A pedestrian-like merging trajectory control method for mixed traffic flow based on random potential game

Abstract

The present invention discloses a mixed traffic flow human-like merging trajectory control method based on random potential game, which belongs to the field of autonomous driving technology. The method includes: establishing a random potential game model that takes into account the influence of communication delay or sensor uncertainty, determining the potential function in the random potential game model, and calibrating the parameters in the random potential game model using an open source data set; collecting real-time status information of all vehicles within the merging area of ​​the highway as input information for trajectory planning; constructing a model predictive controller based on the random potential game model and using a distributed algorithm to solve it; using the real-time vehicle status information, the model calculates the vehicle action combination at the next moment, obtains the vehicle planning trajectory, and updates the vehicle status. The present invention can accurately describe the driving decision control logic of human drivers and generate a planned trajectory, enabling good interaction between autonomous vehicles and manually driven vehicles, and improving the overall traffic efficiency and safety of the merging area.

Description

Technical Field

[0001] The present invention belongs to the field of autonomous driving technology, and in particular relates to a method for controlling human-like merging trajectories in mixed traffic flows based on random potential game. Background Art

[0002] Freeway ramp merging areas are typical traffic bottlenecks, often causing potential conflicts due to merging vehicles. With the development of connected vehicles and autonomous driving technologies, vehicles can achieve microscopic trajectory control through connected vehicles, effectively mitigating these conflicts and significantly improving traffic efficiency and safety in merging areas. Currently, much research is focused on cooperative merging strategies in purely autonomous driving environments, aiming to optimize the interaction and coordination between autonomous vehicles. This research has made significant progress in improving the collaborative performance of autonomous vehicles, laying a solid foundation for the development of autonomous driving technology.

[0003] Despite significant progress in autonomous driving technology, its practical application in mixed traffic environments (where autonomous vehicles and human-driven vehicles coexist) still faces numerous challenges. First, existing research has paid little attention to merging control in mixed traffic environments, making it difficult to effectively address the interaction between autonomous vehicles and human-driven vehicles. Second, research on human-like interaction mechanisms is limited, particularly the impact of different driving styles on merging decisions and control. Furthermore, most existing decentralized control algorithms optimize agent behavior only from an individual perspective, with little consideration of the impact of individual behavior on the global system, resulting in insufficient overall coordination. Finally, existing research often overlooks uncertainties caused by sensor errors or communication delays, which are difficult to address in practical applications. This further reduces the level of interaction between autonomous vehicles and human-driven vehicles and limits the widespread application of autonomous driving technology in mixed traffic environments. Summary of the Invention

[0004] In order to solve the above technical problems, the present invention proposes a mixed traffic flow human-like merging trajectory control method based on random potential game to solve the problems existing in the above-mentioned prior art.

[0005] To achieve the above objectives, in a first aspect, the present invention provides a method for controlling a mixed traffic flow human-like merging trajectory based on a random potential game, comprising:

[0006] Establish a random potential game model that considers the effects of communication delay or sensor uncertainty, determine the potential function in the random potential game model, and calibrate the parameters in the random potential game model using open source datasets;

[0007] Collect real-time status information of all vehicles in the merging area of ​​the highway as input information for trajectory planning;

[0008] Based on the random potential game model, a model predictive controller is constructed and solved using a distributed algorithm. The real-time vehicle status information is used to calculate the vehicle action combination at the next moment through the random potential game model to obtain the vehicle planning trajectory and update the vehicle status.

[0009] Preferably, the process of establishing a random potential game model that considers the effects of communication delay or sensing uncertainty includes:

[0010] According to the vehicle's search for the best action under the random potential game model, the original goal is converted into a mathematical expectation with a random vector, and the deterministic constraints are converted into chance constraints;

[0011] Use sampling average approximation method to transform uncertain mathematical expectations and conditional constraints into deterministic problems;

[0012] Determining a potential function in a game model, wherein the potential function is a continuous and precise potential game potential function;

[0013] Interacting vehicle trajectories in highway merging areas are extracted from open source datasets, and parameter calibration is performed using a model method.

[0014] Preferably, the potential function includes autocorrelation cost and coupling cost, the autocorrelation cost includes traffic efficiency cost, driving comfort cost, and cost of deviation from the target state, and the coupling cost includes safety cost.

[0015] Preferably, the parameter calibration adopts a two-layer programming model, the upper layer model minimizes the mean absolute error between the predicted action and the actual action, and the lower layer model solves the potential game equilibrium.

[0016] Preferably, all vehicle real-time status information includes:

[0017] Collect the vehicle's current lateral and longitudinal position, lateral and longitudinal speed, and lateral and longitudinal acceleration as dynamic input information;

[0018] Collect vehicle type, vehicle lane, and vehicle driving style factors as static input information;

[0019] Collect information on merging area length, road width, and road boundaries.

[0020] Preferably, the step of constructing a model predictive controller includes:

[0021] Establish an MPC optimization model that considers vehicle dynamics constraints, road conditions, traffic regulations, and other vehicle behaviors;

[0022] Decompose the original optimization problem into a vehicle unilateral decision-making problem and an auxiliary variable update problem;

[0023] A distributed algorithm is used for solving the problem.

[0024] Preferably, the steps of solving the problem using a distributed algorithm include:

[0025] With the auxiliary and dual variables fixed, the unilateral decision problem is solved independently for each vehicle to optimize its control strategy.

[0026] When the vehicle control variables and dual variables are fixed, the decision-making among vehicles is coordinated by solving the updating problem of auxiliary variables.

[0027] The dual variables are adjusted according to the update rule of the Lagrange multiplier method to gradually satisfy the constraints.

[0028] Preferably, the trajectory planning is a dynamic process, generating all vehicle trajectories at the next moment according to the state at the current moment.

[0029] In a second aspect, the present invention further discloses a computer-readable storage medium having a computer program stored thereon, which implements the steps of the method described in the first aspect when executed by a processor.

[0030] In a third aspect, the present invention further discloses a computer program product, comprising a computer program, which implements the steps of the method described in the first aspect when executed by a processor.

[0031] Compared with the prior art, the present invention has the following advantages and technical effects:

[0032] The present invention provides a mixed traffic flow human-like merging trajectory control method based on random potential game, comprising: first, establishing a random potential game model that takes into account the influence of communication delay or sensor uncertainty, determining the potential function in the random potential game model, and calibrating the parameters in the random potential game model using an open source data set; second, collecting real-time status information of all vehicles within the merging area of ​​the highway as input information for trajectory planning; finally, constructing a model predictive controller based on the random potential game model and solving it using a distributed algorithm; using the real-time status information of the vehicles, calculating the vehicle action combination at the next moment through the random potential game model, obtaining the vehicle planning trajectory, and updating the vehicle status.

[0033] The present invention considers random uncertainties caused by communication or sensor errors and solves the optimal control of the autonomous driving vehicle in a stable and efficient manner.

[0034] The present invention constructs a potential function to capture the decision-making logic and behavioral randomness of human drivers; by understanding and learning the mechanism of human driving behavior, it provides real-time human driving behavior prediction for autonomous driving vehicles in highway merging areas.

[0035] The present invention analyzes the driving behavior characteristics of different driving styles in natural driving data, calibrates the weight parameters in the random potential game model, and characterizes the decision-making and response of human drivers under different traffic scene conditions.

[0036] The present invention establishes a model predictive controller to realize human-like merging trajectory planning based on a calibrated game model; and adopts a model predictive control algorithm to solve the optimal control output and plan the future movement of the autonomous driving vehicle.

[0037] This invention builds a distributed solution algorithm, transforming centralized optimization into a distributed optimization framework. This allows for decentralized control of individual autonomous vehicles, improving solution efficiency. Ultimately, this approach enables smooth interaction between autonomous and manually driven vehicles, improving traffic efficiency and safety within merging areas. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] The accompanying drawings, which constitute part of this application, are intended to provide a further understanding of this application. The exemplary embodiments and descriptions of this application are intended to explain this application and do not constitute an improper limitation on this application. In the accompanying drawings:

[0039] Figure 1 is a flow chart of a method according to an embodiment of the present invention;

[0040] Figure 2 A schematic diagram showing how various utilities of an embodiment of the present invention vary with acceleration;

[0041] Figure 3 This is a parameter calibration flow chart of an embodiment of the present invention;

[0042] Figure 4 Generate a planning trajectory and motion state diagram for an embodiment of the present invention. DETAILED DESCRIPTION

[0043] It should be noted that, in the absence of conflict, the embodiments and features of the embodiments in this application can be combined with each other. The present application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0044] It should be noted that the steps shown in the flowcharts of the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and that, although a logical order is shown in the flowcharts, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0045] Example 1

[0046] like Figure 1 As shown, this embodiment provides a mixed traffic flow human-like merging trajectory control method based on random potential game, including:

[0047] S1. Construct a random potential game model, taking into account the uncertainty caused by communication delays or sensing; determine the potential function in the model; and use open source datasets to calibrate the parameters in the model.

[0048] Specifically, the S1 includes:

[0049] S11. By characterizing the mechanism of human driving behavior, a random potential game model is established to find the best action:

[0050] (1)

[0051] (2)

[0052] Where, Indicates the best action, represents a random vector, which indicates the uncertainty caused by communication delay or sensing. express The mathematical expectation of . express The original target is converted to a random vector The mathematical expectation of Deterministic constraints Converted into a chance constraint , Represents the probability of failure.

[0053] S12. Use the sampling average approximation method to transform the uncertain mathematical expectations and conditional constraints in step S11 into deterministic ones.

[0054] (3)

[0055] (4)

[0056] Where, represents the mathematical expectation of cost, Indicates the The samples are drawn independently from the distribution of the vehicle's initial states. Indicates the maximum number of approximate sampling times.

[0057] Specifically, the sampling average approximation method is implemented as follows:

[0058] First, from the random variable Generate N samples in Then, we use the generated samples to construct an approximate deterministic optimization problem. For the objective function , its expected value An approximate replacement can be made by taking the mean of the sample:

[0059] (5)

[0060] Where, is the objective function of the sampling average approximation.

[0061] Then, solve the above deterministic optimization problem and obtain the approximate optimal solution .

[0062] Finally, evaluate the quality of the approximate solution. If the final solution does not meet the requirements, increase the number of samples N and repeat the above steps to improve the accuracy of the approximate solution. The sampling average approximation method can gradually improve the accuracy of the approximate solution by increasing the number of samples.

[0063] S13. The potential function in the game model needs to be a continuous and precise potential game potential function if and only if is a connected set, and there exists a potential function , so that its partial derivative is equal to the cost function of each participant The partial derivatives of are consistent.

[0064] (6)

[0065] Potential function in the game model:

[0066] (7)

[0067] Where, represents the total potential function, and is the driving style weight factor, and denote the autocorrelation cost and coupling cost respectively.

[0068] Specifically, the potential function can be expressed as:

[0069] (8)

[0070] (9)

[0071] (10)

[0072] (11)

[0073] (12)

[0074] Where, The weight factor representing human driving style can reflect the driver's trade-off between different indicators. For example, an aggressive driver may pay more attention to traffic efficiency and ignore safety. The cost of traffic efficiency can be calculated as the sum of the squares of the deviations between the vehicle's lateral and longitudinal speeds and the expected speed. This cost decreases if the longitudinal or lateral speed is closer to the expected speed. and Indicates vehicle The expected longitudinal and lateral velocities. The cost of driving comfort can be calculated as the sum of the squares of the vehicle's longitudinal and lateral accelerations. This cost decreases when the longitudinal or lateral acceleration decreases. The cost of deviation from the target state is composed of three parts: the square of the deviation between the lateral position and the center line of the target lane, the square of the lateral velocity and the square of the lateral acceleration. The vehicle moves upwards while its lateral velocity and lateral acceleration are zero. represents the safety cost, if the vehicle and If the vertical or horizontal spacing between the two is greater than or equal to the safety distance, .otherwise, , the safety cost will gradually increase as the distance between vehicles decreases, thus prompting vehicles to stay away from surrounding vehicles. and Represents the longitudinal and lateral safety distances respectively, and represents the length and width of the vehicle. The utility changes with acceleration as shown in Figure 2 shown.

[0075] S14. Extract the trajectories of vehicles interacting in a highway merging area over a period of time from an open source dataset and calibrate the parameters using the following model method:

[0076] (13)

[0077] (14)

[0078] Where, Represents a given parameter vector The optimal action of the potential game under , the result is the predicted value of the optimal action. Indicates the true acceleration.

[0079] because Since this is a nonlinear programming problem, a heuristic algorithm can be used to solve the optimal parameter vector in the upper model. The lower model is a standard mathematical optimization form, so a global optimization algorithm can be used to solve it.

[0080] The algorithm flow chart of the parameter calibration framework is as follows Figure 3 As shown, set the initial parameter vector , the number of iterations is 0; calculate the parameter vector The cost function under [1] is used; the real trajectory data can be extracted from the open source dataset INTERACTION; a two-layer model is solved, where the upper model minimizes the mean absolute error between the predicted action and the real action, and the lower model solves the potential game equilibrium; if the preset iteration stop condition is reached, the optimal estimated parameters are output. , otherwise continue iterating.

[0081] S2. Collect real-time status information of all vehicles within the merging area of ​​the highway as input information for trajectory planning;

[0082] Specifically, S2 includes:

[0083] Roadside infrastructure is used to collect real-time status information from all vehicles within the merging area of ​​a highway. This information includes current lateral and longitudinal position, lateral and longitudinal speed, and lateral and longitudinal acceleration, which serves as real-time dynamic input for trajectory planning. Static inputs include vehicle type, lane, and driving style factors. Furthermore, structural road information, such as merging area length, road width, and road boundaries, is collected. This information is transmitted to autonomous vehicles for decision-making and control. For manually driven vehicles, only information about vehicles in the preceding lane and target lane is transmitted for following and lane-changing control decisions.

[0084] S3. Build a model predictive controller and use a distributed algorithm to solve it. Based on the real-time status information of the vehicle, use the model to calculate the vehicle action combination at the next moment, obtain the vehicle planning trajectory, and update the vehicle status.

[0085] Specifically, the S3 includes:

[0086] S31. The model predictive controller based on random potential game and distributed algorithm solution is:

[0087] (15)

[0088] (16)

[0089] Where, is an auxiliary variable.

[0090] The model predictive controller based on random potential game considers factors such as vehicle dynamic constraints, road conditions, traffic rules, and the behavior of other vehicles, and calculates the optimal control quantity in real time. The corresponding MPC optimization model can be established:

[0091] (17)

[0092] in, ; ; ; ;

[0093] Where, Indicates from arrive Cumulative cost of potential function within time range . 、 、 、 They are respectively the preset minimum longitudinal acceleration, maximum longitudinal acceleration, minimum lateral acceleration, and maximum lateral acceleration of the vehicle. and Indicates the preset maximum longitudinal and lateral speeds. Indicates the upper boundary of the lane, Indicates the lower boundary of the lane. Indicates the width of the vehicle.

[0094] S32. The original problem can be decomposed into two sub-problems, namely the unilateral decision problem of the vehicle and the auxiliary variable Update issue.

[0095] In each iteration, the specific steps are as follows:

[0096] 1) Update vehicle control variables : In fixed auxiliary variables and dual variables In the case of , the unilateral decision problem is solved independently for each vehicle to optimize its control strategy;

[0097] 2) Update auxiliary variables : Fixed vehicle control variables and dual variables In the case of Update problem, coordinate the decision-making among vehicles;

[0098] 3) Update the dual variables :Adjust the dual variables according to the update rule of Lagrange multiplier method to gradually satisfy the constraints.

[0099] Distributed algorithm can be used to solve:

[0100] (18)

[0101] (19)

[0102] (20)

[0103] (twenty one)

[0104] Where, is the number of iterations of the ADMM algorithm. is the penalty parameter.

[0105] S33, the vehicle state update equation is:

[0106] (twenty two)

[0107] (twenty three)

[0108] (twenty four)

[0109] (25)

[0110] Where, Indicates the vertical position, Indicates horizontal position. represents the longitudinal velocity, Indicates the lateral velocity. Indicates adjacent time intervals.

[0111] The trajectory planning is a dynamic process. According to the state at the current time t, all vehicle trajectories at time t+1 are generated. The generated trajectory is as follows: Figure 4 shown.

[0112] Beneficial effects of this embodiment:

[0113] This embodiment establishes a potential function that takes into account traffic efficiency, driving objectives, comfort, and safety, accurately depicts the decision-making factors of human drivers in merging and lane changing, introduces a random potential game model to capture sensor errors, uses average approximate sampling to convert the random optimization model into a deterministic model, and models the interactive game between vehicles. The established model can reflect the merging and lane changing interactive logic of human drivers. A two-level planning model is used to calibrate the driving style weight factor in the model to more realistically depict human driving behavior. Finally, the random potential game is combined with model predictive control and applied to the trajectory planning and control of autonomous vehicles to achieve human-like merging control. The proposed planning and control framework can improve the interaction between autonomous vehicles and human-driven vehicles in mixed traffic, thereby improving regional traffic efficiency and safety.

[0114] Example 2

[0115] This embodiment further discloses a computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, the steps of the method described in the first embodiment are implemented.

[0116] Example 3

[0117] This embodiment further discloses a computer program product, including a computer program, which implements the steps of the method described in the first embodiment when executed by a processor.

[0118] The above are merely preferred embodiments of the present application, but the scope of protection of the present application is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in this application should be included in the scope of protection of the present application. Therefore, the scope of protection of the present application should be based on the scope of protection of the claims.

Claims

1. A mixed traffic flow human-like merging trajectory control method based on random potential game, characterized by: The following steps are involved: Establish a random potential game model that considers the effects of communication delay or sensor uncertainty, determine the potential function in the random potential game model, and calibrate the parameters in the random potential game model using open source datasets; The process of building a stochastic potential game model that considers the effects of communication delay or sensing uncertainty includes: According to the vehicle's search for the best action under the random potential game model, the original goal is converted into a mathematical expectation with a random vector, and the deterministic constraints are converted into chance constraints; By characterizing the human driving behavior mechanism, a random potential game model is established to find the best action: ； ； Where, Indicates the best action, represents a random vector, indicating uncertainty caused by communication delay or sensing, express The mathematical expectation of express The probability of the original target is converted to a random vector The mathematical expectation of , deterministic constraints Converted into a chance constraint , represents the probability of failure; Use sampling average approximation method to transform uncertain mathematical expectations and conditional constraints into deterministic problems; ； ； Where, represents the mathematical expectation of cost, Indicates the The samples are drawn independently from the distribution of the vehicle's initial state; Indicates the maximum number of approximate sampling; Determining a potential function in a game model, wherein the potential function is a continuous and precise potential game potential function; The trajectories of vehicles interacting in the merging area of ​​a highway are extracted from an open source dataset and the parameters are calibrated using a model method. Collect real-time status information of all vehicles in the merging area of ​​the highway as input information for trajectory planning; Based on the random potential game model, a model predictive controller is constructed and solved using a distributed algorithm. The real-time vehicle status information is used to calculate the vehicle action combination at the next moment through the random potential game model to obtain the vehicle planning trajectory and update the vehicle status.

2. The method according to claim 1, characterized in that The potential function includes autocorrelation cost and coupling cost. The autocorrelation cost includes traffic efficiency cost, driving comfort cost, and cost of deviation from the target state. The coupling cost includes safety cost.

3. The method according to claim 1, characterized in that The parameter calibration adopts a two-level programming model, the upper level model minimizes the mean absolute error between the predicted action and the real action, and the lower level model solves the potential game equilibrium.

4. The method according to claim 1, wherein All vehicle real-time status information includes: Collect the vehicle's current lateral and longitudinal position, lateral and longitudinal speed, and lateral and longitudinal acceleration as dynamic input information; Collect vehicle type, vehicle lane, and vehicle driving style factors as static input information; Collect information on merging area length, road width, and road boundaries.

5. The method according to claim 1, characterized in that The steps to building a model predictive controller include: Establish an MPC optimization model that considers vehicle dynamics constraints, road conditions, traffic regulations, and other vehicle behaviors; Decompose the original optimization problem into a vehicle unilateral decision-making problem and an auxiliary variable update problem; A distributed algorithm is used for solving the problem.

6. The method according to claim 5, characterized in that The steps of solving the problem using a distributed algorithm include: With the auxiliary and dual variables fixed, the unilateral decision problem is solved independently for each vehicle to optimize its control strategy. When the vehicle control variables and dual variables are fixed, the decision-making among vehicles is coordinated by solving the updating problem of auxiliary variables. The dual variables are adjusted according to the update rule of the Lagrange multiplier method to gradually satisfy the constraints.

7. The method according to claim 1, characterized in that The trajectory planning is a dynamic process that generates all vehicle trajectories at the next moment based on the current state.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

9. A computer program product comprising a computer program, characterized in that When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

Images