Constraint optimization driven non-cooperative game collision avoidance control method
By introducing a non-cooperative game collision avoidance control method driven by constraint optimization in game control, combining vehicle model and path planning module, the solution control actions are optimized, and the safety and stability problems caused by not considering constraints in the existing technology are solved, and the safety and stability problems caused by failure to consider constraints are achieved. Safer and stable man-machine co-driving control.
Patent Information
- Application Number
- CN202411396685.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-10-08
- Publication Date
- 2024-11-19
- Estimated Expiration
- Not applicable · inactive patent
AI Technical Summary
The existing game control scheme does not consider system constraints, resulting in the control action that may exceed the physical limit of the actuator, affecting smoothness and comfort, and being unable to ensure safety in complex driving environments, which may lead to collisions or vehicle instability.
A non-cooperative game collision avoidance control method driven by constraint optimization is proposed. Through the path planning module and the constrained non-cooperative game controller, combined with the vehicle model, the optimal control action is optimized and solved, and the physical constraints, safety constraints and state variables of the executor are considered. Constrain and ensure that the control actions are within a reasonable range.
It effectively avoids the control action exceeding the limit of the actuator, ensures the smoothness and comfort of the control action, ensures the safety and stability of the vehicle jointly driven by a man-machine, and avoids the occurrence of side slip and tail swing.
Smart Images

Figure CN118977704A_ABST
Abstract
Description
Technical field:
[0001] The present invention relates to the field of intelligent vehicle human-machine collaborative control, and specifically to a non-cooperative game collision avoidance control method driven by constraint optimization. Background technology:
[0002] In recent years, the degree of automation of automobiles has gradually increased. Intelligent driving can not only improve driving safety and reduce accidents caused by human errors, but also reduce the operating burden of drivers and improve driving comfort. However, unmanned driving technology is still not mature enough. Therefore, human-machine co-driving control is a necessary transition stage from assisted driving to unmanned driving, and human-machine co-driving is still the main research direction of intelligent driving cars. Automated control systems surpass humans in information storage and computing power, while human drivers surpass automation in moral judgment and reasoning; human-machine collaborative control fully combines the respective advantages of drivers and automatic control systems, which is of great significance to the safe driving of vehicles.
[0003] In terms of human-machine collaborative control, game theory can effectively handle the decision-making interaction between the two parties in the game and improve the overall coordination and efficiency of the system. However, existing game control schemes usually solve Nash equilibrium solutions without considering system constraints. Since the physical constraints of the actuator are not considered, the control actions generated by the game may exceed the physical limits of the actuator; at the same time, it will also affect the smoothness of the control action and the comfort of the driver; since safety constraints are not considered, the safety of the game path cannot be guaranteed. When the driving environment is too complex, the path generated by the game between the driver and the automated control system is not necessarily safe, which may lead to collisions. In addition, without considering the state quantity constraints, skidding or tail-swinging may occur during emergency obstacle avoidance, and the stability of the vehicle cannot be guaranteed. Summary of the invention:
[0004] In order to solve the above problems, the present invention proposes a non-cooperative game collision avoidance control method driven by constraint optimization. First, according to the road information, the driver's path planning and the controller's path planning are used to obtain the expected paths of the driver and the controller respectively. Secondly, a vehicle model is established, and the actual motion state of the vehicle is fed back to the constrained non-cooperative game controller. Finally, the optimal control action is optimized and solved by the constrained non-cooperative game controller to achieve active collision avoidance control. Three types of constraints are considered in the constrained non-cooperative game controller: actuator physical constraints; safety constraints; state variable constraints. Among them, the actuator physical constraints are used to prevent the control actions of the game from exceeding the manipulation limit of the actuator. Safety constraints are imposed to ensure the driving safety of human-machine co-driving vehicles. The state variables are constrained to constrain the vehicle state within a safe range to ensure the driving stability of the vehicle.
[0005] The technical solution adopted by the present invention to solve the technical problem is as follows:
[0006] A non-cooperative game collision avoidance control method driven by constraint optimization, characterized in that the method comprises a path planning module, a constrained non-cooperative game controller, and a vehicle model; the path planning module obtains the driver's expected path and the controller's expected path respectively according to road information through driver path planning and controller path planning, and inputs the expected paths of the two into the constrained non-cooperative game controller, wherein an interactive prediction model is constructed for prediction output to obtain the lateral displacement and yaw angle of the vehicle, and on this basis, the actual vehicle state information is output in combination with the expected paths of the driver and the controller and the vehicle model, including the lateral speed v of the vehicle y , yaw angular velocity ω r , lateral displacement y and yaw angle Construct a constrained optimization problem and determine the optimal control action by solving the quadratic programming problem, including the driver's front wheel steering angle δ' d and the controller's front wheel steering angle δ c ', input to the vehicle model, and control the vehicle to achieve collision avoidance control.
[0007] The method comprises the following steps:
[0008] Step 1: Design a path planning module to obtain the desired path of the driver model and the controller, including lateral displacement and yaw angle, by combining the road information with the path planning of the driver and the controller. The process includes the following sub-steps:
[0009] Step 1.1, controller planning: Solve the lateral displacement and yaw angle of the controller reference based on the fifth-order polynomial method to obtain the desired path of the controller;
[0010] Step 1.2: When planning the driver's path, build a driving environment based on the Prescan simulation platform, and perform joint simulation with the Logitech G29 steering wheel to determine the driver's lateral displacement and yaw angle to obtain the driver's desired path.
[0011] Step 2: Design a constrained non-cooperative game controller, which includes the following sub-steps:
[0012] Step 2.1: Establish an interactive prediction model, which includes the following sub-steps:
[0013] Step 2.1.1: Construct the vehicle dynamics model, the expression is as follows:
[0014]
[0015] Among them, v x is the longitudinal velocity of the vehicle, v y is the lateral velocity of the vehicle in the direction of the vehicle's transverse axis, m is the vehicle's mass, ω is the vehicle's yaw rate, is the lateral acceleration of the vehicle, C αf and C αr are the cornering stiffness of the front and rear tires of the vehicle, δ is the front wheel turning angle, l f and l r are the distances from the vehicle's center of mass to the front and rear axles, I z is the yaw moment of inertia about the vertical axis of the vehicle's center of mass, is the vehicle yaw angular acceleration, is the vehicle yaw angle;
[0016] Step 2.1.2: Convert the vehicle dynamics model (1) into a state space equation as follows:
[0017]
[0018] ε is the state variable of the vehicle, including lateral velocity, yaw rate, lateral displacement and yaw angle, A and B are coefficient matrices, and u is the control variable δ;
[0019] in,
[0020]
[0021] Where G is the overall steering gear ratio;
[0022] Step 2.1.3: Discretize the state space equation (2) and construct the interaction model between the driver and the controller based on the discrete result. The expression is as follows:
[0023]
[0024] Among them, δ driver is the control input action of the driver model, δ control is the control input action of the controller, ε is the state variable of the vehicle, is the output matrix, A v , B1, B2, C v is the coefficient matrix;
[0025] Step 2.1.3: Calculate the predicted output of the system. According to the basic principle of model predictive control, the predicted output at time t can be obtained as:
[0026] Y(t)=ψε(t)+Θ d U driver (k)+Θ c U control (k) (4)
[0027] Y(t) is the control output matrix, U driver (k) and U control(k) are the control input matrices of the driver and controller, ε(t) is the state matrix, ψ, Θ d and θ c is the coefficient matrix;
[0028] Step 2.2: The constrained non-cooperative game controller constructs a constrained optimization problem for the predicted output, including the following sub-steps:
[0029] Step 2.2.1. Design the objective function, predict the output vehicle trajectory Y(k) according to the interactive prediction model, and on this basis, combine the driver's expected path R d (k) and the controller expected path R c (k), and the state information fed back by the vehicle to establish the objective function, which is expressed as follows:
[0030]
[0031] Considering the control limits of the vehicle's physical actuators, the control input U driver (k) and the control input U of the controller control (k) is constrained. Considering the safety of human-machine co-driving, constraints are imposed on the vehicle control output Y(k), and its expression is as follows:
[0032]
[0033] in,
[0034]
[0035]
[0036]
[0037] Among them, Q d and Q c are the weights assigned to the driver and the controller, including the lateral displacement weight and the yaw angle weight, R dl and R cl are the front wheel steering weights of the driver and controller, R d (k) and R c (k) are the desired paths of the driver and controller, respectively, U driver (k) and U control (k) are the control variables of the driver and the controller, respectively, U dmin (k) and U dmax (k) represents the set of minimum and maximum values of the driver's control quantity in the control time domain, U cmin (k) and U cmax (k) represents the set of minimum and maximum values of the controller control quantity in the control domain, Ymin (k) and Y max (k) is the set of minimum and maximum values of the control output;
[0038] Step 3: Solve the quadratic programming problem, transform the constrained optimization problem into a QP problem, and obtain the optimal control action through the solver. The specific process is as follows:
[0039] Define the matrix:
[0040] G1=ψε(k)-R d (k), G2=ψε(k)-R c (k), Z = (Θ d Θ c ),
[0041] Among them, G1 and G2 are the tracking errors of the driver and controller respectively, Z is the coefficient matrix, R is the weight matrix, and U is the control amount;
[0042] Establish a quadratic programming mathematical model:
[0043]
[0044] Convert the objective function into quadratic programming form:
[0045]
[0046] in,
[0047]
[0048] In the constraint matrix,
[0049]
[0050] The optimal control sequence U is solved by the quadratic programming method, and the first and Nth M+1 The combination of the first element of the driver control input and the first element of the controller control input is the optimal control action, and the optimal control action instruction is input to the vehicle model to achieve safe collision avoidance control.
[0051] The beneficial effects of the present invention are: considering the physical constraints of the actuator to avoid the control action of the game exceeding the operating limit of the actuator, while ensuring the smoothness of the control action and the driving comfort; considering the safety constraints to ensure the driving safety of the human-machine co-driving vehicle, and in the face of complex working conditions, ensuring the safety of the operating instructions jointly issued by the driver and the controller; considering the state variable constraints, constraining the vehicle state within a safe range, ensuring the driving stability of the vehicle, and avoiding excessive side slip and tail swinging of the vehicle. Description of the drawings:
[0052] Figure 1 It is a schematic diagram of the control system structure of the present invention. Specific implementation method:
[0053] The present invention is described in detail below with reference to the accompanying drawings and embodiments.
[0054] Figure 1 It is a system structure diagram of a non-cooperative game collision avoidance control method driven by constraint optimization of the present invention. The method includes a path planning module 1, a constrained non-cooperative game controller 2, and a vehicle model 3; the path planning module 1 obtains the driver's expected path and the controller's expected path respectively according to road information through driver path planning and controller path planning; the constrained non-cooperative game controller 2 predicts and outputs by constructing an interactive prediction model to obtain the lateral displacement and yaw angle of the vehicle, and on this basis, a constrained optimization problem is constructed by combining the driver model and the controller's expected path as well as the actual vehicle state information output by the vehicle model, and a secondary planning solution is performed to determine the optimal control action, including the driver's front wheel angle and the controller's front wheel angle, which are input to the vehicle model to control the car to avoid collision safely; the vehicle model 3 outputs the actual state quantity of the car, including the car's lateral speed, yaw angular velocity, lateral displacement and yaw angle.
[0055] The design of the path planning module 1 includes the following parts: 1.1 controller path planning; 1.2 driver path planning;
[0056] In Section 1.1, the path planning of the controller is solved by the fifth-order polynomial method to obtain the lateral displacement y and yaw angle Regarding the trajectory curve of the longitudinal displacement X, the expression is as follows:
[0057] Y(X)=C0+C1X+C2X 2 +C3X 3 +C4X 4 +C5X 5 (1)
[0058] Among them, C0~C5 are unknown coefficients; X is the longitudinal displacement; Y(X) is the lateral displacement;
[0059] In Section 1.2, the driver’s path planning is performed on the Prescan simulation platform and co-simulated with the physical Logitech G29 steering wheel to make the simulated driving environment more realistic. Through this system, the driver’s reference lateral displacement and yaw angle can be determined to obtain the driver’s desired path;
[0060] The constrained non-cooperative game module 2 includes the following parts: 2.1 Establishing the interactive prediction model between the driver and the controller; 2.2 Constructing the constrained optimization problem; 2.3 Performing quadratic programming solution;
[0061] In Section 2.1, the interactive prediction model between the driver model and the automatic controller is established, including the following parts: 2.1.1 Establishment of the vehicle dynamics model; 2.1.2 Establishment of the state space equation; 2.1.3 Construction of the driver and controller discretization models; 2.1.4 Interactive model prediction;
[0062] In Section 2.1.1, the vehicle dynamics model is constructed, and the expression is as follows:
[0063]
[0064] Among them, v x is the longitudinal velocity of the vehicle, v y is the lateral velocity of the vehicle in the direction of the vehicle's transverse axis, m is the vehicle's mass, ω is the vehicle's yaw rate, is the lateral acceleration of the vehicle, C αf and C αr are the cornering stiffness of the front and rear tires of the vehicle, δ is the front wheel turning angle, l f and l r are the distances from the vehicle's center of mass to the front and rear axles, I z is the yaw moment of inertia about the vertical axis of the vehicle's center of mass, is the vehicle yaw angular acceleration, is the vehicle yaw angle;
[0065] In Section 2.1.2, the dynamic model in formula (1) is converted into a state space equation as follows:
[0066]
[0067] ε is the state variable of the vehicle, including lateral velocity, yaw rate, lateral displacement and yaw angle, A and B are coefficient matrices, and u is the control variable δ;
[0068] in,
[0069]
[0070] Where G is the overall steering gear ratio;
[0071] In Section 2.1.3, a discretization model is constructed to discretize the continuous-time model (3). Based on the discrete results, the driver model and the controller interaction control model are constructed. The expression is as follows:
[0072]
[0073] Among them, δ driver is the control input action of the driver model, δ controlis the control input action of the controller, ε is the state variable of the vehicle, is the output matrix, A v , B1, B2, C v is the coefficient matrix;
[0074] in,
[0075]
[0076] In Section 2.1.4, interactive model prediction is performed to calculate the predicted output of the system. According to the basic principle of model predictive control, the predicted output at time t can be obtained as:
[0077] Y(t)=ψε(t)+Θ d U driver (k)+Θ c U control (k) (5)
[0078] in,
[0079]
[0080]
[0081] Y(t) is the control output matrix, U driver (k) and U control (k) are the control input matrices of the driver and controller, ε(t) is the state matrix, ψ, Θ d and θ c is the coefficient matrix;
[0082] In Section 2.2, the constrained game controller constructs a constrained optimization problem for the predicted output, including the following parts: 2.2.1 Design of objective function; 2.2.2 Design of constraint conditions;
[0083] In Section 2.2.1, the objective function is designed to predict the motion trajectory Y(k) of the output vehicle according to the interactive prediction model. On this basis, combined with the driver's expected path R d (k) and the controller expected path R c (k), and the state information fed back by the vehicle to establish the objective function, which is expressed as follows:
[0084]
[0085] In Section 2.2.2, the constraint design considers the control limits of the vehicle's physical actuators and the control input U driver (k) and the control input U of the controller control(k) is constrained. Considering the safety of human-machine co-driving, constraints are imposed on the vehicle control output Y(k), and its expression is as follows:
[0086]
[0087] in,
[0088]
[0089]
[0090]
[0091] Among them, Q d and Q c are the weights assigned to the driver and the controller, including the lateral displacement weight and the yaw angle weight, R dl and R cl are the front wheel steering weights of the driver and controller, R d (k) and R c (k) are the desired paths of the driver and controller, respectively, U driver (k) and U control (k) are the control variables of the driver and the controller, respectively, U dmin (k) and U dmax (k) represents the set of minimum and maximum values of the driver's control quantity in the control time domain, U cmin (k) and U cmax (k) represents the set of minimum and maximum values of the controller control quantity in the control domain, Y min (k) and Y max (k) is the set of minimum and maximum values of the control output;
[0092] In Section 2.3, the quadratic programming solution is used to convert the constrained optimization problem into a QP problem, and the optimal control action is obtained by solving the problem with the solver. The specific process is as follows;
[0093] Define the matrix:
[0094]
[0095] Among them, G1 and G2 are the tracking errors of the driver and controller respectively, Z is the coefficient matrix, R is the weight matrix, and U is the control amount;
[0096] Establish a quadratic programming mathematical model:
[0097]
[0098] Convert the objective function into quadratic programming form:
[0099]
[0100] in,
[0101]
[0102] In the constraint matrix,
[0103]
[0104] The optimal control sequence U is solved by the quadratic programming method, and the first and Nth M+1 The combination of the first element of the driver control input and the first element of the controller control input is the optimal control action, and the optimal control action instruction is input to the vehicle model to achieve safe collision avoidance control.
Claims
1. A non-cooperative game collision avoidance control method driven by constraint optimization, characterized in that: The method comprises a path planning module, a constrained non-cooperative game controller and a vehicle model; the path planning module obtains the driver's expected path and the controller's expected path respectively according to the road information through the driver's path planning and the controller's path planning, and inputs the expected path into the constrained non-cooperative game controller; wherein, the method comprises an interactive prediction model, constrained optimization problem construction and quadratic planning solution, firstly, an interactive prediction model is constructed by combining the two-degree-of-freedom vehicle model, the driver and the control actions of the controller, and the lateral displacement and yaw angle of the vehicle are predicted and outputted; on this basis, the actual vehicle state information is outputted by combining the driver's expected path and the controller's expected path, as well as the vehicle model, including the lateral speed v of the vehicle y , yaw angular velocity ω r , lateral displacement y and yaw angle Construct a constrained optimization problem and determine the optimal control action by solving the quadratic programming problem, including the driver's front wheel steering angle δ' d and the controller's front wheel steering angle δ c ', input to the vehicle model, and control the vehicle to achieve collision avoidance control.
2. A non-cooperative game collision avoidance control method driven by constraint optimization as claimed in claim 1, wherein the interactive prediction model is characterized in that: Including the establishment of vehicle dynamics model and interactive prediction model; (1) Construct a vehicle dynamics model, which is expressed as follows: Among them, v x is the longitudinal velocity of the vehicle, v y is the lateral velocity of the vehicle in the direction of the vehicle's transverse axis, m is the vehicle's mass, ω is the vehicle's yaw rate, y is the vehicle's lateral acceleration, C αf and C αr are the cornering stiffness of the front and rear tires of the vehicle, δ is the front wheel turning angle, l f and l r are the distances from the vehicle's center of mass to the front and rear axles, I z is the yaw moment of inertia about the vertical axis of the vehicle's center of mass, is the vehicle yaw angular acceleration, is the vehicle yaw angle; The vehicle dynamics model (1) is converted into the state space equation as follows: ε is the state variable of the vehicle, including lateral velocity, yaw rate, lateral displacement and yaw angle, A and B are coefficient matrices, and u is the control variable δ; in, Where G is the overall steering gear ratio; (2) Establishment of interactive prediction model: The state space equation (2) is discretized, and the interactive control model between the driver model and the controller is constructed based on the discrete results. The expression is as follows: Among them, δ driver is the control input action of the driver model, δ control is the control input action of the controller, is the output matrix, A v , B1, B2, C v is the coefficient matrix; The predicted output of the calculation system can be obtained according to the basic principle of model predictive control as follows: Y(t)=ψε(t)+Θ d AT driver (k)+Θ c AT control (k) (5) Y(t) is the control output matrix, U driver (k) and U control (k) are the control input matrices of the driver and controller, ε(t) is the state matrix, ψ, Θ d and θ c is the coefficient matrix.
3. A non-cooperative game collision avoidance control method driven by constraint optimization as claimed in claim 1, wherein the constrained optimization problem is constructed, characterized in that: Including objective function design and constraint condition design; predict the output vehicle motion trajectory Y(k) according to the interactive prediction model, and on this basis, combine the driver and controller expected path R d (k), R c (k), and the state information fed back by the vehicle to establish the objective function: Considering the control limits of the vehicle's physical actuators, the control input U driver (k) and the control input U of the controller control (k) Constraints are imposed on the vehicle’s control output Y(k) considering the safety of human-machine co-driving; Among them, Q d and Q c are the weights assigned to the driver and the controller, including the lateral displacement weight and the yaw angle weight, R dl and R cl are the front wheel steering weights of the driver and controller, R d (k) and R c (k) are the desired paths of the driver and controller, respectively, U driver (k) and U control (k) are the control variables of the driver and the controller, respectively, U dmin (k) and U dmax (k) represents the set of minimum and maximum values of the driver's control quantity in the control time domain, U cmin (k) and U cmax (k) represents the set of minimum and maximum values of the controller control quantity in the control domain, Y min (k) and Y max (k) is the set of minimum and maximum values of the control output.
4. A non-cooperative game collision avoidance control method driven by constraint optimization as claimed in claim 1, wherein the quadratic programming solution is characterized in that: Convert the constrained optimization problem into a QP solving problem and obtain the optimal control action through the solver; The objective function is as follows: Define the matrix: Among them, G1 and G2 are the tracking errors of the driver and controller respectively, Z is the coefficient matrix, R is the weight matrix, and U is the control amount; Construct a quadratic programming mathematical model: Convert the objective function into quadratic programming form: in, Constraint Matrix The optimal control sequence U is solved by the quadratic programming method, and the first and Nth M+1 The combination of the first element of the driver control input and the first element of the controller control input is the optimal control action, and the optimal control action instruction is input to the vehicle model to achieve safe collision avoidance control.
Citation Information
Patent Citations
Intelligent driving system with autonomous lane changing function and capable of improving lateral safety
CN110356404A
Extended-range distributed drive electric vehicle collision avoidance optimization control system and method
CN112109705A
Intelligent automobile driving right distribution method based on adaptive weight game
CN116443017A
Man-machine cooperation path tracking control method considering driving state of driver
CN116449815A
Method of Adaptive Dynamic Model-base]d Optimal Path Planning for Autonomous Navigation of Unmanned Ground Vehicle and Appratus thereof
KR101897407B1
Cited By
Vehicle bending control method based on lateral acceleration prediction
CN120363900A
Cooperative steering control method considering man-machine mutual reliability
CN121386428A
A cooperative steering control method considering human-machine trust degree
CN121386428B
