Model calibration method, algorithm test method and related device

Abstract

A model calibration method, an algorithm test method and a related device, the related method comprising: obtaining driving behavior information of a target vehicle, the driving behavior information comprising information generated by each test scene in a plurality of test scenes in response to a first operation, risk vehicles in the plurality of test scenes being located in different orientations of the target vehicle; according to the relative position between the risk vehicle and the target vehicle in each test scene and the driving behavior information, a driver decision model in each test scene is calibrated, and the driver decision model indicates the risk avoiding ability of a driver in each test scene. Therefore, the obstacle avoidance capability of the automatic driving algorithm can be tested based on the driver decision model. The scheme of the invention can be applied to the field of intelligent driving, and is helpful for improving the integrating degree of the driver decision model and the human driver decision ability, thereby improving the test result accuracy of an automatic driving algorithm.

Description

Technical Field

[0001] The present application relates to the field of intelligent driving technology, and more specifically, to a model calibration method, an algorithm testing method, and related devices. Background Art

[0002] As vehicles develop towards intelligence and automation, more and more vehicles are equipped with autonomous driving technology. Autonomous driving technology relies on various autonomous driving algorithms to achieve. Among them, the decision-making algorithm is one of the key algorithms for realizing autonomous driving technology, which determines whether the vehicle can smoothly and accurately complete various driving behaviors. Before the actual commercial application of autonomous driving vehicles, it is generally necessary to use simulation testing to test and optimize the autonomous driving algorithms installed on them, so that the safety performance of autonomous driving vehicles meets the requirements.

[0003] When using simulation testing to test the decision-making algorithm, the driving ability of human drivers can be used as a safety boundary to judge whether the safety performance of the decision-making algorithm to be tested meets the standard. In the current technical background, a driver decision-making model is generally constructed based on the driving behavior data of human drivers, and the decision-making ability of human drivers during driving is characterized by the behavior of the driver decision-making model in the simulation scenario, that is, the safety boundary is determined according to the performance of the driver decision-making model in the driving scenario, and then the decision-making algorithm to be tested is tested and / or optimized based on the driver decision-making model.

[0004] However, in the current technical background, the driving scenarios constructed during the calibration of the driver decision-making model are relatively single, and the parameters selected during the calibration of the model are not comprehensive enough, resulting in the existing driver decision-making model being unable to accurately represent the decision-making ability of human drivers during driving, and further making the test results based on this driver decision-making model inaccurate. Summary of the Invention

[0005] The present application provides a model calibration method, an algorithm testing method, and related devices, which helps to improve the degree of fit between the driver decision-making model and the decision-making ability of human drivers, and further improve the accuracy of the test results of the autonomous driving algorithm executed based on this driver decision-making model, thereby improving driving safety.

[0006] In a first aspect, a model calibration method is provided. This method can be executed by an electronic device or by components (such as chips, circuits, etc.) of an electronic device. The method includes: obtaining driving behavior information of a target vehicle, where the driving behavior information includes information generated in each of multiple test scenarios in response to a first operation, and the risk vehicles in the multiple test scenarios are located in different directions of the target vehicle; calibrating a driver decision-making model for each test scenario according to the relative position between the risk vehicle and the target vehicle in each test scenario and the driving behavior information, where the driver decision-making model indicates the risk avoidance ability of the driver in each test scenario.

[0007] In actual implementation, different risk sources (i.e., the orientation of the risk vehicle relative to the target vehicle) have significant differences in their impacts on the driver, which can lead to significantly different decisions made by the driver. Therefore, in the above technical solution, calibrating the driver decision-making model through multiple test scenarios where the risk vehicle is located in different orientations relative to the target vehicle helps improve the accuracy of the driver decision-making model, making the driver decision-making model more in line with the decision-making ability of human drivers.

[0008] Combined with the first aspect, in some implementation manners of the first aspect, the multiple test scenarios include a first test scenario and a second test scenario. The first risk vehicle in the first test scenario is located at a first orientation relative to the target vehicle, and the second risk vehicle in the second test scenario is located at a second orientation relative to the target vehicle, where the first orientation is different from the second orientation. Calibrating the driver decision-making model for each test scenario includes: calibrating the first driver decision-making model based on the relative position between the first risk vehicle and the target vehicle, and the first driving behavior information; calibrating the second driver decision-making model based on the relative position between the second risk vehicle and the target vehicle, and the second driving behavior information; where the first driving behavior information is the information generated in the first test scenario, and the second driving behavior information is the information generated in the second test scenario.

[0009] In the above technical solution, calibrating the driver decision-making model with the driving behavior information generated by risk vehicles located in different orientations relative to the target vehicle, and testing the obstacle avoidance ability of the autonomous driving algorithm based on the calibrated driver decision-making model, helps improve the accuracy of the test results of the autonomous driving algorithm, and further improves the safety of the autonomous driving algorithm.

[0010] Combined with the first aspect, in some implementation manners of the first aspect, the driving behavior information includes multiple steering wheel information. Each steering wheel information in the multiple steering wheel information indicates a steering behavior feature in the first test scenario, and the multiple steering wheel information corresponds to the situation of successful risk avoidance. Calibrating the first driver decision-making model includes: determining at least one of the following steering parameters based on the relative position between the first risk vehicle and the target vehicle and each steering wheel information: decision response duration, maximum steering wheel angle, or steering wheel return duration; calibrating the first driver decision-making model based on the at least one steering parameter.

[0011] In an emergency scenario where a possible collision may occur, in addition to braking, steering can also be used for risk avoidance. Therefore, in the above technical solution, calibrating the driver decision-making model with the steering parameters determined based on the steering behavior helps improve the universality of the driver decision-making model in multiple scenarios.

[0012] In combination with the first aspect, in certain implementations of the first aspect, the driving behavior information includes multiple pedal information, each pedal information in the multiple pedal information indicates a braking behavior feature in the first test scenario, and the multiple pedal information corresponds to a successful risk avoidance situation; calibrating the first driver decision model includes: determining at least one of the following braking parameters based on the relative position between the first risk vehicle and the target vehicle and each pedal information: decision response time, maximum braking deceleration, braking efficiency improvement time, or braking efficiency improvement rate; calibrating the first driver decision model based on at least one braking parameter.

[0013] In the above technical solution, the driver decision model is calibrated by decision response time, maximum braking deceleration, braking efficiency improvement time, and braking efficiency improvement rate, which helps to improve the comprehensiveness and effectiveness of the decision-making ability of the driver decision model, thereby improving the fit between the driver decision model and the decision-making ability of human drivers.

[0014] In combination with the first aspect, in certain implementations of the first aspect, the multiple test scenarios include at least one of the following: the risk vehicle is located in the lane where the target vehicle is located at a first moment, the risk vehicle is located in front of the target vehicle and there is a risk obstacle within a first range in front of the risk vehicle, and the risk vehicle changes to an adjacent lane of the lane where the target vehicle is located at the first moment at a second moment, and the risk obstacle is an obstacle that poses a risk of collision with the target vehicle at the second moment; the risk vehicle is located in the lane where the target vehicle is located at a first moment, the risk vehicle is located in front of the target vehicle and within a second range of the target vehicle, and the speed of the risk vehicle at the first moment is greater than the speed of the risk vehicle at the second moment; or the risk vehicle is located in an adjacent lane of the lane where the target vehicle is located at a first moment, and the risk vehicle is located in the lane where the target vehicle is located at the first moment and the risk vehicle is located in front of the target vehicle at the second moment; wherein the second moment is a moment after the first moment.

[0015] In some implementations, the decision response time varies depending on the first test scenario. For example, in the first scenario, if the risk obstacle is a vehicle, the decision response time is determined based on the time when the risk vehicle cuts out to the target vehicle's driver's seat and can see the lateral quarter of the risk obstacle; in the second scenario, the decision response time is determined based on the time when the risk vehicle's braking deceleration reaches 5m / s. 2 In the third scenario, the decision response time is determined based on the time when the center of mass of the risk vehicle deviates to a preset distance from the driving centerline. The preset distance can be determined based on the model of the risk vehicle.

[0016] In the above technical solution, selecting a scenario with a higher risk level as a test scenario is helpful in obtaining a driver decision-making model in an emergency scenario.

[0017] In combination with the first aspect, in some implementations of the first aspect, calibrating the driver decision-making model for each test scenario includes: when the risk vehicle is in the adjacent lane of the lane where the target vehicle is located, calibrating the driver decision-making model according to the relative position between the risk vehicle and the target vehicle, the vehicle type of the risk vehicle, and the driving behavior information.

[0018] Different vehicle types of vehicles have different impacts on drivers. For example, when there is a queue of large buses on the left side of the target vehicle, the driver of the target vehicle may always pay attention to the behavior dynamics of the large buses and have a psychological expectation of dangerous situations. In addition, the steering and lane-changing behaviors of large vehicles are more obvious, and drivers can clearly distinguish them when the steering behavior just occurs. On the contrary, due to the short length of small vehicles, it is difficult for drivers to distinguish whether there is a collision risk from the head angle deviation. Therefore, in the above technical solution, when the risk vehicle is in the adjacent lane of the lane where the target vehicle is located, calibrating the driver decision-making model based on the vehicle type of the risk vehicle helps to improve the accuracy of the driver decision-making model, that is, makes the driver decision-making model closer to the decisions of real drivers in the same scenario.

[0019] In combination with the first aspect, in some implementations of the first aspect, the first test scenario includes at least one vehicle other than the target vehicle and the risk vehicle, and the at least one vehicle is within the third range of the target vehicle.

[0020] The driving behavior information is the information generated by the test driver during driving in the first test scenario. When only the target vehicle and the risk vehicle exist in the first test scenario, the test driver may make premature judgments and thus take preventive driving behaviors to avoid risks, making the driving behavior information unable to accurately reflect the driver's decisions in emergency scenarios. Therefore, in the above technical solution, setting other vehicles around the target vehicle helps to improve the accuracy of the driving behavior information, thereby improving the accuracy of the calibration result of the driver decision-making model.

[0021] In a second aspect, an algorithm test method is provided. This method can be executed by an electronic device or by components of an electronic device (such as chips, circuits, etc.). The method includes: determining the first risk avoidance success rate of the decision-making algorithm to be tested in the test scenarios indicated by the test scenario set; determining whether the risk avoidance ability of the decision-making algorithm to be tested meets the standard according to the second risk avoidance success rate and the first risk avoidance success rate; where the second risk avoidance success rate is the risk avoidance success rate determined by running the driver decision-making model in the test scenarios indicated by the test scenario set, the driver decision-making model is calibrated according to the driving behavior information of the target vehicle and the relative positions between the risk vehicle and the target vehicle in each test scenario of multiple test scenarios, the driving behavior information includes the information generated in response to the first operation in each test scenario of multiple test scenarios, and the risk vehicles in multiple test scenarios are located in different orientations of the target vehicle.

[0022] In the above technical solution, performing an automatic driving algorithm test according to the driver decision-making model calibrated in the first aspect helps to improve the accuracy of the test results, so that algorithm developers can optimize based on the test results, which helps to improve the safety and reliability of the automatic driving algorithm.

[0023] Combined with the second aspect, in some implementation manners of the second aspect, the driving behavior information includes a plurality of steering wheel information in each test scenario, each steering wheel information in the plurality of steering wheel information indicates a steering behavior feature, and the plurality of steering wheel information corresponds to the situation of successful risk avoidance; and, the driver decision-making model in each test scenario is calibrated according to at least one steering parameter, and the at least one steering parameter is determined according to the relative position between the risk vehicle and the target vehicle and each steering wheel information, and the at least one steering parameter includes at least one of a decision response duration, a maximum steering wheel rotation angle, or a steering wheel return duration.

[0024] Combined with the second aspect, in some implementation manners of the second aspect, the driving behavior information includes a plurality of pedal information in each test scenario, each pedal information in the plurality of pedal information indicates a braking behavior feature, and the plurality of pedal information corresponds to the situation of successful risk avoidance; and, the driver decision-making model in each test scenario is calibrated according to at least one braking parameter, and the at least one braking parameter is determined according to the relative position between the risk vehicle and the target vehicle and each pedal information, and the at least one braking parameter includes at least one of a decision response duration, a maximum braking deceleration, a braking efficiency improvement duration, or a braking efficiency improvement rate.

[0025] Combined with the second aspect, in some implementation manners of the second aspect, the driver decision-making model in each test scenario is calibrated according to the relative position between the risk vehicle and the target vehicle, the vehicle type of the risk vehicle, and the driving behavior information.

[0026] Combined with the second aspect, in some implementation manners of the second aspect, the plurality of test scenarios include at least one of the following:

[0027] At the first moment, the risk vehicle is in the lane where the target vehicle is located, the risk vehicle is in front of the target vehicle, and there is a risk obstacle within the first range in front of the risk vehicle. And at the second moment, the risk vehicle changes to the adjacent lane of the lane where the target vehicle is located at the first moment, and the risk obstacle is an obstacle that has a risk of collision with the target vehicle at the second moment; at the first moment, the risk vehicle is in the lane where the target vehicle is located, the risk vehicle is in front of the target vehicle and within the second range of the target vehicle, and the speed of the risk vehicle at the first moment is greater than the speed of the risk vehicle at the second moment; or at the first moment, the risk vehicle is in the adjacent lane of the lane where the target vehicle is located, and at the second moment, the risk vehicle is in the lane where the target vehicle is located at the first moment and the risk vehicle is in front of the target vehicle; wherein, the second moment is a moment after the first moment.

[0028] In combination with the second aspect, in some implementation manners of the second aspect, each test scenario includes at least one vehicle other than the target vehicle and the risk vehicle, and the at least one vehicle is within the third range of the target vehicle.

[0029] In a third aspect, there is provided an apparatus for performing the method provided in the first aspect above, or for performing the method provided in the second aspect above. Specifically, the apparatus may include units and / or modules for performing the method provided in the first aspect or any one of the above implementation manners of the first aspect. Alternatively, the apparatus may include units and / or modules for performing the method provided in the second aspect or any one of the above implementation manners of the second aspect, such as an acquisition module and a processing module.

[0030] In a fourth aspect, there is provided a model calibration apparatus, the apparatus includes: a memory for storing a computer program; a processor for executing the computer program stored in the memory so that the apparatus performs the method in any one of the possible implementation manners in the first aspect.

[0031] In a fifth aspect, there is provided an algorithm testing apparatus, the apparatus includes: a memory for storing a computer program; a processor for executing the computer program stored in the memory so that the apparatus performs the method in any one of the possible implementation manners in the second aspect.

[0032] In a sixth aspect, there is provided an electronic device, the electronic device includes the apparatus in any one of the possible implementation manners in the third to fifth aspects.

[0033] In a seventh aspect, there is provided a computer program product, the computer program product includes: computer program code, when the computer program code runs on a computer, it causes the computer to perform the method in any one of the possible implementation manners in the first aspect or the second aspect.

[0034] It should be noted that the above computer program code can be stored in whole or in part on the first storage medium, where the first storage medium can be packaged together with the processor or separately packaged from the processor.

[0035] In an eighth aspect, a computer-readable medium is provided, and the computer-readable medium stores instructions that, when executed by a processor, cause the processor to implement the method in any one of the possible implementation manners of the first aspect or the second aspect.

[0036] In a ninth aspect, a chip is provided, and the chip includes a circuit for executing the method in any one of the possible implementation manners of the first aspect or the second aspect. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] Figure 1 FIG. is a schematic diagram of an application scenario of the model calibration method provided by an embodiment of the present application;

[0038] Figure 2 FIG. is a schematic flowchart of obtaining driving behavior data provided by an embodiment of the present application;

[0039] Figure 3 FIG. is a schematic diagram of a test scenario provided by an embodiment of the present application;

[0040] Figure 4 FIG. is a schematic flowchart of the model calibration method provided by an embodiment of the present application;

[0041] Figure 5 FIG. is another schematic flowchart of the model calibration method provided by an embodiment of the present application;

[0042] Figure 6 FIG. is a schematic diagram of a driver decision-making model provided by an embodiment of the present application;

[0043] Figure 7 FIG. is yet another schematic flowchart of the model calibration method provided by an embodiment of the present application;

[0044] Figure 8 FIG. is a schematic diagram of the change of braking deceleration with time provided by an embodiment of the present application;

[0045] Figure 9 FIG. is a schematic flowchart of an algorithm test method provided by an embodiment of the present application;

[0046] Figure 10 FIG. is a schematic block diagram of a related device provided by an embodiment of the present application. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0047] As described above, the test scenarios currently constructed during the calibration of the driver decision-making model are relatively single. For example, generally only the scenario of emergency braking when a neighboring vehicle cuts in is considered. Moreover, the parameters selected during the calibration of the model are not comprehensive enough. For example, only two parameters, namely the decision response time and the maximum braking deceleration, are adopted. In addition, the currently constructed test scenarios are inductive. For example, there is only one other vehicle in the test scenario, which makes it easy for the tested driver to make advance predictions, and thus take preventive driving behaviors to avoid risks, unable to reflect the numerical values of the real decision-making model. The above drawbacks lead to the fact that the existing driver decision-making models cannot accurately represent the decision-making ability of human drivers during driving.

[0048] In view of this, the embodiments of the present application provide a model calibration method, an algorithm testing method, and related devices, which calibrate the driver decision-making model according to the position of the risk vehicle relative to the own vehicle in the test scenario, so as to improve the accuracy of the driver decision-making model.

[0049] To facilitate the understanding of the technical solutions of the present application, the terms related to the present application are introduced below.

[0050] Model calibration: Also known as parameter optimization, it is a process of making the simulation result variables match the measurement data by changing some model parameters.

[0051] Next, the technical solutions in the present application will be described in conjunction with the accompanying drawings.

[0052] Figure 1 It is a schematic diagram of the application scenario of the model calibration method provided by the embodiments of the present application, or a schematic block diagram of an algorithm testing system. Figure 1 The system shown includes a model calibration device 100 and an algorithm testing device 200. Among them, the model calibration device 100 includes an acquisition module 110 and a processing module 120; the algorithm testing device 200 includes a test execution module 210. The acquisition module 110 is used to acquire test data, and the test data may include data generated by different drivers operating the same vehicle (i.e., the own vehicle) in the same test scenario. The test data may include braking avoidance data and steering avoidance data. Among them, the braking avoidance data includes the opening data of the brake pedal during the vehicle's avoidance process, and the steering avoidance data includes the steering angle data of the steering wheel during the vehicle's avoidance process. Further, the processing module 120 calibrates the driver decision-making model according to the relative position between the risk vehicle and the target vehicle in the test scenario, and the test data.

[0053] The driver decision-making model calibrated by the model calibration device 100 can be input into the algorithm testing device 200, and the test execution module 210 can test the decision-making algorithm to be tested based on the driver decision-making model to determine whether the risk avoidance ability of the decision-making algorithm to be tested meets the standard.

[0054] The system related to the present application is introduced above. The method provided by the present application will be introduced in detail below.

[0055] Figure 2 FIG. shows a schematic diagram of a process for collecting data in the early stage of model calibration provided by an embodiment of the present application. The process includes steps S201 to S203.

[0056] S201, construct a simulation scenario.

[0057] Exemplarily, a simulation scenario is constructed in a driving simulation platform. Constructing the simulation scenario includes setting the following basic parameters in response to user operations: weather conditions, road surface adhesion coefficient, maximum braking deceleration, braking force curve, vehicle dynamics model of the host vehicle, physical model of other vehicles, physical feedback of road edges, steering wheel force feedback, and cockpit force feedback. Among them, the weather conditions can be set to sunny and windless during the day; the road surface adhesion coefficient can be set to 1.0; the maximum braking deceleration can be set to 10 m / s^2; the braking force curve can be fitted by three straight lines, that is, the higher the pedal stroke, the slower the braking force increases; the vehicle dynamics model of the host vehicle uses the default dynamics model of the driving simulation platform, including various physical constraints and motion feedback; the physical models of other vehicles, physical feedback of road edges, steering wheel force feedback, cockpit force feedback, etc. all use the default settings of the driving simulation platform.

[0058] It should be understood that the above parameter settings are all exemplary descriptions, and other parameters can also be set in actual implementation.

[0059] Constructing the simulation scenario also includes setting vehicle behavior parameters and the positional relationship between other vehicles and the host vehicle. The vehicle behavior parameters mainly include the behavior of the host vehicle and the behavior of other vehicles. The behavior of other vehicles includes lateral steering behavior and longitudinal speed behavior. The lateral steering behavior includes lane keeping, cutting in, cutting out, etc.; the longitudinal speed behavior includes stationary, constant speed, acceleration, deceleration, etc. Specifically, based on the vehicle behavior parameters and the positional relationship between other vehicles and the host vehicle, the following 3 test scenarios are provided by the embodiment of the present application: the leading vehicle cuts out, the leading vehicle brakes, and the adjacent vehicle cuts in.

[0060] The positional relationship between other vehicles and the host vehicle includes: the other vehicle is in front of the host vehicle, and at this time the other vehicle can be called the leading vehicle (such as the other vehicle at position 2 as shown in Figure 3 ); the other vehicle is in the adjacent lane of the lane where the host vehicle is located, and at this time the other vehicle can be called the adjacent vehicle (such as the other vehicles at positions 1, 4, 6, 3, 5, 8 as shown in Figure 3 ); or, the other vehicle is behind the host vehicle, and at this time the other vehicle can become the following vehicle (such as the other vehicle at position 7 as shown in Figure 3 ).

[0061] Based on the above position relationships, the leading vehicle cutting out can refer to another vehicle at position 2 changing to another lane; the leading vehicle braking can refer to another vehicle at position 2 decelerating; an adjacent vehicle cutting in can refer to a vehicle at any of positions 1, 4, 3, or 5 changing to the lane where the host vehicle is currently located, or an adjacent vehicle cutting in can also be a vehicle at position 6 or 8 respectively driving to positions 1 (or 4), 3 (or 5) and changing to the lane where the host vehicle is currently located.

[0062] In the scenario of an adjacent vehicle cutting in, since the type of the adjacent vehicle may affect the driver's decision-making, the adjacent vehicle cutting in can be further divided into a car cutting in and a large vehicle cutting in. Then, in the scenario of an adjacent vehicle cutting in, constructing a simulation scenario further includes setting the type of the adjacent vehicle. Exemplarily, the above-mentioned cars can include A00-class (mini cars), A0-class (subcompact cars), A-class (compact cars), and large vehicles can include B-class (mid-size cars), C-class (large mid-size cars), D-class (luxury cars).

[0063] It should be noted that in the above-mentioned various test scenarios, in addition to the host vehicle and the risk vehicle, at least one other vehicle is also set around the host vehicle. Among them, the risk vehicle refers to a vehicle that has a collision risk with the host vehicle, such as the vehicle that cuts into the lane where the host vehicle is currently located in the scenario of an adjacent vehicle cutting in, or the leading vehicle in the scenario of the leading vehicle braking, or the other vehicle (such as Figure 3 the other vehicle 2) originally in front of the leading vehicle in the scenario of the leading vehicle cutting out.

[0064] S202, determine the first test scenario.

[0065] Exemplarily, determine the first test scenario from the above four test scenarios (i.e., the leading vehicle braking, the leading vehicle cutting out, the car cutting in, the large vehicle cutting in), that is, the first test scenario is one of the leading vehicle braking, the leading vehicle cutting out, the car cutting in, and the large vehicle cutting in.

[0066] S203, record driving behavior data based on the first test scenario. The driving behavior data includes host vehicle behavior data and risk vehicle behavior data, and is used to calibrate the driver decision-making model.

[0067] In some implementation manners, record driving data based on the operations of each of N drivers in the first test scenario, where N is a positive integer. Exemplarily, N can be 30, or 50, or can also be other values. Among them, the N drivers can be drivers with good driving habits and long driving ages. For example, the N drivers can meet the following requirements: the number of major-responsible traffic accidents in the past three years does not exceed a preset number and the average annual driving mileage is greater than or equal to a preset mileage. Among them, the preset number can be 2, or 3, or other numbers; the preset mileage can be 2000 kilometers, or 5000 kilometers, or can also be other mileage.

[0068] Among them, the self-vehicle behavior data includes the data generated by the self-vehicle in response to the driver's operation within a first duration. For example, it includes the brake pedal information generated in response to the driver's operation of stepping on the brake pedal, and the brake pedal information may include the data of the change in the brake pedal opening over time; or, it may also include the steering wheel information generated in response to the driver's operation of turning the steering wheel, and the steering wheel information may include the data of the change in the steering wheel angle over time; or, it may further include the data of the speed, acceleration, and position of the self-vehicle within the first duration. The risk vehicle behavior data may include the data of the position, speed, and acceleration of the risk vehicle within the first duration. The above first duration may be the total duration of the test process, or it may also be a partial duration of the test process. For example, it may be the duration from when the risk vehicle starts to change lanes or starts to brake until it collides with the self-vehicle, or until the collision risk is eliminated during the test process.

[0069] So far, the driving behavior data has been recorded. Next, a method for model calibration based on this driving behavior data will be introduced.

[0070] Figure 4 The schematic flowchart of the model calibration method provided by an embodiment of the present application is shown. This method can be Figure 1 executed by the model calibration device 100 shown, and this method may include S410 and S420.

[0071] S410, obtain the driving information of the target vehicle. The driving behavior information includes the information generated in response to a first operation in each of multiple test scenarios, and the risk vehicles in the multiple test scenarios are located in different positions relative to the target vehicle.

[0072] Among them, the target vehicle includes the self-vehicle in the above embodiment, and the multiple test scenarios may be the multiple test scenarios in the above embodiment, such as the first test scenario. The first operation may be the operation of stepping on the brake pedal, or it may also be the operation of turning the steering wheel. The driving behavior information is associated with the above driving data. For example, the driving behavior information includes the driving data.

[0073] In some implementation manners, the multiple test scenarios include at least one of the following:

[0074] 1) At a first moment, the risk vehicle is in the lane where the target vehicle is located, the risk vehicle is in front of the target vehicle, and there is a risk obstacle within a first range in front of the risk vehicle. And at a second moment, the risk vehicle changes to an adjacent lane of the lane where the target vehicle is located at the first moment. The risk obstacle is an obstacle that has a collision risk with the target vehicle at the second moment. It can be understood that at this time, the first test scenario corresponds to the scenario where the vehicle in front cuts out.

[0075] Exemplarily, the risk obstacle may be a stationary obstacle, or it may also be a moving obstacle. For example, the risk obstacle may be a stationary vehicle or a vehicle with a slow driving speed; for another example, the risk obstacle may also be an obstacle such as a roadblock or a pedestrian.

[0076] After the risk vehicle cuts out of the lane where the target vehicle is currently located, if the target vehicle does not take braking and / or steering measures, it may collide with the risk obstacle.

[0077] The first range may be within 100 meters in front of the target vehicle, or it may also be within 50 meters in front of the target vehicle, or it may also be other ranges. For example, the first range may also be determined according to the speed of the target vehicle, and the first range may decrease as the speed of the target vehicle increases.

[0078] 2) The risk vehicle is located in the lane where the target vehicle is located at the first moment, the risk vehicle is in front of the target vehicle and within the second range of the target vehicle, and the speed of the risk vehicle at the first moment is greater than the speed of the risk vehicle at the second moment. It can be understood that at this time, the first test scenario corresponds to the scenario of the vehicle in front braking.

[0079] Exemplarily, the second range may be within 50 meters in front of the target vehicle, or it may also be within 30 meters in front of the target vehicle, or it may also be other ranges. For example, the second range may also be determined according to the speed of the target vehicle, and the second range may decrease as the speed of the target vehicle increases.

[0080] 3) The risk vehicle is located in the adjacent lane of the lane where the target vehicle is located at the first moment, and the risk vehicle is located in the lane where the target vehicle is located at the first moment and in front of the target vehicle at the second moment. Wherein, the second moment is the moment after the first moment. It can be understood that at this time, the first test scenario corresponds to the scenario of the adjacent vehicle cutting in.

[0081] In some implementation manners, during the process of the risk vehicle cutting in, it may be due to the target vehicle's failure to avoid danger in time that the risk vehicle collides with the target vehicle. That is, a collision occurs between the risk vehicle and the target vehicle between the first moment and the second moment, resulting in the risk vehicle not being located in the lane where the target vehicle is located at the first moment at the second moment.

[0082] It should be noted that in the above three scenarios, the target vehicle may be in the lane where it is located at the first moment at the second moment, or it may also change to other lanes in response to the driver's operation. The present application does not make specific limitations on this.

[0083] In some implementation manners, each of the above test scenarios includes at least one vehicle other than the target vehicle and the risk vehicle, and the at least one vehicle is located within the third range of the target vehicle.

[0084] Exemplarily, the third range may be a range that may cause a collision risk to the target vehicle. For example, the third range may include Figure 2 the range from position 1 to position 5 shown; or the third range may also be a range within a radius of 50 meters centered on the target vehicle; or, the third range may also be other ranges.

[0085] S420. Calibrate the driver decision-making model for each test scenario according to the relative position between the risk vehicle and the target vehicle and the driving behavior information in each test scenario.

[0086] Among them, the risk vehicle is a vehicle that causes a collision risk to the target vehicle, such as the vehicle in front in the scenario of the vehicle in front braking, the adjacent vehicle in the scenario of the adjacent vehicle cutting in, or the vehicle in front in the scenario of the vehicle in front cutting out. The driver decision-making model indicates the risk avoidance ability of the driver in each test scenario.

[0087] In some implementation manners, the multiple test scenarios include a first test scenario and a second test scenario. The first risk vehicle in the first test scenario is located at the first azimuth of the target vehicle, and the second risk vehicle in the second test scenario is located at the second azimuth of the target vehicle, and the first azimuth is different from the second azimuth; calibrating the driver decision-making model for each test scenario includes: calibrating the first driver decision-making model according to the relative position between the first risk vehicle and the target vehicle and the first driving behavior information; calibrating the second driver decision-making model according to the relative position between the second risk vehicle and the target vehicle and the second driving behavior information; where the first driving behavior information is the information generated in the first test scenario, and the second driving behavior information is the information generated in the second test scenario.

[0088] Exemplarily, the first test scenario may be the above-mentioned scenario of the vehicle in front braking, and the second test scenario may be the above-mentioned scenario of the vehicle cutting in; or, the first test scenario may be the above-mentioned scenario of the car cutting in, and the second test scenario may be the above-mentioned scenario of the large vehicle cutting in; or, it may also be other forms.

[0089] Exemplarily, the first test scenario may be the first test scenario in the above-mentioned embodiment.

[0090] In some implementation manners, the driving behavior information includes a plurality of steering wheel information. Each steering wheel information in the plurality of steering wheel information indicates a steering behavior feature in the first test scenario, and the plurality of steering wheel information corresponds to the situation of successful risk avoidance; the above-mentioned calibration of the driver decision-making model includes: determining at least one of the following steering parameters according to the relative position between the first risk vehicle and the target vehicle and each steering wheel information: decision response duration, maximum steering wheel angle, or steering wheel return duration; calibrating the driver decision-making model according to at least one steering parameter.

[0091] In some other implementation manners, the driving behavior information includes a plurality of pedal information, each pedal information in the plurality of pedal information indicates a braking behavior feature in a first test scenario, and the plurality of pedal information corresponds to the situation of successful risk avoidance; the above-mentioned calibrated driver decision-making model includes: determining at least one of the following braking parameters according to the relative position between the first risk vehicle and the target vehicle and each pedal information: decision response duration, maximum braking deceleration, braking effectiveness improvement duration, or braking effectiveness improvement rate; calibrating the driver decision-making model according to at least one braking parameter.

[0092] Exemplarily, the plurality of steering wheel information and / or the plurality of pedal information may include data other than the data excluding the situation of collision in the driving behavior data. A steering behavior feature can be understood as the feature of the target vehicle when a driver controls the target vehicle to steer in a first test scenario, and this steering behavior feature can be characterized by the change of the steering wheel angle over time. A braking behavior feature can be understood as the feature of the target vehicle when a driver controls the target vehicle to brake in a first test scenario, and this braking behavior feature can be characterized by the change of the brake pedal opening and / or the braking deceleration over time.

[0093] In some implementation manners, the above-mentioned calibrated driver decision-making model includes: calibrating the driver decision-making model according to the relative position between the risk vehicle and the target vehicle, the vehicle type of the risk vehicle, and the driving behavior information. More specifically, in the scenario of a neighboring vehicle cutting in, calibrate the driver decision-making model according to the relative position between the risk vehicle and the target vehicle, the vehicle type of the risk vehicle, and the driving behavior information.

[0094] Exemplarily, the following takes the steering risk avoidance data with each steering wheel information being the change of the steering wheel angle over time and the braking risk avoidance data with each pedal information being the change of the braking deceleration over time as examples to respectively illustrate the implementation manners of determining the steering parameters and the braking parameters.

[0095] The method for determining the above-mentioned steering parameters according to the steering risk avoidance data may include S421 and S422 as shown in Figure 5 the following.

[0096] S421, determining the data corresponding to the risk occurrence moment according to the steering risk avoidance data.

[0097] Exemplarily, the corresponding risk occurrence moments (hereinafter denoted as t0) in different test scenarios are different, and specifically, they may be as follows:

[0098] 1) Adjacent vehicle cutting in: The moment when the centroid of the risk vehicle deviates from the center line of the original driving lane by a preset distance is regarded as the moment when the risk occurs. For example, in the scenario of a car cutting in, the preset distance can be 0.375 meters; in the scenario of a large vehicle cutting in, the preset distance can be 0 meters, or the above preset distance can also be other values.

[0099] 2) Leading vehicle braking: The moment when the braking deceleration of the risk vehicle reaches a preset value is regarded as the moment when the risk occurs. Exemplarily, the preset value can be -8m / s 2 , or it can also be -5m / s 2 , or it can also be other values.

[0100] 3) Leading vehicle cutting out: The moment when the risk vehicle cuts out until the lateral quarter body of the leading vehicle of the risk vehicle can be observed is regarded as the moment when the risk occurs.

[0101] Furthermore, according to the above moment when the risk occurs, determine the number of frames of the data corresponding to the moment when the risk occurs in the steering and evasive data.

[0102] In some implementation manners, before executing S421, perform smoothing and filtering processing on the steering and evasive data to eliminate singular point data.

[0103] S422, determine at least one of the following according to the data corresponding to the moment when the risk occurs: decision response duration, maximum steering wheel angle, steering wheel return duration.

[0104] Exemplarily, select the data of the first 500 frames and the last 500 frames before and after the moment when the risk occurs as the preselected data, and select the 10 frames of data with the largest steering angle after the moment when the risk occurs from the preselected data to determine the maximum steering wheel angle. For example, the average value of the steering wheel angles corresponding to the 10 frames of data can be used as the maximum steering wheel angle.

[0105] Record the moment when the steering wheel angle reaches 20% of the maximum steering wheel angle during the process of increasing the steering wheel angle as t 20 , and record the moment when the steering wheel angle reaches 15% of the maximum steering wheel angle during the process of increasing the steering wheel angle as t 15 , then the moment when the steering wheel starts to rotate (or the moment when steering starts) t1 can be calculated as: t1 = t 15 - 3×(t 20 - t 15 ). Furthermore, the decision response time is the difference between t1 and t0.

[0106] Record the moment when the steering wheel angle reaches 20% of the maximum steering wheel angle during the process of gradually decreasing the steering wheel angle from the maximum as t 20’ , and record the moment when the steering wheel angle reaches 15% of the maximum steering wheel angle during the process of increasing the steering wheel angle as t 15’, then the moment t2 when the steering wheel returns to the straight position can be calculated as: t2 = t 20’ + 3×(t 15’ - t 20’ ). Further, the duration when the steering wheel returns to the straight position is the difference between t2 and t1.

[0107] After determining the steering parameters corresponding to each steering risk avoidance data, the average maximum steering wheel angle and the average duration when the steering wheel returns to the straight position can be obtained by taking the average of the steering parameters corresponding to all steering risk avoidance data. Based on the above average maximum steering wheel angle and average duration when the steering wheel returns to the straight position, the following driver decision-making model can be calibrated:

[0108]

[0109] Among them, δ(t) represents the change of the steering wheel angle with time during risk avoidance, and δ max is the maximum steering wheel angle. This driver decision-making model is as Figure 6 shown.

[0110] The method for determining the above braking parameters based on the braking risk avoidance data may include S423 and S424 as Figure 7 shown.

[0111] S423, determining the data corresponding to the risk occurrence moment based on the braking risk avoidance data.

[0112] Exemplarily, the braking avoidance data can be as Figure 8 shown. The method for determining the data corresponding to the risk occurrence moment based on the braking risk avoidance data can refer to the description in S421 above and will not be elaborated here.

[0113] S424, determining at least one of the following based on the data corresponding to the risk occurrence moment: decision response duration, maximum braking deceleration, braking effectiveness improvement duration, braking effectiveness improvement rate.

[0114] Exemplarily, select the data of the first 500 frames and the last 500 frames before and after the risk occurrence moment as preselected data, and select 10 frames of data with the maximum braking deceleration after the risk occurrence moment (hereinafter denoted as T0) from the preselected data to determine the maximum braking deceleration. For example, the average value of the braking decelerations corresponding to the 10 frames of data can be used as the maximum braking deceleration.

[0115] Denote the moment when the braking deceleration reaches 10% of the maximum braking deceleration during the process of increasing the braking deceleration as T 10 , and denote the moment when the braking deceleration reaches 15% of the maximum braking deceleration during the process of increasing the braking deceleration as T 15 , then the starting braking moment T1 can be calculated as: T1 = T 10 - 2×(T 15 - T 10)。

[0116] During the process of increasing the braking deceleration, the moment when the braking deceleration reaches 90% of the maximum braking deceleration is denoted as T 90 , and during the process of increasing the braking deceleration, the moment when the braking deceleration reaches 85% of the maximum braking deceleration is denoted as T 85 , then the moment when the braking deceleration reaches the maximum value (or the moment of braking saturation) T2 can be calculated as: T2 = T 90 +2×(T 90 -T 85 ). Further, the braking effectiveness improvement duration is the difference between T2 and T1.

[0117] The braking effectiveness improvement rate j can be calculated according to the following formula:

[0118]

[0119] where d max is the maximum braking deceleration.

[0120] Exemplarily, the decision response time can be the difference between T1 and T0. Or, the decision response time can also be the difference between the moment when the braking deceleration reaches 5% of the maximum braking deceleration during the process of increasing the braking deceleration and T0.

[0121] After determining the braking parameters corresponding to each braking avoidance data, the average value of each braking parameter can be obtained by taking the average of the braking parameters corresponding to all braking avoidance data, and used to calibrate the driver decision model.

[0122] It should be noted that the above S421 - S422 and S423 - S424 can be executed synchronously or sequentially, and the embodiments of the present application do not make specific limitations in this regard.

[0123] The model calibration method provided by the embodiments of the present application helps to improve the accuracy of the driver decision model, makes the driver decision model more in line with the decision-making ability of human drivers, and helps to improve the safety of the autonomous driving algorithm.

[0124] Figure 9 shows a schematic flowchart of an algorithm test method provided by the embodiments of the present application. This method can be performed by Figure 1 the algorithm test device 200 shown, and this method can include S910 and S920.

[0125] S910, determining the first avoidance success rate of the decision-making algorithm to be tested in the test scenarios indicated by the test scenario set.

[0126] Exemplarily, the set of test scenarios may include multiple test scenarios. The first risk avoidance success rate may be the ratio of the number of successful risk avoidances of the decision algorithm to be tested in the test scenarios indicated by the set of test scenarios to the total number of test scenarios included in the set of test scenarios.

[0127] S920. Determine whether the risk avoidance ability of the decision algorithm to be tested meets the standard according to the second risk avoidance success rate and the first risk avoidance success rate. The second risk avoidance success rate is the risk avoidance success rate determined by running the driver decision model in the test scenarios indicated by the set of test scenarios. The driver decision model is calibrated based on the driving behavior information of the target vehicle and the relative positions between the risk vehicle and the target vehicle in each of the multiple test scenarios. The driving behavior information includes the information generated in response to the first operation in each of the multiple test scenarios. The risk vehicles in the multiple test scenarios are located in different orientations of the target vehicle.

[0128] Exemplarily, the second risk avoidance success rate may be the ratio of the number of successful risk avoidances of the driver decision model in the test scenarios indicated by the set of test scenarios to the total number of test scenarios included in the set of test scenarios. When the second risk avoidance success rate is greater than the first risk avoidance success rate, it is determined that the risk avoidance ability of the decision algorithm to be tested does not meet the standard. When the second risk avoidance success rate is less than or equal to the first risk avoidance success rate, it is determined that the risk avoidance ability of the decision algorithm to be tested meets the standard.

[0129] It should be noted that the set of test scenarios may or may not include the first test scenario.

[0130] For a more detailed calibration method of the driver decision model, reference can be made to the description in Method 400, which will not be elaborated here.

[0131] The algorithm testing method provided by the embodiments of the present application conducts autonomous driving algorithm testing according to the driver decision model calibrated in Method 400, which helps to improve the accuracy of the test results, enabling algorithm developers to optimize based on the test results and contributing to the improvement of the safety and reliability of the autonomous driving algorithm.

[0132] Figure 10 It is a schematic block diagram of a device provided by the embodiments of the present application. The device 1000 can be used to execute Figure 2 , Figure 4 , Figure 5 , Figure 7 , Figure 9 any of the methods shown in Figure 10The illustrated apparatus 1000 may include: a processor 1010, a transceiver 1020, and a memory 1030. Among them, the processor 1010, the transceiver 1020, and the memory 1030 are connected through an internal connection path. The memory 1030 is used to store instructions, and the processor 1010 is used to execute the instructions stored in the memory 1030 to implement the methods in the above embodiments. Optionally, the memory 1030 can be coupled to the processor 1010 through an interface or integrated with the processor 1010.

[0133] It should be noted that the above transceiver 1020 may include, but is not limited to, a transceiver device such as an input / output interface to implement communication between the apparatus 1000 and other devices or communication networks.

[0134] The memory 1030 may be a volatile memory and / or a non-volatile memory. Among them, the non-volatile memory may be a read-only memory (ROM), a programmable ROM (PROM), an erasable PROM (EPROM), an electrically erasable PROM (EEPROM), or a flash memory. The volatile memory may be a random access memory (RAM). For example, the RAM can be used as an external cache. By way of example and not limitation, the RAM includes the following various forms: static random access memory (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDR SDRAM), enhanced SDRAM (ESDRAM), synchlink DRAM (SLDRAM), and direct rambus RAM (DR RAM).

[0135] The transceiver 1020 uses a transceiver device such as, but not limited to, a transceiver to implement communication between the apparatus 1000 and other devices or communication networks to receive / transmit data / information for implementing the methods in the above embodiments.

[0136] In the embodiments of the present application, a processor is a circuit with signal processing capabilities. In one implementation, the processor can implement certain functions through the logical relationship of a hardware circuit, and the logical relationship of the hardware circuit is fixed or can be reconfigured. For example, the processor is a hardware circuit implemented by an application-specific integrated circuit (ASIC) or a programmable logic device (PLD), such as a field programmable gate array (FPGA). In a reconfigurable hardware circuit, the process of the processor loading a configuration document to implement the configuration of the hardware circuit can be understood as the process of the processor loading instructions to implement the functions of some or all of the above units. In addition, it can also be a hardware circuit designed for artificial intelligence, which can be understood as a type of ASIC, such as a neural network processing unit (NPU), a tensor processing unit (TPU), a deep learning processing unit (DPU), etc.

[0137] In various embodiments of the present application, if there is no special description and logical conflict, the terms and / or descriptions among the various embodiments are consistent and can be referenced to each other. The technical features in different embodiments can be combined to form new embodiments according to their inherent logical relationships.

[0138] The embodiments of the present application further provide an electronic device, which includes device 1000; or includes model calibration device 100; or includes algorithm testing device 200.

[0139] The embodiments of the present application further provide a computer program product, which includes computer program code. When the computer program code runs on a computer, the computer is enabled to implement the methods in the above various embodiments of the present application.

[0140] The embodiments of the present application further provide a computer-readable storage medium, which stores computer instructions. When the computer instructions run on a computer, the computer is enabled to implement the methods in the above various embodiments of the present application.

[0141] The embodiments of the present application further provide a chip, which includes a circuit for executing the methods in the above various embodiments of the present application.

[0142] Those skilled in the art can clearly understand that for the convenience and brevity of description, the specific working processes of the systems, devices, and units described above can refer to the corresponding processes in the foregoing method embodiments and will not be elaborated herein.

[0143] In the description of the embodiments of the present application, unless otherwise specified, " / " means "or". For example, A / B can mean A or B. Herein, "and / or" is an associative relationship describing associated objects, indicating that three relationships can exist. For example, A and / or B can mean: A exists alone, A and B exist simultaneously, and B exists alone. In the present application, "at least one" means one or more, and "a plurality" means two or more. "At least one (item)" or similar expressions thereof refer to any combination of these items, including any combination of single items or plural items. For example, at least one (item) of a, b, or c can mean: a, b, c, a - b, a - c, b - c, or a - b - c, where a, b, c can be single or multiple.

[0144] In the embodiments of the present application, prefix words such as "first" and "second" are only used to distinguish different described objects and have no restrictive effect on the position, order, priority, quantity, or content of the described objects. In the embodiments of the present application, the use of ordinal numbers and other prefix words for distinguishing described objects does not constitute a restriction on the described objects. For the statement of the described objects, refer to the description in the claims or the context of the embodiments, and no redundant restriction should be formed due to the use of such prefix words.

[0145] In several embodiments provided by the present application, it should be understood that the disclosed systems, devices, and methods can be implemented in other ways. For example, the device embodiments described above are merely illustrative. For example, the division of the units is only a logical function division, and there can be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the displayed or discussed coupling or direct coupling or communication connection to each other can be through some interfaces, and the indirect coupling or communication connection of the devices or units can be in electrical, mechanical, or other forms.

[0146] In each embodiment of the present application, if there is no special description and logical conflict, the terms and / or descriptions among the embodiments are consistent and can be referred to each other, and the technical features in different embodiments can be combined to form new embodiments according to their internal logical relationships.

[0147] The unit described as a separation component may or may not be physically separated. The component shown as a unit may or may not be a physical unit, that is, it may be located in one place or may be distributed across multiple network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0148] In addition, each functional unit in various embodiments of the present application may be integrated in a processing unit, may exist separately as individual physical units, or two or more units may be integrated in one unit.

[0149] As described above, the above are only specific embodiments of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art within the technical scope disclosed in the present application can easily think of changes or substitutions, which should be covered by the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claimed rights.

Claims

1. A model calibration method, characterized in that, Including: Obtaining driving behavior information of a target vehicle, where the driving behavior information includes information generated in each of a plurality of test scenarios in response to a first operation, and risk vehicles in the plurality of test scenarios are located at different orientations of the target vehicle; Calibrating a driver decision model for each of the test scenarios according to the relative position between the risk vehicle and the target vehicle in each of the test scenarios and the driving behavior information, where the driver decision model indicates the risk avoidance ability of a driver in each of the test scenarios.

2. The method according to claim 1, wherein The plurality of test scenarios include a first test scenario and a second test scenario. A first risk vehicle in the first test scenario is located at a first orientation of the target vehicle, and a second risk vehicle in the second test scenario is located at a second orientation of the target vehicle, and the first orientation is different from the second orientation; The calibrating the driver decision model for each of the test scenarios includes: Calibrating a first driver decision model according to the relative position between the first risk vehicle and the target vehicle and first driving behavior information; Calibrating a second driver decision model according to the relative position between the second risk vehicle and the target vehicle and second driving behavior information; where the first driving behavior information is information generated in the first test scenario, and the second driving behavior information is information generated in the second test scenario.

3. The method according to claim 2, wherein The driving behavior information includes a plurality of steering wheel information, and each of the plurality of steering wheel information indicates a steering behavior feature in the first test scenario, and the plurality of steering wheel information corresponds to a situation of successful risk avoidance; The calibrating the first driver decision model includes: Determining at least one of the following steering parameters according to the relative position between the first risk vehicle and the target vehicle and each of the steering wheel information: decision response duration, maximum steering wheel angle, or steering wheel return duration; Calibrating the first driver decision model according to the at least one steering parameter.

4. The method according to claim 2 or 3, characterized in that, The driving behavior information includes a plurality of pedal information, and each of the plurality of pedal information indicates a braking behavior feature in the first test scenario, and the plurality of pedal information corresponds to a situation of successful risk avoidance; The calibrating the first driver decision model includes: Determining at least one of the following braking parameters according to the relative position between the first risk vehicle and the target vehicle and each of the pedal information: decision response duration, maximum braking deceleration, braking efficiency improvement duration, or braking efficiency improvement rate; Calibrating the first driver decision model according to the at least one braking parameter.

5. The method according to any one of claims 1 to 4, characterized in that, The plurality of test scenarios include at least one of the following: The risk vehicle is located in the lane where the target vehicle is located at a first moment, the risk vehicle is in front of the target vehicle and there is a risk obstacle within a first range in front of the risk vehicle, and the risk vehicle changes to an adjacent lane of the lane where the target vehicle is located at the first moment at a second moment, and the risk obstacle is an obstacle that has a collision risk with the target vehicle at the second moment; The risk vehicle is located in the lane where the target vehicle is located at the first moment, the risk vehicle is in front of the target vehicle and within a second range of the target vehicle, and the speed of the risk vehicle at the first moment is greater than the speed of the risk vehicle at the second moment; or The risk vehicle is located in an adjacent lane to the lane where the target vehicle is located at the first moment, and the risk vehicle is located in the lane where the target vehicle is located at the first moment and in front of the target vehicle at the second moment; wherein, the second moment is a moment after the first moment.

6. The method according to claim 5, wherein Calibrating the driver decision model in each of the test scenarios includes: When the risk vehicle is located in an adjacent lane to the lane where the target vehicle is located, calibrate the driver decision model according to the relative position between the risk vehicle and the target vehicle, the vehicle type of the risk vehicle, and the driving behavior information.

7. The method according to any one of claims 1 to 6, characterized in that, Each of the test scenarios includes at least one vehicle other than the target vehicle and the risk vehicle, and the at least one vehicle is within a third range of the target vehicle.

8. An algorithm testing method, characterized in that, Including: Determine the first hazard avoidance success rate of the decision algorithm to be tested in the test scenarios indicated by the test scenario set; Determine whether the hazard avoidance ability of the decision algorithm to be tested meets the standard according to the second hazard avoidance success rate and the first hazard avoidance success rate; wherein, the second hazard avoidance success rate is the hazard avoidance success rate determined by running the driver decision model in the test scenarios indicated by the test scenario set, the driver decision model is calibrated according to the driving behavior information of the target vehicle and the relative position between the risk vehicle and the target vehicle in each of the multiple test scenarios, the driving behavior information includes the information generated in response to the first operation in each of the multiple test scenarios, and the risk vehicles in the multiple test scenarios are located in different orientations of the target vehicle.

9. The method according to claim 8, characterized in that, The driving behavior information includes multiple steering wheel information in each of the test scenarios, each steering wheel information in the multiple steering wheel information indicates a steering behavior feature, and the multiple steering wheel information corresponds to the situation of successful hazard avoidance; and, the driver decision model in each of the test scenarios is calibrated according to at least one steering parameter, the at least one steering parameter is determined according to the relative position between the risk vehicle and the target vehicle and each of the steering wheel information, and the at least one steering parameter includes at least one of a decision response duration, a maximum steering wheel angle, or a steering wheel return duration.

10. The method according to claim 8 or 9, characterized in that, The driving behavior information includes multiple pedal information in each of the test scenarios, each pedal information in the multiple pedal information indicates a braking behavior feature, and the multiple pedal information corresponds to the situation of successful hazard avoidance; Moreover, the driver decision-making model under each test scenario is calibrated based on at least one braking parameter, and the at least one braking parameter is determined according to the relative position between the risk vehicle and the target vehicle and each pedal information, and the at least one braking parameter includes at least one of decision response duration, maximum braking deceleration, braking efficiency improvement duration, or braking efficiency improvement rate.

11. The method according to any one of claims 8 to 10, characterized in that The driver decision-making model under each test scenario is calibrated according to the relative position between the risk vehicle and the target vehicle, the vehicle type of the risk vehicle, and the driving behavior information.

12. The method according to any one of claims 8 to 11, characterized in that, The multiple test scenarios include at least one of the following: At the first moment, the risk vehicle is in the lane where the target vehicle is located, the risk vehicle is in front of the target vehicle and there is a risk obstacle within the first range in front of the risk vehicle, and at the second moment, the risk vehicle changes to the adjacent lane of the lane where the target vehicle is located at the first moment, and the risk obstacle is an obstacle that has a collision risk with the target vehicle at the second moment; At the first moment, the risk vehicle is in the lane where the target vehicle is located, the risk vehicle is in front of the target vehicle and within the second range of the target vehicle, and the speed of the risk vehicle at the first moment is greater than the speed of the risk vehicle at the second moment; Or At the first moment, the risk vehicle is in the adjacent lane of the lane where the target vehicle is located, and at the second moment, the risk vehicle is in the lane where the target vehicle is located at the first moment and the risk vehicle is in front of the target vehicle; Wherein, the second moment is a moment after the first moment.

13. The method according to any one of claims 8 to 12, characterized in that Each test scenario includes at least one vehicle other than the target vehicle and the risk vehicle, and the at least one vehicle is within the third range of the target vehicle.

14. A model calibration device, characterized in that, Includes a module for executing the method according to any one of claims 1 to 7.

15. A model calibration device, characterized in that, Includes: A memory for storing a computer program; A processor for executing the computer program stored in the memory, so that the device executes the method according to any one of claims 1 to 7.

16. An algorithm testing device, characterized in that, Includes a module for executing the method according to any one of claims 8 to 13.

17. An algorithm testing device, characterized in that, Includes: A memory for storing a computer program; A processor for executing the computer program stored in the memory, so that the device executes the method according to any one of claims 8 to 13.

18. A computer-readable storage medium, characterized in that, Instructions are stored thereon, and when the instructions are executed by a processor, the method according to any one of claims 1 to 7 is implemented; or, the method according to any one of claims 8 to 13 is implemented.

19. A computer program product, characterized in that, The computer program product includes: computer program code, and when the computer program code is run, the method according to any one of claims 1 to 7 is implemented, or the method according to any one of claims 8 to 13 is implemented.

20. A chip, characterized in that, The chip includes a circuit for executing the method according to any one of claims 1 to 7; or executing the method according to any one of claims 8 to 13.

Images