Multi-driving-style high-risk automatic driving cut-in scene test method and system

Abstract

The invention relates to a multi-driving-style high-risk automatic driving cut-in scene test method and system, and the method comprises the steps: collecting original driving track data, carrying out the clustering of driving styles, and generating a risk cut-in scene track cluster with consistent driving styles through employing a Cutin-TimeGAN network model in combination with physical feasibility constraints; constructing a cost function, calculating a cost function value for each risk cut-in scene trajectory cluster, and selecting a target cut-in reference trajectory as a target reference state vector; constructing a game framework, and combining a target reference state vector to construct a utility function according to the state vectors of the VUT and the test confrontation vehicle in the framework; constructing a risk confrontation utility function based on the predicted collision time; a game optimization problem is constructed and solved, an optimal interaction track of the test confrontation vehicle is obtained, and a closed-loop high-risk automatic driving scene switching test is carried out based on the optimal interaction track; the system is used for implementing the method. Compared with the prior art, the method has the advantages that testing of authenticity and high-risk confrontation of multiple driving styles is considered.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of automatic driving car testing, in particular to a multi-driving style high-risk automatic driving cut-in scene testing method and system. BACKGROUND

[0002] When an automatic driving car faces a risk cut-in scene, it is often in a critical interaction state of strong coupling and high uncertainty, which is easy to trigger a collision or a near-miss accident. The scene-based testing method is an important testing method before the automatic driving car is put into use. The existing scene-based testing method mainly includes: a testing method based on natural driving data playback, which directly uses natural driving data for playback, although it is real, but lacks controllability and high-risk conditions; a testing method based on parameterized script predefinition, which generates scenes by setting rules and parameters, although it is controllable, but lacks real driving style and diversity. At the same time, many high-risk scene generation methods often cannot simultaneously consider real driving style and adversarial enhancement, resulting in insufficient test coverage, especially in critical interaction scenes under different driving styles. In order to solve the above technical problems, Chinese patent application CN120278193A provides an automatic driving risk lane change testing scene generation method, which first generates a risk lane change trajectory of a background vehicle BV using an improved Traj-TimeGAN; then constructs a critical safety distance model, under the premise that the AV takes a given braking behavior, the initial state of the AV is inversely calculated through an analytical formula, so that the two vehicles are in a critical state of just not colliding at a certain time C; then each generated BV trajectory and the AV initial state obtained by inverse calculation are directly used as a critical lane change test case, and then the lane change test scene is generalized; although the testing accuracy of high-risk critical lane change is realized, it still has the following shortcomings: 1) only open-loop critical scenes are generated, there is no closed-loop interaction test with the tested vehicle strategy, and the generated scene forms are basically the same for different tested vehicles, and the most dangerous trajectory cannot be adaptively generated for the specific control strategy of the tested vehicle; 2) each generated trajectory is regarded as an equivalent candidate scene, ignoring the balance between different driving style types of trajectory clusters, and lacking trajectory cluster level controllable optimization.

[0003] Therefore, a high-style cut-in scene testing method is provided, which can realize closed-loop interaction with the tested vehicle strategy, adaptively generate the most dangerous trajectory for the specific control strategy of the tested vehicle, and simultaneously consider the balance of different driving style trajectory clusters and support trajectory cluster level controllable optimization. SUMMARY

[0004] The purpose of the present application is to overcome the defects of the prior art and provide a multi-driving style high-risk automatic driving cut-in scene test method and system, which learns the cut-in trajectory distribution of multi-driving style from natural driving data, generates candidate trajectories with consistent and diversified styles, and uses them as style references to build a high-risk, controllable and repeatable cut-in interaction test scene in dynamic game, thereby comprehensively testing the safety of a vehicle under test (VUT) in a risk cut-in scene.

[0005] The purpose of the present application can be achieved by the following technical solutions: According to a first aspect of the present application, a multi-driving style high-risk automatic driving cut-in scene test method is provided, comprising: Collecting original driving trajectory data and building physical feasibility constraints, clustering the original driving trajectory data by driving style, and using a Cutin-TimeGAN network model combined with the physical feasibility constraints to generate a risk cut-in scene trajectory cluster with consistent driving style; the driving style includes conservative, normal and aggressive; Building a cost function, calculating the cost function value for each risk cut-in scene trajectory cluster, and selecting a target cut-in reference trajectory as a target reference state vector based on the cost function value; Building a Stackelberg game framework, constructing an utility function based on the state vectors of the VUT and the test opponent vehicle in the framework, and combining the target reference state vector; Calculating the predicted collision time of the VUT and the test opponent vehicle, and building a risk confrontation utility function based on the predicted collision time; Based on the utility function and the risk confrontation utility function, a risk cut-in interaction Stackelberg game optimization problem is constructed and solved to obtain the optimal interaction trajectory of the test opponent vehicle, and a closed-loop high-risk automatic driving cut-in scene test is performed based on the optimal interaction trajectory.

[0006] As a preferred technical solution, the physical feasibility constraints include: For each vehicle, its state always moves forward, i.e. the vehicle's longitudinal displacement at the previous time is less than the vehicle's longitudinal displacement at the next time; For a vehicle after the end of the scene cut-in, the vehicle's lateral displacement point at the end time is located within the left and right boundaries of the lane; For each vehicle, the difference in lateral displacement between adjacent time frames does not exceed a difference threshold.

[0007] As a preferred technical solution, the loss function of the Cutin-TimeGAN network model includes an embedder reconstruction loss, a discriminator loss and a generator adversarial loss. The embedder reconstruction loss is: , represents a calculation mathematical expectation operation, represents input driving trajectory data, represents a driving trajectory reconstructed by the embedder; The discriminator loss is: , represents a discrimination result output by the discriminator based on the latent vector H, represents a cross-entropy loss calculation operation, represents a discrimination result output by the discriminator based on the generated latent representation , represents a weight coefficient, represents a discrimination result output by the discriminator based on the representation generated by random noise , The generator adversarial loss is: , represents an adversarial loss function, and ; represents a supervision loss, and , represents a real latent representation at t+1 time, represents a prediction result at t+1 time; represents a statistical matching loss, and , represents a mean calculation operation, represents trajectory data generated by the Cutin-TimeGAN network model; represents a standard deviation calculation operation.

[0008] As a preferred technical solution, for the ith trajectory, the corresponding cost function is: , wherein, , and all represent style weight coefficients; represents a total planning step length; represents a unit time length; represents a vehicle longitudinal speed at t time, and represents a vehicle longitudinal acceleration at t time; represents a jth planning step time; represents a reference vehicle speed; represents a vehicle lateral acceleration at t time.

[0009] ​​​As a preferred technical solution, in the Stackelberg game framework, the VUT is set as a follower F, and the test opponent vehicle is set as a leader L, and the state vectors of the two are in the form of: , , Wherein, represents the state vector of the follower F at the kth moment; represents the state vector of the leader L at the kth moment; and respectively represent the longitudinal position of the follower and the leader at the kth moment; and respectively represent the longitudinal speed of the follower and the leader at the kth moment; and respectively represent the lateral position of the follower and the leader at the kth moment; and respectively represent the lateral speed of the follower and the leader at the kth moment; The state weight coefficient matrix is constructed as: wherein, represents the weight proportion coefficient of different states; The control input weight matrix is constructed as: wherein, represents the weight proportion coefficient of different control inputs; and the control input vector is set as: , represents the longitudinal control input of the vehicle at the kth moment, represents the lateral control input of the vehicle at the kth moment; The state transition of the follower and the leader follows: , , wherein, represents the control input vector of the follower at the kth moment; represents the state transition matrix; represents the input gain matrix; represents the control input vector of the leader at the kth moment.

[0010] As a preferred technical solution, the utility function is: , , wherein, represents the utility function of the follower, i.e. the VUT; ​x k + 1 represents the state vector of the follower at the k+1 th moment; x k + 1 represents the target reference state vector of the follower; x k + 1 represents the state weight coefficient matrix; x k represents the control input vector of the follower at the k th moment; x k represents the control input weight matrix; x k represents the total step length of planning; u k represents the utility function of the leader, i.e., the test confrontation vehicle; x k + 1 represents the state vector of the leader at the k+1 th moment; x k + 1 represents the target reference state vector of the leader; x k represents the control input vector of the leader at the k th moment.

[0011] As a preferred technical solution, the risk confrontation utility function is: , wherein, x k represents the total step length of planning; x k represents the length of the trajectory traveled by the follower; x k represents the speed of the follower; x k represents the length of the trajectory traveled by the leader; x k represents the speed of the leader.

[0012] As a preferred technical solution, the risk cut-in interaction Stackelberg game optimization problem is: , , , wherein, x k represents the control input vector of the leader; x k represents the risk driving adjustment coefficient; x k represents the compliance natural driving adjustment coefficient; u k represents the utility function of the leader, i.e., the test confrontation vehicle; u k represents the risk confrontation utility function; x k represents the control input vector of the follower at the k th moment; x k represents the state transition matrix; x k represents the input gain matrix; x k represents the control input vector of the leader at the k th moment; x k represents the longitudinal position of the leader at the k th moment; x k represents the longitudinal position of the follower at the k th moment; x k represents the preset safety distance; x k represents the control input vector of the i th vehicle; x k represents the lower bound of the control input vector; This indicates the upper bound of the control input vector; Indicates the speed of the i-th vehicle; Indicates the maximum vehicle speed; This represents the state vector of the follower at time k. This represents the leader's state vector at time k.

[0013] As a preferred technical solution, the method for solving the aforementioned risk-entry interaction Stackelberg game optimization problem is as follows: The aforementioned risk-based Stackelberg game optimization problem is transformed into a leader-optimal control problem and a follower-optimal control problem, namely: ,in, Represents the leader's state vector; This represents the leader's control input vector; This represents the follower's optimal control input vector; This represents the follower control input vector; Represents the follower state vector; Based on its convex optimization property, the follower optimal control problem is transformed into a single-layer convex optimization problem by introducing KKT conditions, which are: , , , , in, and Both represent the Lagrange coefficients; This indicates the risk driving adjustment factor; Indicates the compliant natural driving adjustment coefficient; This represents the utility function of the leader, i.e., the test vehicle. This represents the risk mitigation utility function; This represents the control input vector of the follower at time k. Represents the state transition matrix; Represents the input gain matrix; This represents the leader's control input vector at time k. This represents the leader's vertical position at time k. This represents the vertical position of the follower at time k. Indicates the preset safety distance; This represents the control input vector for the i-th vehicle; Indicates the lower bound of the control input vector; This indicates the upper bound of the control input vector; denotes the speed of the i-th vehicle; denotes the maximum vehicle speed; denotes the Lagrangian function; denotes the state vector of the follower at the k-th time; denotes the state vector of the leader at the k-th time; The single-layer convex optimization problem is solved by using CasADi and Ipopt solvers.

[0014] According to the second aspect of the present application, a multi-driving style high-risk automatic driving cut-in scene test system is provided for implementing the above method.

[0015] Compared with the prior art, the present application has the following beneficial effects: 1) In view of the problem in the prior art that the closed-loop interactive test of the measured vehicle strategy cannot be realized, after obtaining the risk cut-in trajectory cluster with multiple driving styles based on TimeGAN, the present application optimizes a trajectory with the minimum current cost function value as a reference trajectory, and incorporates the reference trajectory into the construction of the game optimization problem, so that the scene generation process becomes a closed-loop game process of VUT and the opponent vehicle; in the closed-loop game process, the risk driving adjustment coefficient and the compliance natural driving adjustment coefficient are introduced, and the above two parameters are dynamically adjusted for each different driving style, so that the generated cut-in trajectory is automatically adjusted according to the control strategy of the measured vehicle, compared with the one-way critical scene in the prior art, the present application can find a risk interaction process that is more consistent with the real behavior of the measured vehicle, significantly improves the test diversity and antagonism, and at the same time, the multiple driving style authenticity and high-risk antagonism are taken into account.

[0016] 2) The present application does not need to manually collect any trajectory, but realizes the automatic generation of different groups of cut-in trajectories to form test scenes through the TimeGAN algorithm, and realizes dynamic interactive testing. BRIEF DESCRIPTION OF DRAWINGS

[0017] Figure 1 is a flowchart of the method of the present application; Figure 2 is a driving trajectory graph and a speed change graph in the risk cut-in scene test process in the embodiment of the present application. DETAILED DESCRIPTION

[0018] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor should belong to the protection scope of the present application.

[0019] To solve the problems in the prior art, the present application provides a multi-driving style high-risk automatic driving cut-in scene test method, the flow of which is as shown in the figure, and in detail comprises the following steps: Figure 1 S1, collect original driving trajectory data and construct physical feasibility constraints, after driving style clustering of the original driving trajectory data, generate driving style consistent risk cut-in scene trajectory clusters by using a Cutin-TimeGAN network model combined with physical feasibility constraints.

[0020] S11, original driving trajectory data collection and processing.

[0021] In this embodiment, the collected original driving trajectory data is derived from the trajectory data of the risk cut-in scene of the automatic driving vehicle on the expressway extracted from the public data set.

[0022] S111, load multi-period vehicle trajectory data set, wherein the vehicle trajectory data includes timestamp, vehicle ID and vehicle driving state information, including position, speed, heading and lane identification; filter the driving state information of each vehicle according to the vehicle ID, and identify the cut-in event according to the lane identification of the ego vehicle and the driving state information of the target vehicle, and filter to obtain complete cut-in trajectory data.

[0023] S112, filter outliers in the cut-in trajectory data, adopt Z-score standardization processing, and obtain the final complete n vehicle cut-in trajectory set , wherein the trajectory i of the i-th vehicle is , T is the vehicle cut-in duration, and the vehicle cut-in state trajectory at time t contains the following state information: longitudinal position , lateral position , longitudinal speed , lateral speed and heading angle , which can be expressed as: .

[0024] S12, driving style clustering.

[0025] In order to identify the driving style of different drivers in the cut-in process, the cut-in duration T is selected to construct the feature vector, wherein for the i-th vehicle, the feature vector is: .

[0026] The feature vectors of the cut-in scene data are clustered by using a density-based clustering algorithm DBSCAN, in this embodiment, the core idea of DBSCAN is to define the feature vector of the i-th vehicle as​ - neighborhood: , where, denotes the feature vector of the jth vehicle; if the number of samples of a point in its neighborhood is not less than the minimum sample number , it is called a core point. The core points are gradually expanded through "density reachable" and "density connected" between the core points to form a cluster set . In the cluster set, the driving style is divided into conservative, ordinary and aggressive according to the trajectory feature data.

[0027] S13, constructing physical feasibility constraints.

[0028] In order to ensure that the generated trajectory meets the actual requirements, the following constraints are constructed in the present application: i) for each vehicle, its state always moves forward, that is, the longitudinal displacement of the vehicle at the previous time is less than that at the next time, and taking the ith vehicle as the column, it can be expressed as: , denotes the longitudinal displacement of the ith vehicle at time t.

[0029] ii) for the vehicle after the end of the scene cut-in, the lateral displacement point of the vehicle at the end time is located within the left and right boundaries of the lane, and taking the ith vehicle as the column, it can be expressed as: , denotes the left boundary constraint of the lane, denotes the lateral displacement point of the ith vehicle at time t, denotes the right boundary constraint of the lane.

[0030] iii) for each vehicle, the difference value of the lateral displacement of adjacent time frames does not exceed the difference value threshold, and taking the ith vehicle as an example, it can be expressed as: .

[0031] S14, risk cut-in scene trajectory cluster generation of multiple driving styles.

[0032] On the basis of the traditional time series generative adversarial network (TimeGAN), a time series generative adversarial network (Cutin-TimeGAN) for high-speed cut-in scene is proposed. The specific improvements include: introducing traffic behavior feature constraints to ensure the physical feasibility of the generated trajectory; introducing prior style to ensure that the generated trajectory is consistent under different driving styles. In the Cutin-TimeGAN network provided in the present application, an embedder E , a generator G , and a restorer R, supervisor S and discriminator D , in detail, have: , embedder E : input the cut-in trajectory sequence data after driving style clustering, encode the trajectory sequence data into low-dimensional latent vector through multi-layer recurrent neural network , reserve the time-dependent feature.

[0033] , restorer R : input latent vector H , decode the latent vector to reconstruct the trajectory through the symmetric network structure , ensure the effectiveness of encoding.

[0034] , generator G : input random noise sequence Z and driving style label , output fake latent vector , simulate the time sequence trajectory of the real risk cut-in scene.

[0035] , supervisor S : input real latent vector H and generated latent vector , output time prediction result .

[0036] , discriminator D : classify and distinguish real and fake latent vectors, drive generator optimization.

[0037] When training Cutin-TimeGAN network, three types of collaborative loss functions, embedder reconstruction loss, discriminator loss and generator adversarial loss, are constructed to ensure the real distribution and time sequence rationality of the generated driving trajectory.

[0038] In order to make the input get through the embedder , and then get the reconstructed trajectory through the restorer R, the reconstruction loss of this process is minimized, and the embedder reconstruction loss is constructed as: , represents the calculation mathematical expectation operation, represents the input driving trajectory data, represents the driving trajectory reconstructed by the embedder.

[0039] The goal of the discriminator is to distinguish between real latent representation , generated latent representation and representation generated only by random noise , and the discriminator loss can be: , represents a discrimination result output by the discriminator based on the latent vector H, represents a cross-entropy loss calculation operation, represents a discrimination result output by the discriminator based on the generated latent representation , represents a weight coefficient (in the embodiment ), represents a discrimination result output by the discriminator based on the representation generated by the random noise .

[0040] The objective of the generator is to make the discriminator believe that , are real, so the generator adversarial loss can be constructed as: , wherein, represents an adversarial loss function, which is used to measure the difference between x and the real label, since the objective of the generator is to "deceive" the discriminator, the label is taken in this loss term, that is, it is hoped that the discriminator believes that the generated sample is a real sample, and ; represents a supervision loss, which aims to constrain the temporal consistency, and requires the generated latent representation to be able to predict the real , , represents a real latent representation at t+1 time, represents a prediction result at t+1 time; represents a statistical matching loss, in order to ensure the consistency of the overall statistical distribution, the matching of the first moment (mean) and the second moment (standard deviation) is introduced, and , represents a mean calculation operation, represents trajectory data generated by the Cutin-TimeGAN network model; represents a standard deviation calculation operation.

[0041] After training the network model by using the above loss function, according to actual requirements, determine the network type, hidden layer dimension, number of layers, training iteration number, batch size and other key parameters, load the driving trajectory data, and use the trained Cutin-TimeGAN network to generate a risk cut-in scene trajectory cluster, by inputting the trajectory data of different driving styles after clustering, it is ensured that the driving style of the network model is consistent in each generation of the risk cut-in scene trajectory cluster.

[0042] For the generated risk cut-in scenario trajectory cluster, the cut-in trajectory satisfying the above physical feasibility constraint is screened as an effective trajectory segment, and the final risk cut-in trajectory cluster meeting the requirements is obtained.

[0043] S2, construct a cost function, calculate the cost function value for each risk cut-in scenario trajectory cluster, and select the target cut-in reference trajectory as the target reference state vector based on the cost function value.

[0044] S21, cost function construction.

[0045] Taking the ith trajectory as an example, the corresponding cost function can be constructed as: , Among them, , and all represent style weight coefficients; represent the total step length; represent the unit time; represent t j the vehicle longitudinal speed at time t, and represent the time at the jth planning step; represent the reference vehicle speed; represent t j the vehicle longitudinal acceleration at time t; represent t j the vehicle lateral acceleration at time t.

[0046] According to the above constructed cost function, the corresponding function value is calculated, and under the premise of meeting the safety, feasibility and style constraints, the trajectory with the minimum cost function value is selected as the target cut-in reference trajectory.

[0047] S22, target reference state vector construction.

[0048] According to the target cut-in reference trajectory data, the state vector at the target time is constructed as: , , and represent the target reference longitudinal position of VUT and the test opponent vehicle respectively; and represent the target reference longitudinal speed of VUT and the test opponent vehicle respectively; and represent the target reference lateral position of VUT and the test opponent vehicle respectively; and respectively represent the target reference lateral velocity of the VUT and the test opposing vehicle.

[0049] S3, constructing a Stackelberg game framework, constructing a utility function according to the state vectors of the VUT and the test opposing vehicle in the framework, and combining the target reference state vector.

[0050] In the Stackelberg game framework, there are: Taking the VUT as a follower F and the test opposing vehicle as a leader L, the state vectors of the two are set as: , , wherein, represents the state vector of the follower F at the kth moment; represents the state vector of the leader L at the kth moment; and respectively represent the longitudinal position of the follower and the leader at the kth moment; and respectively represent the longitudinal velocity of the follower and the leader at the kth moment; and respectively represent the lateral position of the follower and the leader at the kth moment; and respectively represent the lateral velocity of the follower and the leader at the kth moment; the initial state can be set as: , ; The state transition of the follower and the leader follows: , , wherein, represents the control input vector of the follower at the kth moment; represents a state transition matrix; represents an input gain matrix; represents the control input vector of the leader at the kth moment; then in the prediction interval , based on the initial state , the future state trajectory can be obtained through the above state transition as: .

[0051] The state weight coefficient matrix is constructed as: wherein, represents the weight proportion coefficient of different states.

[0052] The control input weight matrix is constructed as: wherein, and denote the weight proportion coefficient of different control inputs; and set the control input vector as: , denotes the longitudinal control input of the vehicle at time k, denotes the lateral control input of the vehicle at time k.

[0053] On the basis of the above architecture, for the time interval , the utility functions of the follower and the leader are respectively designed as: , , wherein, denotes the utility function of the follower, i.e., the VUT; denotes the state vector of the follower at time k+1; denotes the target reference state vector of the follower; denotes the state weight coefficient matrix; denotes the control input vector of the follower at time k; denotes the control input weight matrix; denotes the total planning step; denotes the utility function of the leader, i.e., the test counter vehicle; denotes the state vector of the leader at time k+1; denotes the target reference state vector of the leader; denotes the control input vector of the leader at time k.

[0054] S4, calculate the predicted collision time of the VUT and the test counter vehicle, and construct a risk counter utility function based on the predicted collision time.

[0055] In addition, the leader needs to actively counter the VUT in addition to meeting the driving compliance and driving smoothness, and therefore needs to additionally design a risk counter utility function, which is: , wherein, denotes the total planning step; denotes the length of the trajectory traveled by the follower; denotes the speed of the follower; denotes the length of the trajectory traveled by the leader; denotes the speed of the leader.

[0056] S5, constructing a risk-cutting interaction Stackelberg game optimization problem based on the utility function and the risk-antagonistic utility function and solving the problem to obtain an optimal interaction trajectory of the test-antagonistic vehicle, and performing closed-loop high-risk automatic driving cut-in scene testing based on the optimal interaction trajectory.

[0057] S51, constructing a risk-cutting interaction Stackelberg game optimization problem.

[0058] On the basis of the functions constructed in the foregoing steps, in order to ensure the interaction safety of the follower and the leader, the longitudinal collision constraint needs to be met: In combination with the utility function and the risk-antagonistic utility function, the optimization problem can be obtained as: , , , wherein, represents a control input vector of the leader; represents a risk driving adjustment coefficient; represents a compliance natural driving adjustment coefficient; represents a utility function of the leader, i.e., the test-antagonistic vehicle; represents a risk-antagonistic utility function; represents a control input vector of the follower at the kth moment; represents a state transition matrix; represents an input gain matrix; represents a control input vector of the leader at the kth moment; represents a longitudinal position of the leader at the kth moment; represents a longitudinal position of the follower at the kth moment; represents a preset safety distance; represents a control input vector of the ith vehicle; represents a lower bound of the control input vector; represents an upper bound of the control input vector; represents a speed of the ith vehicle; represents a maximum vehicle speed; represents a state vector of the follower at the kth moment; represents a state vector of the leader at the kth moment.

[0059] S52, solving the game optimization problem.

[0060] In the Stackelberg game framework, the optimization solution problem of the leader and the follower is a typical bi-level optimal control problem, the follower makes a response action according to the strategy of the leader, and therefore the upper optimization is the leader optimal control and the lower optimization is the follower optimal control, that is, the risk-cut interactive Stackelberg game optimization problem can be transformed into the leader optimal control problem and the follower optimal control problem, which is expressed as: , wherein, denotes the leader state vector; denotes the leader control input vector; denotes the follower optimal control input vector; denotes the follower control input vector; denotes the follower state vector.

[0061] In the foregoing modeling, the utility function of the follower is quadratic, the constraints are composed of system dynamics equations and collision constraints, and both are in the affine form, which is a convex optimization problem, then according to the convex optimization property of the above-mentioned, the KKT condition is introduced to transform the risk-cut interactive Stackelberg game optimization problem into a single-layer convex optimization problem, which is: , , , , wherein, and both denote the Lagrange coefficient; denotes the risk driving regulation coefficient; denotes the compliance natural driving regulation coefficient; denotes the utility function of the leader, that is, the test confrontation vehicle; denotes the risk confrontation utility function; denotes the control input vector of the follower at the kth moment; denotes the state transition matrix; denotes the input gain matrix; denotes the control input vector of the leader at the kth moment; denotes the longitudinal position of the leader at the kth moment; denotes the longitudinal position of the follower at the kth moment; denotes the preset safety distance; denotes the control input vector of the ith vehicle; denotes the lower bound of the control input vector; denotes the upper bound of the control input vector; denotes the speed of the ith vehicle; represents the maximum vehicle speed; represents the Lagrangian function; represents the state vector of the follower at the kth moment; represents the state vector of the leader at the kth moment.

[0062] Finally, the single-layer convex optimization problem is solved by using CasADi and Ipopt solvers to obtain the optimal trajectory of the leader interaction strategy.

[0063] The obtained optimal trajectory of the leader interaction strategy is sent to a simulation environment (such as CARLA) or a real vehicle test platform in real time for closed-loop measurement and setting, and statistical indicators (TTC, driving speed change curve) are output.

[0064] After one round of S1-S5 is executed, it is selected according to the requirement whether to continue testing, if it is required to continue, the driving trajectory data generated after the current round of testing is executed is collected, and S1-S5 is executed again.

[0065] In order to verify the feasibility of the method, the method provided by the application is used to test a high-risk automatic driving cut-in scene, and a result graph as shown in Figure 2 can be drawn according to the simulation result, and it can be seen from Figure 2 that the curve in the relative distance-time step diagram shows that the distance between the two vehicles gradually decreases during the test, and the collision risk gradually increases; the curve in the speed-time step diagram shows that the test vehicle can adjust the trajectory in real time according to the speed change of the tested vehicle, and form a dynamic interaction, that is, the method provided by the application is feasible.

[0066] In addition, the application provides a multi-driving style high-risk automatic driving cut-in scene test system and an electronic device, which are used to implement the above method. The electronic device of the application includes a central processing unit (CPU), which can execute various appropriate actions and processes according to computer program instructions stored in a read-only memory (ROM) or computer program instructions loaded from a storage unit into a random access memory (RAM). Various programs and data required for device operation can also be stored in the RAM. The CPU, ROM and RAM are connected to each other through a bus. An input / output (I / O) interface is also connected to the bus.

[0067] The multiple components in the device are connected to the I / O interface, including: an input unit such as a keyboard, a mouse, etc.; an output unit such as various types of displays, speakers, etc.; a storage unit such as a magnetic disk, an optical disk, etc.; and a communication unit such as a network card, a modem, a wireless communication transceiver, etc. The communication unit allows the device to exchange information / data with other devices through a computer network such as the Internet and / or various telecommunications networks.

[0068] The processing units perform the various methods and processes described above, such as methods S1-S5. For example, in some embodiments, methods S1-S5 can be implemented as a computer software program tangibly embodied in a machine readable medium, such as a storage unit. In some embodiments, portions or all of the computer program can be loaded and / or installed onto the device via the ROM and / or the communication unit. When the computer program is loaded onto the RAM and executed by the CPU, one or more steps of the above-described methods S1-S5 can be performed. Alternatively, in other embodiments, the CPU can be configured to perform methods S1-S5 by any other suitable means, such as by way of firmware.

[0069] The functionality described herein above can be performed, at least in part, by one or more hardware logic components. For example, and without limitation, illustrative types of hardware logic components that can be used include Field-programmable Gate Arrays (FPGAs), Application-specific Integrated Circuits (ASICs), Application-specific Standard Products (ASSPs), System-on-a-chip systems (SOCs), Complex Programmable Logic Devices (CPLDs), etc.

[0070] Program code for carrying out the methods of the present application can be written in any combination of one or more programming languages. This program code can be provided to a processor or controller of a general purpose computer, special purpose computer, or other programmable data processing apparatus to produce a machine, such that the program code, when executed by the processor or controller, causes the machine to perform the functions / acts specified in the flow diagrams and / or block diagrams. The program code can be embodied in whole or in part within a machine readable medium, which can be any medium for storing or transmitting the program code. The program code can be transmitted in the form of signals over a transmission medium via a data signal or carrier wave, or it can be provided on a machine readable medium.

[0071] In the context of the present application, a machine-readable medium can be any tangible medium that can contain, or store a program for use by or in connection with an instruction execution system, apparatus, or device. The machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. The machine-readable medium can include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any suitable combination of the foregoing. More specific examples of the machine-readable storage medium will include one or more lines of a computer program code, a portable computer diskette, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or Flash memory), an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0072] The above merely illustrates the specific embodiments of the present application, but the protection scope of the present application is not limited thereto, and any skilled person in the art can easily think of various equivalent modifications or replacements within the technical range disclosed by the present application, and these modifications or replacements shall be covered within the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the protection scope of the claims.

Claims

1. A multi-driving style high-risk autonomous driving cut-in scene test method, characterized in that, The method comprises the following steps: Collecting original driving trajectory data and constructing physical feasibility constraints, clustering the original driving trajectory data by driving style, and using a Cutin-TimeGAN network model combined with the physical feasibility constraints to generate a risk cut-in scenario trajectory cluster with consistent driving style; the driving style includes conservative, ordinary and aggressive; Constructing a cost function, calculating the cost function value for each risk cut-in scenario trajectory cluster, and selecting a target cut-in reference trajectory as a target reference state vector based on the cost function value; Constructing a Stackelberg game framework, constructing an utility function based on the state vectors of the VUT and the test adversarial vehicle in the framework, and combining the target reference state vector; Calculating the predicted collision time of the VUT and the test adversarial vehicle, and constructing a risk confrontation utility function based on the predicted collision time; Based on the utility function and the risk confrontation utility function, a risk cut-in interaction Stackelberg game optimization problem is constructed and solved to obtain the optimal interaction trajectory of the test adversarial vehicle, and based on the optimal interaction trajectory, a closed-loop high-risk automatic driving cut-in scenario test is performed.

2. The multi-driving style high-risk automatic driving cut-in scene test method according to claim 1, wherein The physical feasibility constraints include: For each vehicle, its state always moves forward, i.e. the vehicle's longitudinal displacement at the previous time is less than the vehicle's longitudinal displacement at the next time; For the vehicle after the end of the scene cut-in, the vehicle's lateral displacement point at the end time is located within the left and right boundaries of the lane; For each vehicle, the difference in lateral displacement between adjacent time frames does not exceed a difference threshold.

3. The multi-driving style high-risk automatic driving cut-in scene test method according to claim 1, wherein The loss function of the Cutin-TimeGAN network model includes an embedder reconstruction loss, a discriminator loss and a generator adversarial loss; The embedder reconstruction loss is: , represents a calculation mathematical expectation operation, represents input driving trajectory data, represents a driving trajectory reconstructed by the embedder. The discriminator loss is: , denotes a discrimination result output by the discriminator based on the latent vector H, denotes a cross-entropy loss calculation operation, denotes a discrimination result output by the discriminator based on the generated latent representation , denotes a weight coefficient, denotes a discrimination result output by the discriminator based on the representation generated by the random noise , The generator adversarial loss is: , represents an adversarial loss function, has ; represents a supervised loss, has , represents a true latent representation at t+1 time, represents a prediction result at t+1 time; represents a statistical matching loss, has , represents a mean calculation operation, represents trajectory data generated by the Cutin-TimeGAN network model; represents a standard deviation calculation operation.

4. The multi-driving style high-risk automatic driving cut-in scene test method according to claim 1, wherein For the ith trajectory, the corresponding cost function is: ， wherein, , and all represent style weight coefficients; represents a total step length of planning; represents a unit time length; represents a vehicle longitudinal speed at time t, and represents a time at the jthplanning step; represents a reference vehicle speed; represents t j a vehicle longitudinal acceleration at time t; represents a vehicle lateral acceleration at time t.

5. The multi-driving style high-risk automatic driving cut-in scene test method according to claim 1, wherein In the Stackelberg game framework, the VUT is set as the follower F, and the test adversarial vehicle is set as the leader L, and the state vector forms of the two are set as: The state transition of the follower and the leader follows: ， ， wherein, denotes the state vector of the follower F at time k; denotes the state vector of the leader L at time k; and denote the longitudinal position of the follower and the leader at time k, respectively; and denote the longitudinal velocity of the follower and the leader at time k, respectively; and denote the lateral position of the follower and the leader at time k, respectively; and denote the lateral velocity of the follower and the leader at time k, respectively. The state weight coefficient matrix is constructed as: wherein, represents the weight proportion coefficient of different states; The control input weight matrix is constructed as: wherein, represents the weight proportion coefficient of different control inputs; and the control input vector is set as: , represents the longitudinal control input of the vehicle at time k, represents the lateral control input of the vehicle at time k; The utility function is: ， ， wherein, denotes the control input vector of the follower at time k; denotes the state transition matrix; denotes the input gain matrix; denotes the control input vector of the leader at time k.

6. The multi-driving style high-risk automatic driving cut-in scene test method according to claim 1, wherein The risk confrontation utility function is: ， ， wherein, represents the utility function of the follower, i.e. the VUT; represents the state vector of the follower at time k+1; represents the target reference state vector of the follower; represents the state weight coefficient matrix; represents the control input vector of the follower at time k; represents the control input weight matrix; represents the total step size of the planning; represents the utility function of the leader, i.e. the test opponent vehicle; represents the state vector of the leader at time k+1; represents the target reference state vector of the leader; represents the control input vector of the leader at time k.

7. The multi-driving style high-risk automatic driving cut-in scene test method according to claim 1, wherein The risk cut-in interaction Stackelberg game optimization problem is: ， wherein, denotes a planned total step length; denotes a follower traveled trajectory length; denotes a follower speed; denotes a leader traveled trajectory length; denotes a leader speed.

8. The multi-driving style high-risk automatic driving cut-in scene test method according to claim 1, wherein The method for solving the risk cut-in interaction Stackelberg game optimization problem is: ， ， ， wherein, denotes the leader's control input vector; denotes the risk driving regulation coefficient; denotes the compliance natural driving regulation coefficient; denotes the leader's utility function, i.e. the test opponent vehicle; denotes the risk opponent utility function; denotes the follower's control input vector at time k; denotes the state transition matrix; denotes the input gain matrix; denotes the leader's control input vector at time k; denotes the leader's longitudinal position at time k; denotes the follower's longitudinal position at time k; denotes the preset safety distance; denotes the i-th vehicle's control input vector; denotes the control input vector lower bound; denotes the control input vector upper bound; denotes the i-th vehicle's speed; denotes the maximum vehicle speed; denotes the follower's state vector at time k; denotes the leader's state vector at time k.

9. The multi-driving style high-risk automatic driving cut-in scene test method according to claim 1, wherein According to the convex optimization property of the follower optimal control problem, the KKT condition is introduced to convert the risk cut-in interaction Stackelberg game optimization problem into a single-layer convex optimization problem, which is: The risk-cutting interactive Stackelberg game optimization problem is converted into a leader optimal control problem and a follower optimal control problem, namely: wherein, xLrepresents a leader state vector; uLrepresents a leader control input vector; uF*represents a follower optimal control input vector; uFrepresents a follower control input vector; xFrepresents a follower state vector; The single-layer convex optimization problem is solved by using CasADi and Ipopt solvers. ， ， ， ， wherein, and both represent Lagrange coefficients; represents a risk driving regulation coefficient; represents a compliance natural driving regulation coefficient; represents a leader, i.e. a test opponent vehicle, utility function; represents a risk opponent utility function; represents a follower k-th time instant control input vector; represents a state transition matrix; represents an input gain matrix; represents a leader k-th time instant control input vector; represents a leader k-th time instant longitudinal position; represents a follower k-th time instant longitudinal position; represents a preset safety distance; represents an i-th vehicle control input vector; represents a control input vector lower bound; represents a control input vector upper bound; represents an i-th vehicle speed; represents a maximum vehicle speed; represents a Lagrange function; represents a follower k-th time instant state vector; represents a leader k-th time instant state vector; The system is used to implement the method of any one of claims 1-9.

10. A multi-driving style high-risk automated driving cut-in scenario testing system, characterized in that, ​

Images