex patterns (e.g., arcs, objects, and/or the like).

[0078] In some embodiments, CNN 440 performs the first convolution function based on CNN 440 multiplying the values provided as input to each of the one or more neurons included in first convolution layer 442 with the values of the filter that corresponds to each of the one or more neurons. For example, CNN 440 can multiply the values provided as input to each of the one or more neurons included in first convolution layer 442 with the values of the filter that corresponds to each of the one or more neurons to generate a single value or an array of values as an output. In some embodiments, the collective output of the neurons of first convolution layer 442 is referred to as a convolved output. In some embodiments, where each neuron has the same filter, the convolved output is referred to as a feature map.

[0079] In some embodiments, CNN 440 provides the outputs of each neuron of first convolutional layer 442 to neurons of a downstream layer. For purposes of clarity, an upstream layer can be a layer that transmits data to a different layer (referred to as a downstream layer). For example, CNN 440 can provide the outputs of each neuron of first convolutional layer 442 to corresponding neurons of a subsampling layer. In an example, CNN 440 provides the outputs of each neuron of first convolutional layer 442 to corresponding neurons of first subsampling layer 444. In some embodiments, CNN 440 adds a bias value to the aggregates of all the values provided to each neuron of the downstream layer. For example, CNN 440 adds a bias value to the aggregates of all the values provided to each neuron of first subsampling layer 444. In such an example, CNN 440 determines a final value to provide to each neuron of first subsampling layer 444 based on the aggregates of all the values provided to each neuron and an activation function associated with each neuron of first subsampling layer 444.

[0080] At step 460, CNN 440 performs a first subsampling function. For example, CNN 440 can perform a first subsampling function based on CNN 440 providing the values output by first convolution layer 442 to corresponding neurons of first subsampling layer 444. In some embodiments, CNN 440 performs the first subsampling function based on an aggregation function. In an example, CNN 440 performs the first subsampling function based on CNN 440 determining the maximum input among the values provided to a given neuron (referred to as a max pooling function). In another example, CNN 440 performs the first subsampling function based on CNN 440 determining the average input among the values provided to a given neuron (referred to as an average pooling function). In some embodiments, CNN 440 generates an output based on CNN 440 providing the values to each neuron of first subsampling layer 444, the output sometimes referred to as a subsampled convolved output.

[0081] At step 465, CNN 440 performs a second convolution function. In some embodiments, CNN 440 performs the second convolution function in a manner similar to how CNN 440 performed the first convolution function, described above. In some embodiments, CNN 440 performs the second convolution function based on CNN 440 providing the values output by first sub sampling layer 444 as input to one or more neurons (not explicitly illustrated) included in second convolution layer 446. In some embodiments, each neuron of second convolution layer 446 is associated with a filter, as described above. The filter(s) associated with second convolution layer 446 may be configured to identify more complex patterns than the filter associated with first convolution layer 442, as described above.

[0082] In some embodiments, CNN 440 performs the second convolution function based on CNN 440 multiplying the values provided as input to each of the one or more neurons included in second convolution layer 446 with the values of the filter that corresponds to each of the one or more neurons. For example, CNN 440 can multiply the values provided as input to each of the one or more neurons included in second convolution layer 446 with the values of the filter that corresponds to each of the one or more neurons to generate a single value or an array of values as an output.

[0083] In some embodiments, CNN 440 provides the outputs of each neuron of second convolutional layer 446 to neurons of a downstream layer. For example, CNN 440 can provide the outputs of each neuron of first convolutional layer 442 to corresponding neurons of a subsampling layer. In an example, CNN 440 provides the outputs of each neuron of first convolutional layer 442 to corresponding neurons of second subsampling layer 448. In some embodiments, CNN 440 adds a bias value to the aggregates of all the values provided to each neuron of the downstream layer. For example, CNN 440 adds a bias value to the aggregates of all the values provided to each neuron of second subsampling layer 448. In such an example, CNN 440 determines a final value to provide to each neuron of second subsampling layer 448 based on the aggregates of all the values provided to each neuron and an activation function associated with each neuron of second subsampling layer 448. [0084] At step 470, CNN 440 performs a second subsampling function. For example, CNN 440 can perform a second subsampling function based on CNN 440 providing the values output by second convolution layer 446 to corresponding neurons of second subsampling layer 448. In some embodiments, CNN 440 performs the second subsampling function based on CNN 440 using an aggregation function. In an example, CNN 440 performs the first subsampling function based on CNN 440 determining the maximum input or an average input among the values provided to a given neuron, as described above. In some embodiments, CNN 440 generates an output based on CNN 440 providing the values to each neuron of second subsampling layer 448.

[0085] At step 475, CNN 440 provides the output of each neuron of second subsampling layer 448 to fully connected layers 449. For example, CNN 440 provides the output of each neuron of second sub sampling layer 448 to fully connected layers 449 to cause fully connected layers 449 to generate an output. In some embodiments, fully connected layers 449 are configured to generate an output associated with a prediction (sometimes referred to as a classification). The prediction may include an indication that an object included in the image provided as input to CNN 440 includes an object, a set of objects, and/or the like. In some embodiments, perception system 402 performs one or more operations and/or provides the data associated with the prediction to a different system, described herein.

[0086] FIGS. 5A-5D show implementations 500A-500D of a camera to LiDAR calibration and/or validation model. Calibration ensures the measurement accuracy of an instrument satisfies a known standard. Validation ensures a process or equipment operates according to its stated operating specifications. In some embodiments, calibration ensures that image data as captured by a camera and point cloud data as captured by a LiDAR accurately identify objects in the environment. Additionally, in some embodiments, validation ensures that image data as captured by a camera and point cloud data as captured by a LiDAR accurately detect objects such that the calibration parameters of the respective camera or LiDAR are within an acceptable tolerance range.

[0087] The present techniques build a camera-to-LiDAR calibration and validation model by representing data from sensors and devices in a unified data representation. In examples, a scene refers to a sequence of continuous action represented by various sensor modalities. The sensor modalities may be, for example, a sensor suite of an autonomous system, such as autonomous system 202 including one or more devices such as cameras 202a, LiDAR sensors 202b, radar sensors 202c, and microphones 202d. Autonomous vehicles, as well as other robotic systems, rely on sensors to perceive their environment. Most systems use a variety of sensors, including cameras, LiDARs, as well as radars. Fusing data from multiple sensors is used to leverage the advantages of each sensor. In examples, to fuse sensor data the precise value of their extrinsic calibration is used. This precise value is a transformation in the Lie group SE(3), represented by a 6-degree of freedom (DoF) transformation consisting of the relative translation and rotation between the two sensor poses. Camera-to-LiDAR (C2L) calibration is a sensor fusion strategy that combines the visual information obtained by cameras with the spatial and occupancy information obtained by LiDARs. With this information, downstream tasks such as mapping, localization, and planning can be carried out.

[0088] Traditionally, C2L calibration is performed with offline methods. This usually involves pre-collecting a series of frame in a controlled environment to carry out iterative optimization. However, this does not consider the scenario where sensor position changes during the normal operation of a robot, by weather conditions, mechanical vibrations, or collision. This implies the need for online methods that can perform calibration in real-time. Existing solutions to perform calibration in real-time rely on geometric feature detection and optimization. Traditional deep learning methods include dual- branch architectures that process the camera and LiDAR features separately before passing the features through a matching layer before regressing the output transformation. For the traditional deep learning methods, scenes are collected by a single camera located at the front of the vehicle.

[0089] The present techniques include a camera-to-LiDAR calibration and validation model. The camera-to-LiDAR calibration and validation model establishes an architecture that can be used as a baseline for future works. Unlike the traditional dual-branch architecture, the camera-to-LiDAR calibration and validation model as described herein merges the raw inputs directly (e.g., unified data representation) and relies on a single backbone to perform the calibration and/or validation task. Its single-branch architecture enables the model to be more lightweight. It is also extensible as feature encoders are used to modify the input. The camera-to-LiDAR calibration and validation model uses a transformers-based backbone, (e.g., MobileViT) to leverage self-attention mechanisms and identify areas with the most significant features. In some embodiments the camera-to-LiDAR calibration and validation model is evaluated using multiple datasets, such as a Kitti dataset. The camera-to-LiDAR calibration and validation model is also evaluated using handcrafted datasets, such as datasets collected from multiple autonomous vehicle logs, referred to drivelog data. Unlike Kitti, the drivelog data includes varying vehicles and features a suite of cameras surrounding the vehicle which provides various points of view. In examples, the camera- to-LiDAR calibration and validation model enables real-time calibration and/or validation, with the output parameters having being on par or better compared to traditional techniques.

[0090] In some embodiments, the camera-to-LiDAR calibration and validation model performs the task of C2L calibration validation. The task is to determine the validity of a given set of calibration parameters for a pair of sensors. This is relevant for autonomous vehicle use cases as it can be used to determine instantly if a vehicle is well-calibrated before allowing it to operate. This problem can be reduced to a binary classification task (calibrated/de-calibrated), where the calibrated class is defined as having a de-calibration smaller than a chosen sensitivity margin. Training a camera-to-LiDAR calibration and validation model for a calibration task and then using transfer learning with a classification head to determine an output of classification (calibrated/de- calibrated) with enables better calibration than directly learning this classification with existing calibration architectures.

[0091] Accordingly, the present techniques enable a camera-to-LiDAR calibration and validation model including a lightweight single-branch architecture which has up to 3 to 10 times less parameters than traditional architectures. The camera-to-LiDAR calibration and validation model achieves up to 7 times lower mean average error compared to RegNet. The camera-to-LiDAR calibration and validation model enables an early fusion of all inputs in a unified data representation. Further, the camera-to-LiDAR calibration and validation model leverages transformers and self-attention to learn meaningful features from unstructured environments. A transfer learning technique as described herein improves camera-to- LiDAR calibration validation by training a classification head on top of frozen calibration network weights, achieving 98% accuracy. In examples, transfer learning is applied to the trained single branch backbone network and trained regression head to form a camera-to-LiDAR validation model that obtains as input a unified data representation and outputs validation parameters in real time.

[0092] Referring now to FIG. 5A, an implementation 500A using a camera to LiDAR calibration and validation model is shown. In the implementation 500A, a user device 550A is shown. The user device 550A is, for example, a device such as a client device, a mobile device (e.g., a cellular telephone, a tablet, and/or the like). The user device 55OA transmits data 510A associated with a request for services associated with the vehicle, such as autonomous package delivery, robotaxi services, or any combinations thereof. In examples, the data 510A includes a date, time, starting location, ending, location, and user identification. In examples, the data 510A is transmitted to a V2I device, a remote AV system, a fleet management system, and/or V2I system 118 via a network. In examples, the V2I device, remote AV system, fleet management system, V2I system, and network are the same as or similar to the V2I device 110, remote AV system 114, fleet management system 116, V2I system 118, and network 112 shown in FIG. 1.

[0093] In examples, the data 510A is obtained by or transmitted to a vehicle 502A. In examples, the vehicle 502A is the same as or similar to vehicles 102 of FIG. 1 and/or vehicle 200 of FIG. 2. The vehicle 502A includes an AV compute 400A. In examples, the AV compute 400A is the same as or similar to the AV compute of FIG. 4. In some embodiments, the AV compute 400A includes a perception system (e.g., perception system 402 of FIG. 4) that detects the at least one physical object) in an environment and classifies the at least one physical object. In such an example, perception system classifies at least one physical object (e.g., objects 104a-104n of FIG. 1) based on one or more groupings of physical objects (e.g., bicycles, vehicles, traffic signs, pedestrians, and/or the like). In some embodiments, perception system transmits data associated with the classification of the physical objects to a planning system (e.g., planning system 404 of FIG. 4) based on perception system 402 classifying the physical objects. In examples, the planning system determines a trajectory for the vehicle 502A based on, at least in part, the data output by the perception system.

[0094] Referring now to FIG. 5B shows an implementation 500B using a camera to LiDAR calibration and validation model. In the implementation 500B, an AV compute 400B of a vehicle 502B is shown. In examples, the AV compute 400B is the same as or similar to the AV compute 400 of FIG. 4. A request for transport 512B is sent to a planning system 404B. In examples, the planning system 404B is the same as or similar to the planning system 404 of FIG. 4. In examples, the request for transport 512B is a request for services associated with the vehicle 502B, such as autonomous package delivery, robotaxi services, or any combinations thereof. In examples, the request for transport 512B includes a date, time, starting location, ending, location, and user identification. In examples, the request for transport 512B is obtained from a V2I device, a remote AV system, a fleet management system, and/or V2I system 118 via a network. In examples, the V2I device, remote AV system, fleet management system, V2I system, and network are the same as or similar to the V2I device 110, remote AV system 114, fleet management system 116, V2I system 118, and network 112 shown in FIG. 1.

[0095] In examples, the request for transport 512B obtained by or transmitted to a planning system. Similar to the planning system of the AV compute 400A (FIG. 5A), the planning system 404B performs tactical function-related tasks to operate vehicle on-road traffic and off-road traffic. In such an example, perception system 402B classifies at least one physical object (e.g., objects 104a-104n of FIG. 1) based on one or more groupings of physical objects (e.g., bicycles, vehicles, traffic signs, pedestrians, and/or the like). In some embodiments, perception system 402B transmits data associated with the classification of the physical objects to a planning system 404B based on perception system 402B classifying the physical objects. In examples, the planning system 404B determines a trajectory for the vehicle 502B based on, at least in part, the data output by the perception system.

[0096] Referring now to FIG. 5C shows an implementation 500C using a camera to LiDAR calibration and validation model. In the implementation 500C, an AV compute 400C of a vehicle 502C is shown. In examples, the AV compute 400C is the same as or similar to the AV compute 400 of FIG. 4. A route 514C is determined by a planning system 404C. In examples, the planning system 404C is the same as or similar to the planning system 404 of FIG. 4. In examples, the route 514C is generated in response to a request for services associated with the vehicle 502C, such as autonomous package delivery, robotaxi services, or any combinations thereof. In examples, the route 514C is the same as or similar to the routes 106a-106n of FIG. 1. In examples, the route 514C is obtained from a V2I device, a remote AV system, a fleet management system, and/or V2I system 118 via a network. In examples, the V2I device, remote AV system, fleet management system, V2I system, and network are the same as or similar to the V2I device 110, remote AV system 114, fleet management system 116, V2I system 118, and network 112 shown in FIG. 1.

[0097] In examples, the route 514C is generated by a planning system. Similar to the planning systems of the AV compute 400A (FIG. 5A) and the AV compute 400B (FIG. 5B), the planning system 404C performs tactical function-related tasks to operate vehicle on-road traffic and offroad traffic. A route is transmitted (516C) from the planning system to a control system 408C. In examples, the control system 408C is the same as or similar to the control system 408 of FIG. 4. In examples, the control system 408C receives data associated with at least one route or trajectory from planning system 404C and control system 408C controls operation of the vehicle. In some embodiments, a perception system transmits data associated with the classification of the physical objects to a planning system 404C based on classifications of the physical objects output by the perception system. In examples, the planning system 404C determines a route 514C for the vehicle 502C based on, at least in part, the data output by the perception system. Similar to the implementations 500A and the implementation 500B, the implementation 500C accurately classifies objects by using the perception system 402B that relies on sensor data captured by a sensor suite of an autonomous system with varying modalities.

[0098] Referring now to FIG. 5D shows an implementation 500D using a camera to LiDAR calibration and validation model. In the implementation 500D, an AV compute 400D of a vehicle 502D is shown. In examples, the AV compute 400D is the same as or similar to the AV compute 400 of FIG. 4. A control signal 518D is generated by a control system 408D. In examples, the control system 408D is the same as or similar to the control system 408 of FIG. 4. In examples, the control system 408D receives data associated with at least one route or trajectory a planning system, such as the planning system 404 of FIG. 4. In some embodiments, perception system transmits data associated with the classification of the physical objects to a planning system (e.g., planning system 404 of FIG. 4) based on perception system classifying the physical objects. In examples, the planning system determines a trajectory for the vehicle 502D based on, at least in part, the data output by the perception system. In examples, the control signal 518D is generated in response to a request for services associated with the vehicle 502D, such as autonomous package delivery, robotaxi services, or any combinations thereof. In examples, control signal 520D is transmitted to a DBW system, such as the DBW system 202h of FIG. 2. In some examples, control system 408D receives data associated with at least one trajectory from a planning system and control system 408D controls operation of the vehicle 502D by generating and transmitting control signals to cause a powertrain control system (e.g., DBW system 202h, powertrain control system 204 of FIG. 2, and/or the like), a steering control system (e.g., steering control system 206 of FIG. 2), and/or a brake system (e.g., brake system 208 of FIG. 2) to operate. In examples, the control signal 518D is generated by the control system 408D based on, at least in part, data from a planning system such as the AV computes 400, 400A, 400B, and 400C of FIGs. 4-5C. The control signal 518D is based on sensor data captured by a sensor suite of the autonomous system, such as the vehicle 502D. [0099] In the implementations 500A, 500B, 500C, and 500D, in order to accurately classify objects, the perception system relies on sensor data captured by a sensor suite of an autonomous system with varying modalities, such as autonomous system 202 including one or more devices such as cameras 202a, LiDAR sensors 202b, radar sensors 202c, and microphones 202d. In examples, sensor data from sensors with varying modalities is represented in a dataset, such as a publicly available dataset or a handcrafted dataset. For example, the sensor data is captured by cameras, LiDARs, radars, microphones or any combinations thereof. In some embodiments, the sensor data is transformed into a unified data representation are used to train a single backbone network with a regression head to output a binary classification of calibrated/de-calibrated and/or validated/de-validated. In this manner, the camera to LiDAR calibration and validation model performs an early fusion of the input image data (captured by cameras) and point cloud data (captured by LiDARs) by aggregating camera image channels and LiDAR mappings into a multichannel unified data representation before extracting their features jointly with a single-branch architecture. Its single-branch architecture makes the camera to LiDAR calibration and validation model lightweight, which is desirable in applications with restrained resources such as autonomous driving. As described herein, the camera to LiDAR calibration and validation model achieves improved results compared to traditional methods. Moreover, through transfer learning, weights learned on the calibration task can be applied to a calibration validation task without re-training. [0100] The camera-to-LiDAR calibration and validation model as described herein enables sensor calibration, including camera-to-LiDAR (C2L) extrinsic calibration. Traditionally, sensor calibration can be target based sensor calibration, targetless sensor calibration, and machine learning based. Target based sensor calibration is often solved with offline, target-based methods. In examples, target based sensor calibration uses a checkerboard target, which can be seen by both LiDAR and camera. In examples, target based sensor calibration explores different target shapes. For example, traditional techniques are limited to target shapes, such as a circle and a hole. Some traditional techniques are limited to calibration with multiple checkerboard targets. Traditional target-based methods are limited to using specific equipment and environment. Moreover, users of traditional target-based methods should gain experience for optimal target positioning. In traditional target-based methods, a large number of parameters are tuned, which introduces heuristics and requires iterative optimization which takes time to converge. Traditional targetbased methods are slow and costly to use regularly on a vehicle. In examples, traditional targetless methods are limited to using structure from motion and semantic information to treat the task as a point cloud registration problem. Other traditional targetless methods use natural edge features in both modalities to align them. Accordingly, traditional targetless methods rely on computationheavy optimization and require a feature rich environment to provide a correct distribution of the selected features.

[0101] Traditional machine learning methods such as RegNet are limited to training using generating artificially de-calibrated samples from the ground truth. Traditional machine learning methods also are limited to a dual-branch architecture, using one branch for camera feature extraction from RGB image and a separate branch for LiDAR feature extraction from projected depth. Features are then merged at a later stage before being matched by a third backbone and finally regressed by the head. Moreover, traditional machine learning methods are limited to multi-scale refinement passing the input successively through different networks at the time of inference to refine the output.

[0102] Traditional target based sensor calibration, targetless sensor calibration, and machine learning based sensor calibration achieve calibration but leave room for improvement in accuracy, performance, and reliability in unstructured environments. The networks used are typically large as they follow a dual-branch architecture which requires two separate feature encoders for the camera branch and the LiDAR branch. As previous publications evaluate their results on different splits of the Kitti dataset, Table I and Table II show that the results can vary significantly depending on the dataset split chosen.

[0103] The present techniques enable camera-to-LiDAR calibration and validation model. In some embodiments, the model is a trained deep learning model that performs calibration and validation. In some embodiments, the model is a trained deep learning model that performs calibration or validation. In examples, the present techniques use a Transformer-based networks for the task of camera-to-LiDAR calibration. The data is transformed to a unified data representation. To perform the calibration task, data is obtained in the space of autonomous vehicles consisting of driving scenes. Each scene sample is a synchronized capture of a camera image and a LiDAR point cloud. For training, the corresponding sensor input pair and the “ground truth” extrinsic calibration between them are obtained. In examples, the extrinsic calibration is obtained in a simulation. In examples, the intrinsic parameters are already calibrated. For example, the intrinsic parameters are predetermined by a manufacturer of the sensor or device. [0104] In some embodiments, samples are generated from a dataset. In examples, the dataset consists of calibrated scenes. De-calibrated samples are generated to feed as input to camera-to- LiDAR calibration and validation model. A random de-calibration is generated on all 6 DoF transformation parameters according to a uniform distribution on a chosen de-calibration range. In examples, the range 10cm is selected for translation and 1° is selected for rotation, which is an estimate of the levels of perturbation experienced by the sensors during vehicle operation. Additionally, T a transformation with three rotation parameters and three translation parameters. This de-calibration Tdecai is then applied to the ground truth T_gt to get the de-calibrated initial transformation mt, as explained in Eq. 1. The LiDAR points are projected into the camera frame using these de-calibrated parameters. FIG. 6 shows the projection of LiDAR points onto a corresponding camera image. As shown in FIG. 6, the top image 602 is de-calibrated on rotation and translation parameters. The bottom image is re-calibrated with a camera-to-LiDAR calibration and validation model as described herein.

[0105] Using Tint as the initial calibration value, camera-to-LiDAR calibration and validation model estimates the transformation Tdecai as T_pred, such that applying T ' 'pred mt should result in the ground truth value T_gras described in Eq. 2.

Tinit = T_decaiT_gt (1)

Tgt ~ T_p^_edT_init (2)

[0106] The camera-to-LiDAR calibration and validation model uses a unified data representation as input. For example, the unified data representation includes sensors of varying modalities. Features are extracted directly from this unified data representation. This unified data representation is an N-channel pseudo-image. Each channel corresponds to a different input source from a sensor (at least one channel per sensor). An advantage of this representation is that it requires fewer parameters than a dual-branch architecture, hence the camera-to-LiDAR calibration and validation model would require less computational resources. Moreover, it also makes the camera-to-LiDAR calibration and validation model more modular and extensible. The model can easily be extended to experiment with new input sources, whether they are from a different sensor (to try to solve a different calibration task) or a processed input from the same sensor (for example adding edge extractions on the camera image). In examples, the camera-to-LiDAR calibration and validation model uses grayscale from the cameras, and depth and intensity from the LiDARs. A unified data representation, such as a pseudo-image with 3 channels, is generated as input to the camera-to-LiDAR calibration and validation model.

[0107] FIG. 7 shows an architecture of a camera-to-LiDAR calibration and/or validation model. As shown in the example of FIG. 7, a camera-to-LiDAR calibration and/or validation model 700 includes a backbone network and a regression head network. The regression head obtains as input an image 704. In examples, the image 704 is an N-channel pseudo-image. The image 704 is generated from sensor data, such as cameras 202a and LiDAR 202b. In examples, the cameras 202a and LiDAR 202b are the same as or similar to cameras 202a and LiDAR 202b of FIG. 2, respectively. The camera captures data associated with the environment in a first modality. In examples, the camera outputs a grayscale image of dimension IxHxW. The LiDAR captures data associated with the environment in a second modality. In examples, the LiDAR outputs a point cloud with dimension Nx4. In examples, the LiDAR outputs a LiDAR map with dimension 2xHxW. The LiDAR map includes depth and intensity data as captured by LiDARs 202b.

[0108] As shown in FIG. 7, the camera-to-LiDAR calibration and validation model 700 outputs predicted transformation parameters 710. Predicted transformation parameters 710 are compared to an initial calibration value 712. The camera-to-LiDAR calibration and validation model 700 estimates the 3D transformation. In examples, the camera-to-LiDAR calibration and validation 700 is created to be a lightweight, modular, and efficient baseline that can easily be extended.

[0109] The unified data representation shown by image 704 enables the use of a single-branch architecture. camera-to-LiDAR calibration and validation model 700 directly learns to match the different modalities with a single branch backbone network as opposed to the traditional dualbranch architectures that use three separate backbones for feature extraction and feature matching. This makes the camera-to-LiDAR calibration and validation model 700 lighter in comparison. Moreover, the architecture of the camera-to-LiDAR calibration and validation model 700 is not customized for the input type. The use of a unified data representation means that changing the inputs does not results in other changes to the backbone. In some examples, the channel number of the backbone is expanded based on a number of inputs.

[0110] In examples, the backbone 701 has a transformer-based architectures that leverages selfattention mechanisms to vision tasks and learns jointly on different modalities using convolutional operations. For example, MobileViT is used as the backbone 701. In examples, the backbone 701 implements a convolutional operation in which the local matrix multiplication is replaced by a global operation through a stack of transformer layers. It combines advantages from both convolutional networks (such as spatial bias) and transformers (self-attention). Moreover, it was designed to be lightweight and to run on embedded systems with constrained resources. As a result, the camera-to-LiDAR calibration and validation model 700 is more lightweight, with approximately 5.7 million trainable parameters. As a comparison, a traditional models such as ResNetl 8 backbone has around 11 million trainable parameters. A network using a ResNetl 8- based backbone in a 2-branch architecture, could have up to 33 million parameters for its backbone alone.

[0U1] In examples, the head 703 is a regression head represented by fully connected layers to regress the calibration parameters which are 3 translation parameters (x, y, z) and 3 rotation parameters (roll, pitch, yaw). The regression head 703 consists of a common first layer which then splits into two branches to separately regress the translation and rotation components.

[0112] In examples, regression loss 708 is implemented. The camera-to-LiDAR calibration and validation model 700 is trained using supervised learning. In examples, Mean Square Error regression losses are used for rotation as in Eq. 3 and translation as in Eq. 4 to compare the prediction and the ground truth de-calibration. In Eq. 3 and Eq. 4, r and t are respectively the rotation and translation parameters of the transformation. Both losses in Eq. 3 and Eq. 4 are then be averaged for the batch.

[0113] In examples, spatial losses 706 are implemented. Spatially-aware losses are used to improve convergence during training. Two such losses are used to compare the correct point cloud and the point cloud after re-calibration. The first loss is center loss 705. In examples, center loss 705 is the distance between the center of those two point clouds as in Eq. 5 where C_Pd is the center of the point cloud. This loss must then be averaged for the batch. c ⁼ (T_gtC_pci — T_pre_dTi_nitC_pcl)² (5)

[0114] The second loss is point cloud loss 707. Point cloud loss is the distance between the corresponding points in those two point clouds (there is no need for matching as the data remains ordered) as in Eq. 6 where pk is a point from the point cloud with index k, and K the number of points in the point cloud. This loss must then be averaged for the batch.

[0115] In examples, the backbone 701 it is adapted to embedded applications and brings benefits of both convolutional networks and transformers. Self-attention mechanisms used in transformers enable the camera-to-LiDAR calibration and validation model 700 to give more weight to features deemed more relevant by the model. Attention also brings some form of explainability as it can be displayed as a heatmap for us to see the areas the network found most relevant to solve the calibration task. The heatmaps generated by the camera-to-LiDAR calibration and validation model 700 during testing are illustrated in FIG. 8. Images 802 and 804 are attention heatmaps overlayed on images from our datasets. Images are converted to grayscale and resized. In examples, image 802 is from a publicly available dataset. Image 804 is available from a handcrafted dataset.

[0116] In some examples, the camera-to-LiDAR calibration and validation model 700 weights its attention in different zones depending on the dataset on which it was trained. For examples, on a first dataset, the camera-to-LiDAR calibration and validation model 700 gives more attention weight to roads and cars. On a second dataset, the camera-to-LiDAR calibration and validation model 700 tends to highlight various salient objects in the image, especially objects that offer clear lines visible with both sensors. Those zones with high attention closely resemble those that humans experimented with calibrating would look at to spot de-calibration. Accordingly, in some embodiments the camera-to-LiDAR calibration and validation model 700 relies on elements found on the road, such as when trained using datasets with the a front camera where the road is always visible. In some embodiments, the camera-to-LiDAR calibration and validation model 700 is trained using datasets where the road is not visible.

[0117] The camera-to-LiDAR calibration and validation model 700 can be implemented using various datasets. For example, a publicly available dataset is Kitti, a reference dataset on the autonomous driving scene. Different splits of the Kitti dataset are shown in Table I.

[0118] From the splits presented in Table I, choose a as a reference. As shown by results in Table II a uses samples from separate days for training/validation and testing, with different camera intrinsics. In comparison, splits y, and 8 include in their testing sets samples recorded the same day and with the same camera intrinsics as their training sets. As shown in Table II, 4 uses spatially redundant samples (some scenes are captured in the same location in training and testing sets). Table II shows that the same network, such as the camera-to-LiDAR calibration and validation model 700, when trained and tested on different splits will perform differently. The splits with more similarities between the training and testing sets will get seemingly more accurate and precise results, whether those similarities are mostly in the camera intrinsics as in ?, or even in the location where the samples were collected, as in 3.

[0119] In examples, the camera-to-LiDAR calibration and validation model 700 is implemented using a handcrafted dataset. For example, a handcrafted dataset is generated from autonomous vehicle driving logs captured by autonomous vehicles driving across cities e.g., (Las Vegas, Santa Monica, Pittsburgh, Boston). In examples, 89 driving logs were captured consisting of 28 unique vehicles. The driving log data includes, for example, data from one main LiDAR and 8 different cameras surrounding the vehicle. The handcrafted dataset is split into 21995 training data, 3299 validation data, and 2173 testing data. In examples, training and validation data consists of logs taken from certain locations (e.g., only Las Vegas and Santa Monica) whereas the testing data has logs from each location. In examples, the testing data has no overlapping vehicles with the training set.

[0120] Further, in constructing a camera-to-LiDAR calibration and validation model 700, soft augmentation parameters are used to fit real-life situations as much as possible. For example, an assumption is that the vehicle will lie flat on the road and the ground will be approximately horizontal or rotated with a limited angle. We thus augmented with random rotations of up to 2°, translation of 0.01% of the image dimensions.

[0121] The results of camera-to-LiDAR calibration and validation models (e.g., UniCai) trained using various datasets are shown in Table II. For a first publicly available dataset, such as Kitti, compare in Table II the results obtained on the different splits presented in Table I. The models compared have been trained to solve the task on different de-calibration ranges. However, they rely on multi-scale refinement to bring the final task to the same de-calibration range: a cascade of networks are trained on different ranges of de-calibrations with each step refining the parameters to a range suitable for the next network. The final network, trained on the smallest de-calibration range, receives parameters that are already in its range. This last network can thus be considered independently as operations happening beforehand are transparent to it. It is then possible to compare results from this final network with results from a network trained on the same decalibration range.

[0122] MAE: Mean Average Error; STD: Standard Deviation; TL: Transfer Learning

[0123] On Kitti, observe that UniCai outperforms most net- works from the state of the art, notably RegNet, Net-Calib, and CalibNet. LCCNet relies on multiple scenes for each prediction to refine its results. By contrast, the camera-to-LiDAR calibration and validation model 700 (e.g., UniCai) estimates the calibration parameters with a single shot.

[0124] The results of camera-to-LiDAR calibration and validation models (e.g., UniCai) trained using handcrafted data is described herein. In examples, handcrafted data is captured from multiple vehicles, captured in different conditions and locations, and intrinsic calibration, as well as extrinsic ground truth, can be imperfect. Further, in examples handcrafted data using data from all available cameras on the vehicle compared to only one camera in Kitti experiments and other traditional techniques. The camera-to-LiDAR calibration and validation model 700 calibrates multiple cameras that have wildly different points of view and positions on the vehicle (front, back, sides, etc.). The present techniques achieve a mean average error of 0.13° on rotation and 1.9cm on translation. The present techniques are at least twice more accurate than the 0.28° and 6cm achieved by RegNet on Kitti for a single camera.

[0125] In examples, the camera-to-LiDAR calibration and validation model executes from a graphics card or a graphics processing unit. For example, by profiling the camera-to-LiDAR calibration and validation model on an NVIDIA GeForce RTX 2070 SUPER, an inference time of around 1 .67ms for a batch size of 1 is obtained. This means that the camera-to-LiDAR calibration and validation model performs calibration about 85 times per second, which is enough for realtime applications. The calibration can thus be performed in milliseconds while the vehicle is operating. A traditional techniques are much slower, with data acquisition taking about 20s during which the vehicle must not move, and the processing pipeline takes another 60s.

[0126] Table III shows the influence of different choices on the camera-to-LiDAR calibration and validation model, conducted with a publicly available dataset, using the split a presented in Table I. Results in Table III show that applying a soft data augmentation improves results in rotation and translation. Data augmentation enables the camera-to-LiDAR calibration and validation model to leam more features and generalize better.

[0127] In some embodiments, adding more information enables the camera-to-LiDAR calibration and validation model to leam new features. Some datasets, like Kitti, provide intensity information from the LiDAR sensor. As shown in Fig. 9, intensity can be helpful in perceiving two-dimensional visual features and patterns that are not perceived in the depth map. This is the case in Fig. 9 where road surface markings, which are visible in the intensity map, are not visible in the depth map since they are two-dimensional.

[0128] Using a ResNet backbone (ResNet is a popular CNN architecture for vision tasks that does not use attention mechanisms) results in about 25% higher Mean Average Error (MAE) and Standard Deviation (STD) on rotation and translation parameters compared to the single branch backbone according to the present techniques as shown in Table III. This shows that the selected single branch backbone outperforms traditional architectures. Relevant features are correctly found by attention mechanisms in our heatmap visualizations. However, results obtained with ResNet are still satisfactory and close to the state-of-the-art, showing that the unified data representation will work with different network backbones.

[0129] In addition to calibration, the camera-to-LiDAR calibration and validation model is used for validation. Given a corresponding pair of camera image and LiDAR point cloud, the task is to detect if the calibration is correct (within an acceptable tolerance range). This is a binary classification task. A positive classification would be the calibration as given by the dataset, and a negative classification indicates a de- calibration. The calibration is perturbed by up to 1° on each rotation axis and up to 10cm on each translation axis.

[0130] Initially, switch out the network heads for classification heads trained the network heads using a publicly available dataset, such as Kitti. The camera-to-LiDAR calibration and validation model may be referred to as UniVai. As shown in Table IV, the RegNet model severely overfits during training while our UniVai model fairs significantly better. In examples, transfer learning is used where the network is pre-trained on the calibration task and the network layers froze, then fine-tuned with a validation (classification) head. In this example, the backbone weights trained on calibration are expressive enough to be used in the validation task. Results show that the performance using this method improves from simply training UniVai without transfer learning.

[0131] FIG. 10 is a process flow diagram of a process 1000 for a camera-to-LiDAR calibration and validation model. At block 1002, scene samples are obtained from at least one scene, wherein the scene samples include unified data representation extracted from datasets (e.g., datasets, drivelog datasets). In examples a scene refers to data captured in the environment by sensors and devices. In some embodiments, the scene samples include at least one camera image and a LiDAR point cloud at a synchronized point in time. In some embodiments, the scene samples include multiple camera images and multiple LiDAR point clouds at a synchronized point in time. In some embodiments, the scene samples are generated by applying a random de-calibration on 6 DoF transformation parameters according to a uniform distribution on a chosen de-calibration range.

[0132] At block 1004, a single branch backbone network is trained jointly to output calibration parameters using a convolutional operation based on a stack of transformer layers. In examples, the transformers are a deep learning architecture based on a multi-head attention mechanism, with no recurrent units. In examples, transformers use less training time than previous recurrent neural architectures, such as long short-term memory (LSTM). In examples, the single branch backbone network includes spatial inductive biases and less sensitivity to data augmentation with input- adaptive weighting and global processing. For example, the single branch backbone encodes both local and global information in a tensor effectively and replaces local processing in convolutions with global processing using transformers.

[0133] At block 1006, a regression head network is trained to regress the calibration parameters. In examples, the regression head network comprises fully connected layers. In examples, the trained single branch backbone network and the trained regression head network form a camera- to-LiDAR calibration and validation model. In examples, the camera-to-LiDAR calibration and validation model is a trained deep learning network that obtains as input a unified data representation of sensor data and outputs calibration parameters for respective data sources (e.g., cameras, LiDARs, etc.) in real time. In examples, the single branch backbone network and regression head network are trained simultaneously. In examples, the single branch backbone network and regression head network are trained sequentially.

[0134] As described herein, self-attention-based vision can be leveraged to improve learning on driving scenes and proposed a single-branch architecture that outperforms the standard dualbranch architecture introduced by RegNet. Besides that, the present techniques are able to use transfer learning from the calibration task to outperform regular training on the calibration validation task without retraining the calibration network weights. In some embodiments, refinements such as the multi-frame iteration or the LSTM-based refinement are used for as temporal filtering. In some embodiments, large de-calibration is achieved using scale-iterative refinement.

[0135] In some embodiments, the present techniques are used with varying modalities. For example, the modality of input used in the unified data representation is modified. This enables calibration of other modalities such as Camera-to-Camera calibration or LiDAR- to-LiDAR calibration. In some embodiments, the camera intrinsic parameters are regressed (focal length, principal point, distortion coefficients) and intrinsic calibration performed as well.

[0136] According to some non-limiting embodiments or examples, provided is a method, including: obtaining, with at least one processor, scene samples from at least one scene, wherein the scene samples extracted from a dataset; training, with the at least one processor, a single branch backbone network to jointly output calibration parameters using a convolutional operation based on a stack of transformer layers; and training, with the at least one processor, a regression head network regress the calibration parameters, wherein the trained single branch backbone network and trained regression head form a camera-to-LiDAR calibration model that obtains as input a unified data representation and outputs calibration parameters in real time.

[0137] According to some non-limiting embodiments or examples, provided is a system, including: at least one processor, and at least one non-transitory storage media storing instructions that, when executed by the at least one processor, cause the at least one processor to: obtain scene samples from at least one scene, wherein the scene samples extracted from a dataset; train a single branch backbone network to jointly output calibration parameters using a convolutional operation based on a stack of transformer layers; and train a regression head network regress the calibration parameters, wherein the trained single branch backbone network and trained regression head form a camera-to-LiDAR calibration model that obtains as input a unified data representation and outputs calibration parameters in real time.

[0138] According to some non-limiting embodiments or examples, provided is at least one non- transitory storage media storing instructions that, when executed by at least one processor, cause the at least one processor to: obtain scene samples from at least one scene, wherein the scene samples extracted from a dataset; train a single branch backbone network to jointly output calibration parameters using a convolutional operation based on a stack of transformer layers; and train a regression head network regress the calibration parameters, wherein the trained single branch backbone network and trained regression head form a camera-to-LiDAR calibration model that obtains as input a unified data representation and outputs calibration parameters in real time. [0139] Further non-limiting aspects or embodiments are set forth in the following numbered clauses:

[0140] Clause 1 : A method, including: obtaining, with at least one processor, scene samples from at least one scene, wherein the scene samples extracted from a dataset; training, with the at least one processor, a single branch backbone network to jointly output calibration parameters using a convolutional operation based on a stack of transformer layers; and training, with the at least one processor, a regression head network regress the calibration parameters, wherein the trained single branch backbone network and trained regression head form a camera-to-LiDAR calibration model that obtains as input a unified data representation and outputs calibration parameters in real time. [0141] Clause 2: The method of any preceding clause, wherein the unified data representation includes data associated with sensors with varying sensor modalities.

[0142] Clause 3: The method of any preceding clause, wherein the unified data representation is an N-channel pseudo-image.

[0143] Clause 4: The method of any preceding clause, wherein the single branch backbone network and regression head network are trained using supervised machine learning.

[0144] Clause 5: The method of any preceding clause, wherein the regression head network includes two branches to separately regress translation and rotation components of the calibration parameters.

[0145] Clause 6: The method of any preceding clause, wherein transfer learning is applied to the trained single branch backbone network and trained regression head to form a camera-to-LiDAR validation model that obtains as input a unified data representation and outputs validation parameters in real time.

[0146] Clause 7: A system, including: at least one processor, and at least one non-transitory storage media storing instructions that, when executed by the at least one processor, cause the at least one processor to: obtain scene samples from at least one scene, wherein the scene samples extracted from a dataset; train a single branch backbone network to jointly output calibration parameters using a convolutional operation based on a stack of transformer layers; and train a regression head network regress the calibration parameters, wherein the trained single branch backbone network and trained regression head form a camera-to-LiDAR calibration model that obtains as input a unified data representation and outputs calibration parameters in real time.

[0147] Clause 8: The system of any preceding clause, wherein the unified data representation includes data associated with sensors with varying sensor modalities.

[0148] Clause 9: The system of any preceding clause, wherein the unified data representation is an N-channel pseudo-image.

[0149] Clause 10: The system of any preceding clause, wherein the single branch backbone network and regression head network are trained using supervised machine learning.

[0150] Clause 11 : The system of any preceding clause, wherein the regression head network includes two branches to separately regress translation and rotation components of the calibration parameters. [0151] Clause 12: The system of any preceding clause, wherein transfer learning is applied to the trained single branch backbone network and trained regression head to form a camera-to-LiDAR validation model that obtains as input a unified data representation and outputs validation parameters in real time.

[0152] Clause 13: At least one non-transitory storage media storing instructions that, when executed by at least one processor, cause the at least one processor to: obtain scene samples from at least one scene, wherein the scene samples extracted from a dataset; train a single branch backbone network to jointly output calibration parameters using a convolutional operation based on a stack of transformer layers; and train a regression head network regress the calibration parameters, wherein the trained single branch backbone network and trained regression head form a camera-to-LiDAR calibration model that obtains as input a unified data representation and outputs calibration parameters in real time.

[0153] Clause 14: The at least one non-transitory storage media of any preceding clause, wherein the unified data representation includes data associated with sensors with varying sensor modalities.

[0154] Clause 15: The at least one non-transitory storage media of any preceding clause, wherein the unified data representation is an N-channel pseudo-image.

[0155] Clause 16: The at least one non-transitory storage media of any preceding clause, wherein the single branch backbone network and regression head network are trained using supervised machine learning.

[0156] Clause 17: The at least one non-transitory storage media of any preceding clause, wherein the regression head network includes two branches to separately regress translation and rotation components of the calibration parameters.

[0157] Clause 18: The at least one non-transitory storage media of any preceding clause, wherein transfer learning is applied to the trained single branch backbone network and trained regression head to form a camera-to-LiDAR validation model that obtains as input a unified data representation and outputs validation parameters in real time.

[0158] In the foregoing description, aspects and embodiments of the present disclosure have been described with reference to numerous specific details that can vary from implementation to implementation. Accordingly, the description and drawings are to be regarded in an illustrative rather than a restrictive sense. The sole and exclusive indicator of the scope of the invention, and what is intended by the applicants to be the scope of the invention, is the literal and equivalent scope of the set of claims that issue from this application, in the specific form in which such claims issue, including any subsequent correction. Any definitions expressly set forth herein for terms contained in such claims shall govern the meaning of such terms as used in the claims. In addition, when we use the term “further comprising,” in the foregoing description or following claims, what follows this phrase can be an additional step or entity, or a sub-step/sub-entity of a previously- recited step or entity.

Claims

WHAT IS CLAIMED IS:

1. A method, comprising: obtaining, with at least one processor, scene samples from at least one scene, wherein the scene samples extracted from a dataset; training, with the at least one processor, a single branch backbone network to jointly output calibration parameters using a convolutional operation based on a stack of transformer layers; and training, with the at least one processor, a regression head network regress the calibration parameters, wherein the trained single branch backbone network and trained regression head form a camera-to-LiDAR calibration model that obtains as input a unified data representation and outputs calibration parameters in real time.

2. The method of claim 1, wherein the unified data representation comprises data associated with sensors with varying sensor modalities.

3. The method of claim 1, wherein the unified data representation is an N-channel pseudo-image.

4. The method of claim 1, wherein the single branch backbone network and regression head network are trained using supervised machine learning.

5. The method of claim 1, wherein the regression head network comprises two branches to separately regress translation and rotation components of the calibration parameters.

6. The method of claim 1, wherein transfer learning is applied to the trained single branch backbone network and trained regression head to form a camera-to-LiDAR validation model that obtains as input a unified data representation and outputs validation parameters in real time.

7. A system, comprising: at least one processor, and at least one non-transitory storage media storing instructions that, when executed by the at least one processor, cause the at least one processor to: obtain scene samples from at least one scene, wherein the scene samples extracted from a dataset; train a single branch backbone network to jointly output calibration parameters using a convolutional operation based on a stack of transformer layers; and train a regression head network regress the calibration parameters, wherein the trained single branch backbone network and trained regression head form a camera-to- LiDAR calibration model that obtains as input a unified data representation and outputs calibration parameters in real time.

8. The system of claim 7, wherein the unified data representation comprises data associated with sensors with varying sensor modalities.

9. The system of claim 7, wherein the unified data representation is an N-channel pseudo-image.

10. The system of claim 7, wherein the single branch backbone network and regression head network are trained using supervised machine learning.

11. The system of claim 7, wherein the regression head network comprises two branches to separately regress translation and rotation components of the calibration parameters.

12. The system of claim 7, wherein transfer learning is applied to the trained single branch backbone network and trained regression head to form a camera-to-LiDAR validation model that obtains as input a unified data representation and outputs validation parameters in real time.

13. At least one non-transitory storage media storing instructions that, when executed by at least one processor, cause the at least one processor to: obtain scene samples from at least one scene, wherein the scene samples extracted from a dataset; train a single branch backbone network to jointly output calibration parameters using a convolutional operation based on a stack of transformer layers; and train a regression head network regress the calibration parameters, wherein the trained single branch backbone network and trained regression head form a camera-to-LiDAR calibration model that obtains as input a unified data representation and outputs calibration parameters in real time.

14. The at least one non-transitory storage media of claim 13, wherein the unified data representation comprises data associated with sensors with varying sensor modalities.

15. The at least one non-transitory storage media of claim 13, wherein the unified data representation is an N-channel pseudo-image.

16. The at least one non-transitory storage media of claim 13, wherein the single branch backbone network and regression head network are trained using supervised machine learning.

17. The at least one non-transitory storage media of claim 13, wherein the regression head network comprises two branches to separately regress translation and rotation components of the calibration parameters.

18. The at least one non-transitory storage media of claim 13, wherein transfer learning is applied to the trained single branch backbone network and trained regression head to form a camera-to-LiDAR validation model that obtains as input a unified data representation and outputs validation parameters in real time.