Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
The embodiment of the invention provides a method and a device for controlling an unmanned vehicle and the unmanned vehicle. The unmanned vehicle Control device may be integrated into an Electronic device, and the Electronic device may be a device such as an automobile-specific controller (ECU) or a server.
The unmanned vehicle is an unmanned automatic control vehicle as the name implies, and can also be called as an intelligent vehicle for realizing autonomous traveling along a road under an unmanned state. For example, a vehicle-mounted sensing system senses the road environment, automatically plans a driving route and controls the vehicle to reach a preset place.
So-called unmanned driving allows a vehicle to safely and reliably travel on a road by sensing the surroundings of the vehicle using an in-vehicle sensor and controlling the steering and speed of the vehicle based on the road, vehicle position, and obstacle information obtained by the sensing.
For example, referring to fig. 1, taking an example that the unmanned vehicle control device is integrated in an electronic device, vehicle state information of the unmanned vehicle is obtained, where the vehicle state information includes real-time state information and target state information, a road condition environment is identified according to the real-time state information to obtain current road condition information, a control parameter in the target state information is adjusted based on the current road condition information to obtain an adjusted parameter, a target calibration parameter corresponding to the adjusted parameter is screened from preset calibration parameter information, the target calibration parameter is compensated according to error information between the real-time state information and the target state information to obtain a compensated parameter, and the unmanned vehicle is controlled based on the compensated parameter.
Among them, the unmanned vehicle control device can be applied to control of driving, braking, steering, and the like of a chassis electronic control system in the unmanned vehicle. For example, an accelerator pedal, a brake pedal, a steering wheel, and the like of the vehicle are controlled according to a post-compensation parameter output by the unmanned vehicle control device, which may be a pedal amount or a steering amount. The following are detailed below. It should be noted that the following description of the embodiments is not intended to limit the preferred order of the embodiments.
The present embodiment will be described from the perspective of an unmanned vehicle control apparatus, which may be specifically integrated in an electronic device, which may be a server or an ECU.
A method of controlling an unmanned vehicle, comprising: the method comprises the steps of obtaining vehicle state information of the unmanned vehicle, wherein the vehicle state information comprises real-time state information and target state information, identifying road condition environment according to the real-time state information to obtain current road condition information, adjusting control parameters in the target state information based on the current road condition information to obtain adjusted parameters, screening target calibration parameters corresponding to the adjusted parameters from preset calibration parameter information, compensating the target calibration parameters according to error information between the real-time state information and the target state information to obtain compensated parameters, and controlling the unmanned vehicle based on the compensated parameters.
As shown in fig. 2, the specific flow of the unmanned vehicle control method is as follows:
101. vehicle state information of the unmanned vehicle is acquired, and the vehicle state information comprises real-time state information and target state information.
The unmanned vehicle is an unmanned automatic control vehicle as the name implies, and can also be called as an intelligent vehicle for realizing autonomous traveling along a road under an unmanned state. For example, a vehicle-mounted sensing system senses the road environment, automatically plans a driving route and controls the vehicle to reach a preset place.
The vehicle state information may include state information describing the vehicle in operation, such as speed, acceleration, torque, vehicle body internal information, and/or vehicle body external information. The vehicle state information may be classified into two categories, for example, real-time state information and target state information.
The real-time status information may include real-time speed, acceleration, position of the vehicle, real-time engine information, and/or external environment information of the vehicle. The target state information may be information such as target speed, acceleration, position, and/or engine information set according to the path plan.
For example, the real-time status information may be obtained by an external sensor, such as a speed and acceleration sensor outside the vehicle, an environmental sensor inside and outside the vehicle, and an engine monitoring sensor, to obtain information such as a real-time speed, torque, and/or power of an engine of the vehicle in the current status. For example, when the vehicle performs path planning, information such as target speed, acceleration, position and/or engine information is set in advance, and the target state information is stored in a local database of the vehicle control system, and can be directly called in the local database of the vehicle control system when the target state information is obtained. The target state information may also be directly input to the unmanned vehicle control device by the user, for example, the user converts each target state information in the path plan set in advance into information that can be recognized by the unmanned vehicle control device and then directly uploads the information to the unmanned vehicle control device, or the target state information generated in the path plan test software may also be directly uploaded.
It is emphasized that the target state information may be adjusted when obtaining the vehicle target state information, for example, the format of the target state information may be adjusted so as to better identify the information therein, and the target state information may be converted, for example, km/h may be converted into m/s, etc. The acquired vehicle state information can be screened, and vehicle state information related to vehicle control is selected.
102. And identifying the road condition environment according to the real-time state information to obtain the current road condition information.
The road condition environment may be a road condition type and a road surface environment where the vehicle runs, for example, a road surface type, such as a concrete road surface, an asphalt road surface and/or a dirt road surface, a dry road surface or a wet road surface, a wet degree of a wet road surface, and the like.
The current road condition information may be road condition information of a road surface on which the vehicle is currently driving, and may include, for example, a type of the road surface, such as a paved or unpaved road surface, an asphalt road surface, a concrete road surface, or a dirt road surface, and may further include a road surface environment, such as a dry or wet road surface, and may further obtain a road surface water film thickness for a wet road surface to classify the wet road surface, for example, in three stages, a first-stage wet road surface water film thickness is the largest, a second-stage wet road surface water film thickness is intermediate, and a third-stage wet road surface water film thickness is the smallest.
(1) And classifying the real-time state information according to the type of the external environment of the vehicle.
For example, the real-time status information is classified according to the type of the external environment of the vehicle, and may be classified into road real-time status information, noise real-time status information, wind resistance real-time status information, temperature real-time status information, and the like.
(2) And extracting road condition state information related to the road condition environment from the classified real-time state information.
The road condition status information may include information that the road condition affects the vehicle when the vehicle is running under the current road condition, such as speed, acceleration, vehicle vibration information, and information related to vehicle tires.
For example, the traffic status information related to the traffic environment is extracted from the classified real-time status information, for example, the traffic status information such as speed, acceleration, vehicle vibration information, and vehicle tire related information is extracted from the classified real-time status information.
(3) And identifying the road condition state information to obtain the current road condition information.
And S1, screening out the stress information and the real-time speed of the tire from the road condition information.
The tire stress information may be information related to tire stress, such as speed, acceleration, tire pressure, tire size, and the like of the vehicle, which are used for calculating the tire stress of the vehicle.
For example, information related to tire stress and real-time speed are screened from the road condition information, for example, information related to tire stress, such as acceleration, tire pressure and tire size of the vehicle, is screened from vehicle tire related information of the road condition information. Meanwhile, the real-time speed of the unmanned vehicle under the current road condition is screened from the road condition state information, for example, 0.01 or 0.02 second can be set as a detection period, and the real-time speed of the unmanned vehicle in the detection period is obtained. It should be noted that the detection period may be set according to actual application requirements, and may be a positive value with an arbitrary value different from zero.
It should be emphasized here that the value of the detection period may be any value, and needs to be set according to practical applications, and the smaller the period, the more accurate the corresponding control effect.
And S2, calculating the stress information of the tires to obtain the ground friction under the current road condition.
For example, the tire force information is calculated to obtain the friction force of the ground under the current road surface, for example, the tire force information may be calculated by using an ackerman bicycle model and a friction circle model in vehicle dynamics, where the ackerman bicycle model is shown in fig. 3. The constants that characterize the attributes of a vehicle in the ackermann bicycle model are: l
f,l
rRespectively, front and rear suspension length, C
f,C
rThe roll stiffness coefficients of the front wheel and the rear wheel are respectively, M is the mass of the whole vehicle, and M is
dFor equivalent mass of rotating parts, I
zThe moment of inertia of the whole vehicle. These constants are available at the time of vehicle shipment. The information selected from the vehicle information about tire stress may be v
y,v
xRespectively, transverse and longitudinal speed, a
y,a
xRespectively, the transverse acceleration and the longitudinal acceleration are the steering wheel rotation angles,
the angular velocity and the angular acceleration of the whole vehicle,
is the engine driving force. According to the stress information of the screened tyreThe front and rear suspension speed direction angle can be calculated through a formula, and the specific formula is as follows:
wherein theta isfIs the front suspension speed direction angle, thetarIs the rear suspension speed direction angle.
The front and rear wheel lateral friction force can be calculated through a formula according to the screened stress information with the tire, and the specific formula is as follows:
Ff=2Cf(-θf),Fr=2Crθr
wherein, FfIs front wheel side friction force, FrA rear wheel side frictional force, Cf,Cr-front and rear wheel roll stiffness factor, -steering wheel angle.
And then, calculating longitudinal friction force of the front wheel and the rear wheel according to the screened stress information of the tires, wherein the specific formula is as follows:
wherein f isfLongitudinal friction of the front wheel, frLongitudinal friction of the rear wheel, ay,axRespectively, the transverse acceleration and the longitudinal acceleration are steering wheel corners, M is the mass of the whole vehicle, and M is the mass of the whole vehicledFor rotating parts of equivalent mass, FfIs front wheel side friction force, FrIs the rear wheel side friction force.
According to the longitudinal and lateral friction forces of the front and rear wheels, the friction force of the front and rear wheels of the unmanned vehicle can be calculated by adopting a friction circle stress model, and the specific formula is as follows:
wherein f is1For friction of front wheels of unmanned vehicles, f2Friction force of rear wheel of unmanned vehicle, ffLongitudinal friction of the front wheel, frLongitudinal friction of rear wheels, FfIs front wheel side friction force, FrIs the rear wheel side friction force.
It should be emphasized that, when calculating the total friction force borne by the front wheel and the rear wheel of the unmanned vehicle, the force law of the vehicle tire needs to be considered, and the force law of the vehicle tire needs to be analyzed based on the friction circle model. The friction circle model is shown in fig. 4, and the friction circle model is established based on the fact that: the limit of tire friction is the load multiplied by the road surface coefficient of friction, regardless of the direction of the friction force. The frictional force may be distributed to the tire in a lateral direction, a longitudinal direction, or a combination thereof, depending on the tire slip angle. The lateral and longitudinal forces of the tire are drawn on the same plane, so that a friction circle model as shown in fig. 4 is obtained, and the radius of the friction circle can be changed along with different road conditions.
And S3, determining the sliding rate under the current road condition according to the real-time speed.
The slip rate can be the degree of slip generated by friction between the tire and the ground, and is used for describing the friction factor of the current road surface, so that the current road condition information is determined according to the friction factor.
For example, the real-time speed under the current road condition is screened out from the vehicle information, and the slip rate under the current road condition is obtained according to the real-time speed of the unmanned vehicle. For example, the real-time speed and the wheel rotation speed of the vehicle are obtained from the vehicle information, and the slip ratio under the current road condition can be calculated through the relationship between the real-time speed and the wheel rotation speed of the unmanned vehicle, wherein the specific calculation formula is as follows:
wherein, ω is wheel speed (in rad), and r is the tread radius, takes tire mark as the standard, and v is the speed of a motor vehicle, needs to process the real-time speed of gathering and obtains, for example can obtain through locating data differential filtering.
And S4, generating the current road condition information according to the ground friction force and the sliding rate.
For example, according to the ground friction, first road condition information corresponding to the ground friction is screened out from preset road condition information. For example, the first road condition information is screened out in the road condition classifier according to the ground friction and the real-time speed under the current road condition, for example, the calculated ground friction of the front wheel and the rear wheel under the current road condition is f1And f2Real-time speed of vehicle is v1Screening out the value v in the road condition classifier1At speed, ground friction is f1And f2And corresponding road surface type, wherein if the road surface type is a wet road surface, the wet grade, namely the water film thickness, of the wet road surface can be further obtained according to the ground friction force, and the ground friction force is f1And f2And acquiring the moisture grade of the wet road surface according to the corresponding road surface water film thickness, taking three levels as an example, assuming that the water film thickness is larger and exceeds the threshold value of the second level water film thickness, determining that the moisture grade of the wet road surface is the third grade, and determining that the first road condition information can be the third grade wet road surface.
For example, according to the slip rate, the second road condition information corresponding to the slip rate is screened out from the preset road condition information. For example, the second road condition information is screened in the road condition classifier according to the slip rate and the real-time speed under the current road condition, for example, the calculated slip rate under the current road condition is γ, and the real-time speed of the vehicle is v1Screening out the value v in the road condition classifier1At speed, the slip rate is the road type corresponding to γ, and assuming that the road type is a wet concrete road, the second road condition information is the wet concrete road.
For example, the first road condition information and the second road condition information are fused to obtain the real-time road condition information. For example, when the first road condition information may be a third-level wet road surface, the second road condition information is a wet concrete road surface, the first road condition information and the second road condition information are fused, and the obtained road condition information is the third-level wet concrete road surface, the real-time road condition information may be the third-level wet concrete road surface.
The road condition classifier can be a segmented mapping model, the mapping model is a mapping model of sliding rate and friction force at a plurality of real-time speeds, and the road condition classifier can also be a support vector machine and a neural network model for road condition identification. Taking the segmented mapping model as an example, building the segmented mapping model requires collecting parameters related to tire stress of a target vehicle at a plurality of target speeds under a plurality of known road conditions in advance, calculating slip rates and friction forces corresponding to the plurality of target speeds under different road conditions, and fitting a segmented quadratic boundary curve to the slip rates and friction forces corresponding to the plurality of target speeds under different road conditions, wherein the boundary curve can be a segmented curve of the slip rates at different speeds with respect to the friction force, so that the segmented mapping model under different road conditions can be obtained. And inputting the obtained slip rate and the friction force into a segmented mapping model to respectively obtain first road condition information and second road condition information.
It should be emphasized that the present invention directly adopts a nonlinear vehicle dynamics model to acquire the current real-time traffic information, so that the present invention is not only suitable for the working conditions with higher linearity, but also suitable for the working conditions with higher nonlinearity, such as high speed, large turning angle, etc.
103. And adjusting the control parameters in the target state information based on the current road condition information to obtain adjusted parameters.
The control parameter can be a parameter for automatically controlling the vehicle to run, and a control instruction can be generated based on the control parameter so as to control the unmanned vehicle to run automatically. For example, the control parameters may be preset control parameters such as a target speed, a maximum acceleration, a minimum radius limit, a first-order delay time of vehicle response, and a control model parameter of a feedback branch from the point a to the point B of the unmanned vehicle according to the path planning information.
(1) And screening out the external environment information of the vehicle from the real-time state information.
The vehicle external environment information may include environment information such as a noise environment, an air environment, and/or a temperature environment outside the vehicle, for example, information such as a noise decibel outside the vehicle, a wind speed, a wind resistance, and an air humidity outside the vehicle, and/or a vehicle external temperature.
For example, environment information outside the vehicle, such as noise environment, air environment, and/or temperature environment outside the vehicle, is screened out from the classified real-time status information.
(2) And fusing the external environment information of the vehicle and the current road condition information to obtain interference information.
The interference information may be information interfering with a driving state of the vehicle, such as noise information, current road condition information, air information outside the vehicle, temperature information outside the vehicle, and the like. For example, taking the road surface or road condition information as an example, when the vehicle is driven from a paved road surface to a muddy dirt road, different working conditions may be generated, and if the target speed or acceleration needs to be maintained, the requirement for the traction force of the vehicle may be greatly changed, thereby interfering with the normal driving of the vehicle.
For example, the current road condition information and the environment information outside the vehicle are fused to obtain the interference information outside the vehicle, and the specific fusion mode may be various, for example, assuming that the current road condition information is a third-level wet concrete pavement, the initial interference information is environment information such as wind speed information, noise information, tire pressure, current pavement friction factor, and temperature information inside and outside the vehicle, and the current road condition information and the initial interference information are directly combined to obtain the interference information outside the vehicle. The current road condition information can also be adjusted according to the initial interference information, for example, according to the internal and external temperature information of the vehicle, the road surface temperature and other information are added in the current road condition information.
(3) And adjusting the control parameters in the target state information according to the interference information to obtain adjusted parameters.
And A1, classifying the interference information according to the interference type.
The interference type may be a factor that interferes with the driving of the vehicle, such as noise interference, road condition interference, air interference, temperature interference, and other interference types.
For example, the interference information may be classified according to interference types, for example, the interference information may be classified into direct interference or indirect interference, the direct interference may be further classified into road interference, temperature interference, wind speed interference in air interference, the indirect interference may be further classified into noise interference, humidity interference in air interference, and the like.
And A2, adjusting the interference information of different types to obtain first interference values of different types.
For example, the interference information of different types is adjusted to obtain first interference values of different types, and the adjustment modes may be multiple, for example, the interference information of different types may be converted in a unified manner, for example, decibels of noise are converted into interference values, interference information of speed, road condition type, and the like is converted into interference values of a unified unit, and the conversion process may specifically be conversion according to preset conversion information, for example, an interference value of 10 corresponding to 50 decibels in the preset noise conversion information, an interference value of a concrete road surface in the preset road condition conversion information is 5, an interference value corresponding to an asphalt road surface is 8, and only the corresponding interference value needs to be directly screened from the preset road surface conversion information according to the road surface information, and the corresponding interference values are obtained by the same method for other interference information such as temperature, air, and the like.
It is emphasized that the first interference value may include a plurality of interference information, and each type of interference information corresponds to one first interference value.
And A3, weighting the first interference values of different types according to preset weight coefficients to obtain a second interference value of the control parameter.
The preset weight coefficient may include interference degrees of different types of interference information on the target control coefficient.
For example, the different types of interference values are weighted according to preset weight coefficients to obtain a second interference value of the target control parameter, for example, taking the first interference values of the different types including a first interference value of a road condition a1, a first interference value of noise a2, a first interference value of air A3 and a first interference value of temperature a4 as an example, according to different types of interference information, the interference degrees of the target control parameter are different, the preset weight coefficients corresponding to the four types of first interference values are x1, x2, x3 and x4, the sum of the four types of weight coefficients is 1, the four types of first interference values are multiplied by the weight coefficients, and the products of the four types of first interference values and the weight coefficients are accumulated to obtain the second interference value of the target control parameter.
And A4, adjusting the control parameters in the target state information according to the second interference value to obtain adjusted parameters.
For example, the control parameter in the target state information is adjusted according to the second interference value to obtain the adjusted parameter, for example, the second interference value may be directly calculated with a corresponding value in the control parameter, and the calculation method may include simple four arithmetic operations or may include complex arithmetic operations. For example, the control parameter is a target value of the acceleration, the acceleration value may be directly subtracted from or added to the second disturbance value, the second disturbance value may be input into a predetermined calculation model to obtain a compensation value affecting the target value of the acceleration, and the target value of the acceleration may be directly added to the compensation value to obtain the adjusted parameter.
It is emphasized here that interference can be divided into structural interference and non-structural interference, the latter often appearing in the form of white noise or pink noise. The structural interference mainly comprises wind resistance, gradient and other quantities which can be directly measured or indirectly calculated, and the parts can be directly measured and compensated at the control input end; the non-structural interference mainly comprises various measurement noises and the like, and the part can be removed by using classical filters such as low-pass filtering and the like or modern filters such as Kalman filtering and the like without additional compensation; however, non-structural interference imposes a requirement on the robustness of the control system, which may need to be based on H when the noise is too loud∞The poles of the local linearization control system are theoretically reconfigured to sacrifice rapidity in exchange for stability, so that the adjustment of the control parameters can be partial adjustment of the control parameters or adjustment of all the control parametersFor example, for structural interference, all control parameters influencing structural interference items can be adjusted, and for non-structural interference, filtering is adopted to filter interference, and the control parameters are not required to be adjusted.
104. And screening target calibration parameters corresponding to the adjusted parameters from the preset calibration parameter information.
The preset calibration parameter information may include parameter information for calibrating the adjusted parameter, and usually exists in the form of a calibration table.
Calibration is to be understood as calibration of a control parameter, and the main process is to obtain a calibration value of a target control parameter in a calibration table according to the target control parameter, wherein the calibration table is a reference table containing a calculation model obtained by fitting data of a test data set to acquire a plurality of test data sets.
(1) And screening a calibration table corresponding to the current road condition information from the preset calibration parameter information, wherein the calibration table comprises a plurality of state areas and calculation models corresponding to the state areas.
For example, the calibration table corresponding to the current road condition information is screened out from the preset calibration parameter information, for example, the preset calibration parameter information includes calibration tables corresponding to various road condition types, and taking the current road condition information as a third-stage wet concrete road surface, the first target control parameter as acceleration, the initial second target parameter as pedal amount as an example, the calibration table of the third-stage wet concrete road surface is screened out from the preset calibration parameter information, as shown in fig. 5, a is acceleration, v is speed, u is pedal amount, a is a model parameter, and the parameter is a constant. The calibration table is divided into a plurality of state areas, namely a brake section, a dead zone and an accelerator section, by adopting an analytical model instead of a numerical value form, wherein 0-order continuity is kept among the areas, and each state area corresponds to one calculation model.
It should be emphasized that each road condition information corresponds to a calibration table, and a continuous generalized control quantity-pedal quantity formed by combining a brake and an accelerator in acceleration can be obtained in the calibration table. When the pedal amount is less than 0, the brake is represented, and when the pedal amount is more than 0, the accelerator is represented. By the method, the state runaway or sudden change caused by the discontinuity of the second target control parameter at the accelerator/brake switching position can be eliminated, and the stability and the adaptability of the control of the unmanned vehicle in the constant-speed or quasi-constant-speed cruising state are ensured.
(2) And determining a target state area corresponding to the adjusted parameters in a calibration table.
For example, the current state of the vehicle is determined from the vehicle state information of the unmanned vehicle, and the target state region corresponding to the adjusted parameter may be determined in the calibration table according to the state of the vehicle. For example, taking the adjusted parameter as the first acceleration as an example, the vehicle real-time state information is used to determine that the unmanned vehicle is at a standstill, an idle speed or a high speed at this time, if the speed of the vehicle is 0 and the acceleration is also 0, then the unmanned vehicle is at a standstill at this time, the state area corresponding to the first acceleration is a brake segment, if the speed of the unmanned vehicle is small at this time but the acceleration is large, then the unmanned vehicle is at an idle state (a starting stage) at this time, the state area corresponding to the first acceleration is a dead segment, and if the speed of the unmanned vehicle is large at this time and the acceleration is also large at this time, then the unmanned vehicle is in an accelerator increasing state at this time, and then the state area corresponding to the first acceleration is an accelerator segment.
(3) And calibrating the adjusted parameters by adopting a calculation model of the target state area to obtain target calibration parameters.
For example, the adjusted parameter is calibrated by using the calculation model of the target state area to obtain the initial target calibration parameter, for example, taking the adjusted parameter as the first acceleration, the target calibration parameter as the initial pedal amount, and the target state area corresponding to the first acceleration as the throttle section as an example, the calculation model for calibrating the first acceleration can be obtained as follows from fig. 5:
α(v,u)=VTAu,V=(1,v,v2)T,u=(1,u,u2)T,A∈R3×3(segmentation)
Where α is a first acceleration, v is a real-time velocity of the unmanned vehicle, u is an initial pedal amount, and A is a model parameter, which is a segmented 3 × 3 parameter matrix.
The first acceleration is calculated by referring to the calculation model, and an initial pedal amount for calibrating the first acceleration in the accelerator section can be obtained.
It should be noted that different road surfaces lead to very different vehicle dynamics characteristics, the model collects data such as speed, acceleration, pedal amount and the like of the vehicle in the forward and backward driving processes under different road conditions (such as rain, snow, fog and the like) in advance and fits different off-line mapping models, the current road condition is judged in real time according to the current state and performance of the vehicle in the automatic running process of the vehicle, and then different longitudinal controllers are switched and different parameter calibration tables are loaded according to different road conditions to achieve a satisfactory control effect.
The preset calibration parameter information may be generated by other devices and provided to the unmanned vehicle control device, or may be generated by the unmanned vehicle control device, that is, before the step of "screening out the calibration table corresponding to the current traffic information in the preset calibration parameter information", the unmanned vehicle control method may further include:
acquiring vehicle test data under various road conditions;
identifying target anchor point data of vehicle characteristics under different road conditions according to the test data;
fitting boundary curves of the state areas of the calibration table corresponding to different road conditions according to the target anchor point data;
and fitting continuous curved surfaces of each state area of the calibration table corresponding to different road conditions according to the boundary curve to obtain a calculation model of each state area.
For example, the preset calibration parameter information is generated by mainly performing piecewise quadratic fitting on each state area by using the key anchor point information. In actual drive test, the nonlinear degree of the mapping of the accelerator section is very high, and the other two sections are basically linear except for a low-speed section. This is because the throttle section contains complex engine characteristics, while the brake section only requires the hydraulic mechanism to apply a certain deceleration force. In addition, the acceleration characteristic of the accelerator stage has a large relationship with the current speed of the vehicle, and the secondary acceleration characteristic of the high-speed stage under the same horsepower is much inferior to the starting acceleration characteristic of the low-speed stage. Based on this, different fitting methods can be adopted for the throttle section and the other two sections, specifically as follows:
(1) an accelerator section: and carrying out segmented quadratic fitting on the speed and the pedal quantity.
(2) Dead zone, braking section: and carrying out high-low piecewise quadratic fitting on the speed, and only carrying out linear fitting on the pedal amount.
For example, taking the computational model fitting of the throttle segment as an example, a point-line-plane fitting idea is mainly adopted. As shown in fig. 6, the throttle section also includes a plurality of regions according to the difference of speed or acceleration, specifically as follows:
and C1, obtaining vehicle test data under various road conditions.
For example, the sensor or the real-time vehicle information collecting device collects test data under various road conditions, such as speed, acceleration, pedal amount, and path related information in one or more test periods under various road conditions.
And C2, identifying target anchor point data of the vehicle characteristics under different road conditions according to the vehicle test data.
The target anchor point data may include key data points describing characteristics of the vehicle, for example, a position point of the bottom of the wheel in the planned path from a to B, for example, a center point of a rear axle, and the anchor point data may include a speed, an acceleration, and a pedal amount of the position point.
For example, key anchor points describing characteristics of the vehicle are identified in the test data, for example, one or more fixed points in the planned path may be determined according to the characteristics of the vehicle, and key data of the one or more fixed points may be obtained and stored in a lattice form with a size of m × n. The speed and the pedal amount of the anchor point are uniformly distributed in the rectangular state space, and the acceleration is set as the average value of the internal sample points of the tee ball near the corresponding working condition point in the state space, specifically as follows:
amn=Average{a|a(v,u),(v,u)∈B[(vm,un),]}
wherein, amnThe acceleration of the working point is a first acceleration, v is a real-time speed, u is a pedal amount under the first acceleration, and B is a dot matrix set of the acceleration and the pedal amount.
And C3, fitting the boundary curve of the accelerator section of the calibration table corresponding to different road conditions according to the target anchor point data.
For example, according to the target anchor point data, the boundary regions of the status regions of the calibration table corresponding to different road conditions are fitted. For example, according to different road conditions, each road condition needs to be fitted into a calibration table separately, and the region division is performed according to a specific speed, for example, for a specific speed (longitudinal direction), the pedal amount curve is divided into (n-1) sections; for a certain pedal amount (lateral direction), the velocity curve is divided into (m-1) segments. The principle of fitting is to keep the 1 st order continuous while making the curve as smooth as possible, i.e. the modulus of the quadratic coefficient is as small as possible. Taking the jth longitudinal pedal magnitude curve as an example, the fitting process is to solve the following quadratic convex optimization problem with linear constraint, specifically as follows:
2aiui+βi=2ai+1ui+βi+1,i=1,...,n-2
wherein alpha, beta and gamma are respectively a quadratic term, a primary term and a constant term coefficient, and u and a are respectively anchor values of pedal amount and acceleration. It can be seen that the scale of the parameter to be fitted is 3(n-1), and the number of constraints is 2(n-1) + (n-2), and there is an optimization space of 1 degree of freedom. The problem can be converted into a least square problem by using the existing Lagrange multiplier method in convex optimization and solved.
And C4, fitting the continuous curved surface of the throttle section in the calibration table corresponding to different road conditions according to the boundary curve to obtain a calculation model of the throttle section.
For example, according to the boundary curves of different areas, continuous curved surfaces of throttle sections in the calibration table corresponding to different road conditions are fitted to obtain a calculation model of the throttle sections. For example, a continuous curved surface of the slice quadratic is fitted according to the boundary curve, and the slice granularity is (m-1) × (n-1). The principle of fitting is to keep 0 th order continuous and make the curved surface as smooth as possible, i.e. the modulus of the quadratic coefficient is as small as possible. Taking the curved surface (p, q) as an example, the fitting process is to solve the following quadratic convex optimization problem with linear constraint, specifically as follows:
A∈A3×3
wherein alpha, beta and gamma are respectively a quadratic term, a primary term and a constant term coefficient of a pedal amount curve, v, mu and lambda are respectively a quadratic term, a primary term and a constant term coefficient of a speed curve, u and v are respectively anchor points of pedal amount and speed, and A is a parameter matrix to be fitted.
From the above calculation process, the scale of the parameter to be fitted is 9, the number of constraints is 8 (the 12 constraints are linearly related, the rank number is 8, and there are 4 redundant constraints from the curve overlap at the anchor point), and there is an optimization space with 1 degree of freedom. The problem can be converted into a least square problem by using the existing Lagrange multiplier method in convex optimization and solved.
In the piecewise linear fitting algorithm, the piecewise granularity and the mxn scale of the target anchor point data lattice determine the effect of the calculation model to a large extent, and can be adjusted appropriately according to different vehicle characteristics. Fig. 7 depicts the effect of different segmentation granularities on the output model, with the vehicle type selected as lincoln MKZ. Wherein: (I) sparse (2 × 1) particle size, (II) reasonable (4 × 5) particle size, (III) dense (8 × 9) particle size. As can be seen from fig. 7, the too sparse granularity cannot accurately fit the trend of the drive test data, and the too dense granularity is easily affected by interference items such as noise of the drive test data, resulting in overfitting.
105. And compensating the target calibration parameters according to the error information between the real-time state information and the target state information to obtain compensated parameters.
(1) The current value of the control parameter is extracted from the real-time status information, and the target value of the control parameter is extracted from the target status information.
For example, the current value of the control parameter is extracted from the real-time status information, and the target value of the control parameter is extracted from the target status information. For example, taking the control parameter as the acceleration of the unmanned vehicle as an example, the current value of the vehicle acceleration under the current road condition of the vehicle is extracted from the real-time state information, and the target value of the vehicle acceleration is extracted from the target state information.
(2) The current value is compared to a target value to obtain error information.
For example, the current value is compared with the target value to obtain error information, for example, the current value of the acceleration of the unmanned vehicle in the actual driving process in the planned path area is compared with the target value to obtain a difference value of the acceleration of the unmanned vehicle in the planned path area, taking the control parameter as the acceleration as an example, and the difference value is taken as the error information.
(3) And compensating the target calibration parameters according to the error information to obtain compensated parameters.
For example, a compensation value corresponding to the error information may be filtered from the preset compensation information to compensate the target calibration parameter, for example, a proportional-integral-derivative control model (ProportionInt) may be usedProportional differential, PID) to screen a compensation value corresponding to the error information from preset compensation information, and compensating a target calibration parameter using the compensation value, the PID may be placed in a floor controller of the unmanned vehicle control device, the floor controller having a structure as shown in fig. 8, where a
tar,a
cmd,a
cur,a
errTarget value, command value, current value and error value, u, respectively, representing acceleration
ff,u
fb,u
cmdA feedforward value representing the amount of pedal, respectively, a feedback value and a command value omega representing disturbance terms,
represents observed/calculated/estimated values of the interference term. Specifically, when the error value of the acceleration is obtained, the error value of the acceleration is subjected to 3 operation modes of proportion, integration and differentiation to obtain the compensation value of the pedal amount under the current road condition and the current speed.
Wherein, the PID controller comprises a proportional unit (P), an integral unit (I) and a differential unit (D). The specific calculation formula of the corresponding compensation value obtained according to the error information is as follows:
u(t)=kp(e(t)+1/TI∫e(t)dt+TD×de(t)/dt)
wherein e (t) is the input error information, u (t) is the output compensation value, kpIs a proportionality coefficient, TI is an integral time constant, TD is a differential time constant, and k ispTI, and TD may be set according to practical applications.
It should be noted here that the structure of the bottom layer controller as described in fig. 8 may include a feed-forward part
And a feedback section (PID). Wherein the feed forward part
Ensuring the dynamic characteristics of the vehicle, such as rapid acceleration and rapid braking, the feedback Part (PID) corrects the error of the feedforward control to ensure the stable stateThe tracking characteristics of (1). In this unmanned vehicle control apparatus, the main vehicle characteristics are approximated as three parts in cascade, in the order:
(1) nonlinear mapping from pedal amount to steady acceleration, which belongs to a nonlinear link;
(2) the acceleration lag characteristic of the engine is approximated as a first-order lag element
The first-order lag belongs to a linear minimum phase link, and the time constant T can be obtained by the existing system identification method based on least square based on a step response curve of drive test;
(3) pure delay e of CAN and actuator-τsIt belongs to a linear non-minimum phase element.
Wherein, the inverse system of the nonlinear mapping in (1) is realized in the form of a calibration table of piecewise nonlinearity. Because the order of the denominator of the control system is generally not lower than the numerator, and the inverse system of the first-order lag in (2) is designed into an advance correction link, the details are as follows:
wherein Ts and T0s are time constants, respectively.
Wherein, the links such as pure delay brought by the actuating mechanism in (3) can not design a causal and stable inverse system, so the feedforward branch circuit does not consider the links. The final feedforward branch is formed by cascading two parts, namely a lead correction link and a nonlinear calibration model, namely a mapping table from acceleration to pedal amount.
Due to the influence of high-frequency noise, the feedback branch can also design the PID controller as a PI controller or a low-pass filter.
106. And controlling the unmanned vehicle based on the compensated parameters.
For example, based on the compensated parameters, corresponding control commands may be generated, according to which the unmanned vehicle is controlled. For example, taking the compensated parameter as the pedal amount, a control instruction corresponding to the pedal amount is generated according to the pedal amount, and the control instruction is input to a control system of an engine of the unmanned vehicle, so that the unmanned vehicle is controlled by the pedal amount corresponding to the action of a corresponding accelerator pedal or brake pedal, and the target value of the acceleration of the unmanned vehicle under the current road condition can be realized.
As can be seen from the above, in the embodiment of the present invention, after the vehicle state information of the unmanned vehicle is obtained, the vehicle state information includes real-time state information and target state information, the road condition environment is identified according to the real-time state information to obtain current road condition information, then, the control parameter in the target state information is adjusted based on the current road condition information to obtain an adjusted parameter, a target calibration parameter corresponding to the adjusted parameter is screened out from preset calibration parameter information, then, the target calibration parameter is compensated according to error information between the real-time state information and the target state information to obtain a compensated parameter, and the unmanned vehicle is controlled based on the compensated parameter; according to the scheme, real-time road condition recognition is carried out on the basis of vehicle dynamic characteristics by collecting tire stress information of the vehicle, and multiple factors such as control parameters are adjusted and compensated according to the external environment and response information of the vehicle, so that the accuracy of unmanned vehicle control is greatly improved.
The method described in the above examples is further illustrated in detail below by way of example.
In the present embodiment, the unmanned vehicle control apparatus is specifically integrated into an electronic device, and the electronic device may be an ECU or a server, taking the control parameter as the target acceleration and the post-compensation parameter as the pedal amount as an example.
As shown in fig. 9, a method for controlling an unmanned vehicle includes the following specific steps:
201. the electronic device obtains vehicle state information of the unmanned vehicle, wherein the vehicle state information comprises real-time state information and target state information.
For example, the electronic device may obtain real-time speed and acceleration of the vehicle in the current state through a speed and acceleration sensor outside the vehicle, obtain environmental information inside and outside the vehicle through an environmental sensor of the vehicle, and obtain information such as real-time rotation speed, torque and/or power of an engine of the vehicle in the current state through an engine monitoring sensor. For example, when the vehicle performs path planning, information such as target speed, acceleration, position and/or engine information is set in advance, and the target state information is stored in a local database of the vehicle control system, and can be directly called in the local database of the vehicle control system when the target state information is obtained. The target state information may also be directly input to the unmanned vehicle control device by the user, for example, the user converts each target state information in the path plan set in advance into information that can be recognized by the unmanned vehicle control device and then directly uploads the information to the unmanned vehicle control device, or the target state information generated in the path plan test software may also be directly uploaded.
It is emphasized that the target state information may be adjusted when obtaining the vehicle target state information, for example, the format of the target state information may be adjusted so as to better identify the information therein, and the target state information may be converted, for example, km/h may be converted into m/s, etc. The acquired vehicle state information can be screened, and vehicle state information related to vehicle control is selected.
202. The electronic device classifies the real-time status information according to the type of the environment external to the vehicle.
For example, the electronic device classifies the real-time status information according to the type of the external environment of the vehicle, for example, the real-time status information may be classified into road real-time status information, noise real-time status information, wind resistance real-time status information, temperature real-time status information, and the like.
203. And the electronic equipment extracts the road condition state information related to the road condition environment from the classified real-time state information.
For example, the electronic device extracts the road condition status information related to the road condition environment from the classified real-time status information, for example, the road condition status information such as speed, acceleration, vehicle vibration information, and vehicle tire related information is extracted from the classified road surface real-time status information.
204. And the electronic equipment identifies the road condition state information to obtain the current road condition information.
(1) And screening out the stress information and the real-time speed of the tire from the road condition state information.
For example, the electronic device screens information related to tire stress, such as acceleration, tire pressure, and tire size of the vehicle, from vehicle tire related information of the road condition information. Meanwhile, the real-time speed of the unmanned vehicle under the current road condition is screened from the road condition state information, for example, 0.01 second or 0.02 second can be set as a detection period, and the real-time speed of the unmanned vehicle in the detection period is obtained. It should be noted that the detection period may be set according to actual application requirements, and may be a positive value with an arbitrary value different from zero.
(2) And the electronic equipment calculates the stress information of the tire to obtain the ground friction under the current road condition.
For example, the electronic device may calculate tire force information using an ackerman bicycle model and a friction circle model in vehicle dynamics, where the ackerman bicycle model is shown in fig. 3. The constants that characterize the attributes of a vehicle in the ackermann bicycle model are: l
f,l
rRespectively, front and rear suspension length, C
f,C
rThe roll stiffness coefficients of the front wheel and the rear wheel are respectively, M is the mass of the whole vehicle, and M is
dFor equivalent mass of rotating parts, I
zThe moment of inertia of the whole vehicle. These constants are available at the time of vehicle shipment. The information selected from the vehicle information about tire stress may be v
y,v
xRespectively, transverse and longitudinal speed, a
y,a
xRespectively, the transverse acceleration and the longitudinal acceleration are the steering wheel rotation angles,
the angular velocity and the angular acceleration of the whole vehicle,
is the engine driving force. According to the screened stress information of the tire, the front and rear suspension speed direction angle can be calculated through a formula, and the specific formula is as follows:
wherein theta isfIs the front suspension speed direction angle, thetarIs the rear suspension speed direction angle.
The front and rear wheel lateral friction force can be calculated through a formula according to the screened stress information with the tire, and the specific formula is as follows:
Ff=2Cf(-θf),Fr=2Crθr
wherein, FfIs front wheel side friction force, FrA rear wheel side frictional force, Cf,Cr-front and rear wheel roll stiffness factor, -steering wheel angle.
And then, calculating longitudinal friction force of the front wheel and the rear wheel according to the screened stress information of the tires, wherein the specific formula is as follows:
wherein f isfLongitudinal friction of the front wheel, frLongitudinal friction of the rear wheel, ay,axRespectively, the transverse acceleration and the longitudinal acceleration are steering wheel corners, M is the mass of the whole vehicle, and M is the mass of the whole vehicledFor rotating parts of equivalent mass, FfIs front wheel side friction force, FrIs the rear wheel side friction force.
According to the longitudinal and lateral friction forces of the front and rear wheels, the friction force of the front and rear wheels of the unmanned vehicle can be calculated by adopting a friction circle stress model, and the specific formula is as follows:
wherein f is1For friction of front wheels of unmanned vehicles, f2Friction force of rear wheel of unmanned vehicle, ffLongitudinal friction of the front wheel, frLongitudinal friction of rear wheels, FfIs front wheel side friction force, FrIs the rear wheel side friction force.
(3) And the electronic equipment determines the sliding rate under the current road condition according to the real-time speed.
For example, the electronic device obtains the real-time speed and the wheel rotation speed of the vehicle from the vehicle information, and may calculate the slip ratio under the current road condition according to the relationship between the real-time speed and the wheel rotation speed of the unmanned vehicle, where the specific calculation formula is as follows:
wherein, ω is wheel speed (in rad), and r is the tread radius, takes tire mark as the standard, and v is the speed of a motor vehicle, needs to process the real-time speed of gathering and obtains, for example can obtain through locating data differential filtering.
(4) And the electronic equipment generates current road condition information according to the ground friction force and the sliding rate.
For example, the electronic device screens out the first road condition information in the road condition classifier according to the ground friction and the real-time speed under the current road condition, for example, the calculated ground friction of the front and rear wheels under the current road condition is f1And f2Real-time speed of vehicle is v1Screening out the value v in the road condition classifier1At speed, ground friction is f1And f2And corresponding road surface type, assuming the road surface type is wet road surface, the wet grade of the wet road surface can be further obtained according to the ground friction force,i.e. the thickness of the water film, according to the ground friction force f1And f2And acquiring the moisture grade of the wet road surface according to the corresponding road surface water film thickness, taking three levels as an example, assuming that the water film thickness is larger and exceeds the threshold value of the second level water film thickness, determining that the moisture grade of the wet road surface is the third grade, and determining that the first road condition information can be the third grade wet road surface.
For example, the electronic device screens the second road condition information in the road condition classifier according to the slip rate and the real-time speed under the current road condition, for example, the calculated slip rate under the current road condition is γ, and the real-time speed of the vehicle is v1Screening out the value v in the road condition classifier1At speed, the slip rate is the road type corresponding to γ, and assuming that the road type is a wet concrete road, the second road condition information is the wet concrete road.
For example, the electronic device fuses the first road condition information and the second road condition information to obtain the real-time road condition information. For example, when the first road condition information may be a third-level wet road surface, the second road condition information is a wet concrete road surface, the first road condition information and the second road condition information are fused, and the obtained road condition information is the third-level wet concrete road surface, the real-time road condition information may be the third-level wet concrete road surface.
205. And the electronic equipment adjusts the control parameters in the target state information based on the current road condition information to obtain adjusted parameters.
(1) And the electronic equipment screens out the external environment information of the vehicle from the real-time state information.
For example, the electronic device screens out environment information outside the vehicle, such as noise environment, air environment, and/or temperature environment outside the vehicle, from the classified real-time status information.
(2) The electronic equipment fuses the external environment information of the vehicle and the current road condition information to obtain interference information.
For example, the electronic device fuses the current road condition information and the external environment information of the vehicle to obtain the external interference information of the vehicle, for example, assuming that the current road condition information is a third-level wet concrete pavement, the initial interference information is environment information such as wind speed information, noise information, tire pressure, friction factor of the current pavement, and external temperature information of the vehicle, and the current road condition information and the initial interference information are directly combined to obtain the external interference information of the vehicle. The current road condition information can also be adjusted according to the initial interference information, for example, according to the internal and external temperature information of the vehicle, the road surface temperature and other information are added in the current road condition information.
(3) And the electronic equipment adjusts the control parameters in the target state information according to the interference information to obtain adjusted parameters.
And A1, classifying the interference information according to the interference type by the electronic equipment.
For example, the interference information may be classified by the electronic device according to interference types, for example, the interference information may be classified into direct interference or indirect interference, the direct interference may be classified into road interference, temperature interference, wind speed interference in air interference, the indirect interference may be classified into noise interference, humidity interference in air interference, and the like.
And A2, the electronic equipment adjusts the interference information of different types to obtain first interference values of different types.
For example, the interference information of different types is adjusted to obtain first interference values of different types, and the adjustment modes may be multiple, for example, the interference information of different types may be converted in a unified manner, for example, decibels of noise are converted into interference values, interference information of speed, road condition type, and the like is converted into interference values of a unified unit, and the conversion process may specifically be conversion according to preset conversion information, for example, an interference value of 10 corresponding to 50 decibels in the preset noise conversion information, an interference value of a concrete road surface in the preset road condition conversion information is 5, an interference value corresponding to an asphalt road surface is 8, and only the corresponding interference value needs to be directly screened from the preset road surface conversion information according to the road surface information, and the corresponding interference values are obtained by the same method for other interference information such as temperature, air, and the like.
A3, the electronic equipment weights the first interference values of different types according to a preset weight coefficient to obtain a second interference value of the target acceleration.
For example, the electronic device weights the different types of interference values according to preset weight coefficients to obtain a second interference value of the target acceleration, for example, taking the first interference values of the different types including a first interference value of a road condition a1, a first interference value of a noise a2, a first interference value of air A3, and a first interference value of a temperature a4 as an example, according to different interference degrees of the different types of interference information on the target acceleration, the preset weight coefficients corresponding to the four types of first interference values are x1, x2, x3, and x4, respectively, where the sum of the four types of weight coefficients is 1, and the four types of first interference values are multiplied by the weight coefficients, respectively, and the products of the four types of first interference values and the weight coefficients are accumulated to obtain the second interference value of the target acceleration.
And A4, the electronic equipment adjusts the target acceleration in the target state information according to the second interference value to obtain the adjusted first acceleration.
For example, the electronic device adjusts the target acceleration in the target state information according to the second interference value to obtain the adjusted first acceleration, for example, the second interference value may be directly calculated with a corresponding value in the target control parameter, and the calculation method may include simple four arithmetic operations or may include complex calculation. For example, the acceleration value and the second disturbance value may be directly subtracted or added, or the second disturbance value may be input into a predetermined calculation model to obtain a compensation value affecting the target acceleration, and the target acceleration may be directly added to the compensation value to obtain the adjusted first acceleration.
206. And the electronic equipment screens out the initial pedal amount corresponding to the first acceleration from the preset calibration parameter information.
(1) The electronic equipment screens out a calibration table corresponding to the current road condition information from the preset calibration parameter information, wherein the calibration table comprises a plurality of state areas and calculation models corresponding to the state areas.
For example, the electronic device screens out calibration tables corresponding to the current road condition information from the preset calibration parameter information, for example, the preset calibration parameter information includes calibration tables corresponding to various road condition types, taking the current road condition information as a third-stage wet concrete pavement as an example, the calibration tables of the third-stage wet concrete pavement are screened out from the preset calibration parameter information, as shown in fig. 5, a is acceleration, v is speed, u is pedal amount, a are model parameters, and the parameters are constants. The calibration table is divided into a plurality of state areas, namely a brake section, a dead zone and an accelerator section, by adopting an analytical model instead of a numerical value form, wherein 0-order continuity is kept among the areas, and each state area corresponds to one calculation model.
(2) And the electronic equipment determines a target state area corresponding to the adjusted parameters in a calibration table.
For example, the electronic device may determine the current state of the vehicle in the vehicle state information of the unmanned vehicle, and may determine the target state region corresponding to the first acceleration in the calibration table according to the state of the vehicle. For example, the unmanned vehicle is judged to be at a standstill, an idle speed or a high speed through the vehicle real-time state information, if the speed of the vehicle is 0 and the acceleration is 0 at this time, the state area corresponding to the first acceleration is a brake section, if the speed of the unmanned vehicle is small and the acceleration is large at this time, the unmanned vehicle is in an idle state (a starting stage) at this time, the state area corresponding to the first acceleration is a dead section, and if the speed of the unmanned vehicle is large and the acceleration is large at this time, the unmanned vehicle is in an accelerator increasing state at this time, the state area corresponding to the first acceleration is an accelerator section.
(3) The electronic equipment calibrates the first acceleration by adopting a calculation model of the target state area to obtain an initial pedal amount.
For example, the electronic device calibrates the first acceleration by using a calculation model of the target state area to obtain the initial pedal amount, for example, if the target state area corresponding to the first acceleration is the throttle section, the calculation model for calibrating the first acceleration can be obtained as follows from fig. 5:
α(v,u)=VTAu,V=(1,v,v2)T,u=(1,u,u2)T,A∈R3×3(segmentation)
Where α is a first acceleration, v is a real-time velocity of the unmanned vehicle, and A is a model parameter, which is a segmented 3 × 3 parameter matrix.
The first acceleration is calculated by referring to the calculation model, and an initial pedal amount for calibrating the first acceleration in the accelerator section can be obtained.
The preset calibration parameter information may be generated by other devices and provided to the unmanned vehicle control device, or may be generated by the unmanned vehicle control device, that is, before the step of "screening out the calibration table corresponding to the current traffic information in the preset calibration parameter information", the unmanned vehicle control method may further include:
acquiring vehicle test data under various road conditions;
identifying target anchor point data of vehicle characteristics under different road conditions according to the test data;
fitting boundary curves of each state area of the calibration table corresponding to different road conditions according to the target anchor point data;
and fitting continuous curved surfaces of each state area of the calibration table corresponding to different road conditions according to the boundary curve to obtain a calculation model of each state area.
For example, the electronic device generates preset calibration parameter information and mainly performs piecewise quadratic fitting on each state area by using key anchor point information. In actual drive test, the nonlinear degree of the mapping of the accelerator section is very high, and the other two sections are basically linear except for a low-speed section. This is because the throttle section contains complex engine characteristics, while the brake section only requires the hydraulic mechanism to apply a certain deceleration force. In addition, the acceleration characteristic of the accelerator stage has a large relationship with the current speed of the vehicle, and the secondary acceleration characteristic of the high-speed stage under the same horsepower is much inferior to the starting acceleration characteristic of the low-speed stage. Based on this, different fitting methods can be adopted for the throttle section and the other two sections, specifically as follows:
(1) an accelerator section: and carrying out segmented quadratic fitting on the speed and the pedal quantity.
(2) Dead zone, braking section: and carrying out high-low piecewise quadratic fitting on the speed, and only carrying out linear fitting on the pedal amount.
For example, taking the computational model fitting of the throttle segment as an example, a point-line-plane fitting idea is mainly adopted. As shown in fig. 6, the throttle section also includes a plurality of regions according to the difference of speed or acceleration, specifically as follows:
and C1, obtaining vehicle test data under various road conditions.
For example, the sensor or the real-time vehicle information collecting device collects test data under various road conditions, such as speed, acceleration, pedal amount, and path related information in one or more test periods under various road conditions.
And C2, identifying target anchor point data of the vehicle characteristics under different road conditions according to the vehicle test data.
The target anchor point data may include key data points describing characteristics of the vehicle, for example, a position point of the bottom of the wheel in the planned path from a to B, for example, a center point of a rear axle, and the anchor point data may include a speed, an acceleration, and a pedal amount of the position point.
For example, key anchor points describing characteristics of the vehicle are identified in the test data, for example, one or more fixed points in the planned path may be determined according to the characteristics of the vehicle, and key data of the one or more fixed points may be obtained and stored in a lattice form with a size of m × n. The speed and the pedal amount of the anchor point are uniformly distributed in the rectangular state space, and the acceleration is set as the average value of the internal sample points of the tee ball near the corresponding working condition point in the state space, specifically as follows:
amn=Average{a|a(v,u),(v,u)∈B[(vm,un),]}
wherein, amnThe acceleration of the working point is a first acceleration, v is a real-time speed, u is a pedal amount under the first acceleration, and B is a dot matrix set of the acceleration and the pedal amount.
And C3, fitting the boundary curve of the accelerator section of the calibration table corresponding to different road conditions according to the target anchor point data.
For example, according to the target anchor point data, the boundary regions of the status regions of the calibration table corresponding to different road conditions are fitted. For example, according to different road conditions, each road condition needs to be fitted into a calibration table separately, and the region division is performed according to a specific speed, for example, for a specific speed (longitudinal direction), the pedal amount curve is divided into (n-1) sections; for a certain pedal amount (lateral direction), the velocity curve is divided into (m-1) segments. The principle of fitting is to keep the 1 st order continuous while making the curve as smooth as possible, i.e. the modulus of the quadratic coefficient is as small as possible. Taking the jth longitudinal pedal magnitude curve as an example, the fitting process is to solve the following quadratic convex optimization problem with linear constraint, specifically as follows:
2aiui+βi=2ai+1ui+βi+1,i=1,...,n-2
wherein alpha, beta and gamma are respectively a quadratic term, a primary term and a constant term coefficient, and u and a are respectively anchor values of pedal amount and acceleration. It can be seen that the scale of the parameter to be fitted is 3(n-1), and the number of constraints is 2(n-1) + (n-2), and there is an optimization space of 1 degree of freedom. The problem can be converted into a least square problem by using the existing Lagrange multiplier method in convex optimization and solved.
And C4, fitting the continuous curved surface of the throttle section in the calibration table corresponding to different road conditions according to the boundary curve to obtain a calculation model of the throttle section.
For example, according to the boundary curves of different areas, continuous curved surfaces of throttle sections in the calibration table corresponding to different road conditions are fitted to obtain a calculation model of the throttle sections. For example, a continuous curved surface of the slice quadratic is fitted according to the boundary curve, and the slice granularity is (m-1) × (n-1). The principle of fitting is to keep 0 th order continuous and make the curved surface as smooth as possible, i.e. the modulus of the quadratic coefficient is as small as possible. Taking the curved surface (p, q) as an example, the fitting process is to solve the following quadratic convex optimization problem with linear constraint, specifically as follows:
A∈A3×3
wherein alpha, beta and gamma are respectively a quadratic term, a primary term and a constant term coefficient of a pedal amount curve, v, mu and lambda are respectively a quadratic term, a primary term and a constant term coefficient of a speed curve, u and v are respectively anchor points of pedal amount and speed, and A is a parameter matrix to be fitted.
From the above calculation process, the scale of the parameter to be fitted is 9, the number of constraints is 8 (the 12 constraints are linearly related, the rank number is 8, and there are 4 redundant constraints from the curve overlap at the anchor point), and there is an optimization space with 1 degree of freedom. The problem can be converted into a least square problem by using the existing Lagrange multiplier method in convex optimization and solved.
207. And the electronic equipment compensates the initial pedal amount according to the error information between the real-time state information and the target state information to obtain the pedal amount.
(1) The electronic device extracts a current value of the acceleration from the real-time status information and extracts a target value of the acceleration from the target status information.
For example, the electronic device extracts the current value of the control parameter in the real-time status information and extracts the target value of the control parameter in the target status information. For example, the current value of the vehicle acceleration of the vehicle under the current road condition is extracted from the real-time state information, and the target value of the vehicle acceleration is extracted from the target state information.
(2) The electronic device compares the current value to a target value to obtain error information.
For example, the electronic device compares the current acceleration with the target acceleration to obtain an error value, for example, compares the actual value of the acquired acceleration of the unmanned vehicle in the actual driving process in the planned path area with the target value to obtain a difference value of the acceleration of the unmanned vehicle in the planned path area, and uses the difference value as error information.
(3) And the electronic equipment compensates the initial pedal amount according to the error information to obtain the pedal amount.
For example, the electronic device may screen the compensation value corresponding to the error information from the preset compensation information, for example, a proportional-integral-derivative control model (PID) may be used to screen the compensation value corresponding to the error information from the preset compensation information, and the PID may be placed in a bottom controller of the unmanned vehicle control device, where the bottom controller has a structure as shown in fig. 8, where a
tar,a
cmd,a
cur,a
errTarget value, command value, current value and error value, u, respectively, representing acceleration
ff,u
fb,u
cmdA feedforward value representing the amount of pedal, respectively, a feedback value and a command value omega representing disturbance terms,
represents observed/calculated/estimated values of the interference term. Specifically, when the error value of the acceleration is obtained, the error value of the acceleration is subjected to 3 operation modes of proportion, integration and differentiation to obtain the compensation value of the pedal amount under the current road condition and the current speed.
208. The electronic device controls the unmanned vehicle based on the pedal amount.
For example, based on the pedal amount, a corresponding control command may be generated, and the unmanned vehicle may be controlled according to the control command. For example, a control instruction corresponding to the pedal amount is generated according to the pedal amount, and the control instruction is input to a control system of an engine of the unmanned vehicle, so that the unmanned vehicle is controlled by actuating the corresponding pedal amount of the accelerator pedal or the brake pedal, and the target value of the acceleration of the unmanned vehicle under the current road condition can be achieved.
As can be seen from the above, after the electronic device acquires the vehicle state information of the unmanned vehicle, the vehicle state information includes real-time state information and target state information, identifies a road condition environment according to the real-time state information to obtain current road condition information, then adjusts a control parameter in the target state information based on the current road condition information to obtain an adjusted parameter, screens out a target calibration parameter corresponding to the adjusted parameter from preset calibration parameter information, then compensates the target calibration parameter according to error information between the real-time state information and the target state information to obtain a compensated parameter, and controls the unmanned vehicle based on the compensated parameter; according to the scheme, real-time road condition recognition is carried out on the basis of vehicle dynamic characteristics by collecting tire stress information of the vehicle, and multiple factors such as control parameters are adjusted and compensated according to the external environment and response information of the vehicle, so that the accuracy of unmanned vehicle control is greatly improved.
In order to better implement the above method, embodiments of the present invention also provide an unmanned vehicle control apparatus that may be integrated in an electronic device, such as a server or an ECU.
For example, as shown in fig. 10, the unmanned vehicle control apparatus may include an acquisition unit 301, a recognition unit 302, an adjustment unit 303, a screening unit 304, a compensation unit 305, and a control unit 306, as follows:
(1) an acquisition unit 301;
an obtaining unit 301, configured to obtain vehicle state information of the unmanned vehicle, where the vehicle state information includes real-time state information and target state information;
for example, the obtaining unit 301 may be specifically configured to obtain real-time speed and acceleration of the vehicle in the current state through a speed and acceleration sensor outside the vehicle, obtain environmental information inside and outside the vehicle through an environmental sensor of the vehicle, and obtain information such as real-time rotation speed, torque and/or power of an engine of the vehicle in the current state through an engine monitoring sensor.
(2) An identification unit 302;
the identification unit 302 is configured to identify a road condition environment according to the real-time status information to obtain current road condition information;
the identifying unit 302 may include a classifying subunit 3021, a first extracting subunit 3022, and an identifying subunit 3023, as shown in fig. 11, specifically as follows:
a classification subunit 3021 configured to classify the real-time status information according to a type of an external environment of the vehicle;
a first extracting subunit 3022, configured to extract the traffic status information related to the traffic environment from the classified real-time status information;
the identifying subunit 3023 is configured to identify the traffic status information to obtain current traffic information.
For example, the classification subunit 3021 classifies the real-time status information according to the type of the external environment of the vehicle, the extraction subunit 3022 extracts the traffic status information related to the traffic environment from the classified real-time status information, and the identification subunit 3023 identifies the traffic status information to obtain the current traffic information.
(3) An adjustment unit 303;
an adjusting unit 303, configured to adjust a control parameter in the target status information based on the current traffic information, to obtain an adjusted parameter.
The adjusting unit 303 may include a screening subunit 3031, a fusion subunit 3032, and an adjusting subunit 3033, as shown in fig. 12, specifically as follows:
the screening subunit 3031 is used for screening the vehicle external environment information from the real-time state information;
a fusion subunit 3032, configured to fuse the vehicle external environment information and the current road condition information to obtain interference information;
and an adjusting subunit 3033, configured to adjust the control parameter in the target state information according to the interference information, to obtain an adjusted parameter.
For example, the screening subunit 3031 screens the vehicle external environment information from the real-time status information, the fusion subunit 3032 fuses the vehicle external environment information and the current road condition information to obtain interference information, and the adjusting subunit 3033 adjusts the control parameter in the target status information according to the interference information to obtain an adjusted parameter.
(4) A screening unit 304;
a screening unit 304, configured to screen out, from preset calibration parameter information, a target calibration parameter corresponding to the adjusted parameter.
The screening unit 304 may include a screening subunit 3041, a determining subunit 3042, and a calibrating subunit 3043, as shown in fig. 13, specifically as follows:
a screening subunit 3041, configured to screen a calibration table corresponding to the current road condition information from preset calibration parameter information, where the calibration table includes a plurality of state areas and a calculation model corresponding to each state area;
a determining subunit 3042, configured to determine, in the calibration table, a target state area corresponding to the adjusted parameter;
the calibration subunit 3043 is configured to calibrate the adjusted parameters by using the calculation model of the target state area, so as to obtain target calibration parameters.
For example, the screening subunit 3041 screens a calibration table corresponding to the current traffic information from preset calibration parameter information, where the calibration table includes a plurality of status areas and calculation models corresponding to the status areas, the determining subunit 3042 determines a target status area corresponding to the adjusted parameter in the calibration table, and the calibrating subunit 3043 calibrates the adjusted parameter by using the calculation model of the target status area to obtain a target calibration parameter.
(5) A compensation unit 305;
and a compensating unit 305, configured to compensate the target calibration parameter according to error information between the real-time status information and the target status information, so as to obtain a compensated parameter.
The compensation unit 305 may include a second extraction subunit 3051, a comparison subunit 3052, and a compensation subunit 3053, as shown in fig. 14, which are as follows:
a second extraction subunit 3051, configured to extract a current value of the control parameter from the real-time status information, and extract a target value of the control parameter from the target status information;
a comparison subunit 3052, configured to compare the current value with the target value to obtain error information;
and the compensation subunit 3053, configured to compensate the target calibration parameter according to the error information, to obtain a compensated parameter.
For example, the second extraction subunit 3051 extracts the current value of the control parameter from the real-time status information, extracts the target value of the control parameter from the target status information, the comparison subunit 3052 compares the current value with the target value to obtain error information, and the compensation subunit 3053 compensates the target calibration parameter according to the error information to obtain a compensated parameter.
(6) A control unit 306;
and the control unit 306 is used for controlling the unmanned vehicle based on the compensated parameters.
For example, the control unit 306 may be specifically configured to generate a control instruction corresponding to the pedal amount according to the pedal amount, and input the control instruction to a control system of an engine of the unmanned vehicle, so that the unmanned vehicle is controlled by actuating a corresponding pedal amount of an accelerator pedal or a brake pedal, and a target value of the acceleration of the unmanned vehicle under the current road condition can be achieved.