Automatic driving vehicle lane changing method and device based on multi-objective optimization and medium

Abstract

The invention discloses an automatic driving vehicle lane changing method based on multi-objective optimization, and the method comprises the steps: collecting the surrounding environment information in the driving process of an automatic driving vehicle in real time, constructing an environment model, providing a real-time lane changing decision, and providing a constraint condition for track planning; when the environment model provides a lane changing decision, adopting a quintic polynomial to model a lane changing trajectory to obtain a trajectory model; by taking safety, efficiency and comfort as optimization targets, calculating optimal parameters of the trajectory model by adopting adaptive particle swarm optimization to obtain an optimized trajectory; the optimized track is executed by a vehicle drive-by-wire system, and stable lane changing is ensured. According to the method, multiple targets are optimized, so that the vehicle can better meet various requirements during lane changing, the driving experience and safety of the automatic driving vehicle are further improved, and a better solution is provided for stable lane changing of the vehicle under complex road conditions.

Description

Technical Field

[0001] The present invention belongs to the technical field of autonomous driving, and relates to a lane-changing method, device and medium for autonomous driving vehicles based on multi-objective optimization. Background Art

[0002] Most of the existing lane-changing algorithms for autonomous driving vehicles are based on single-objective optimization, such as giving priority to the shortest time or the shortest path. However, these algorithms often have difficulty taking into account the comfort and safety of the vehicle during the lane-changing process, and are prone to cause adverse experiences such as sudden deceleration or acceleration during the lane-changing process, affecting the comfort of passengers and the safety of the overall driving. Therefore, the lane-changing method based on single-objective optimization has obvious limitations in a changing driving environment and cannot provide sufficient adaptability in different traffic scenarios. Summary of the Invention

[0003] Object of the Invention: The object of the present invention is to provide a lane-changing method, device and medium for autonomous driving vehicles based on multi-objective optimization, and solve the problems in the existing lane-changing algorithms for autonomous driving vehicles, such as difficulty in taking into account the comfort and safety of the vehicle during the lane-changing process, prone to cause adverse experiences such as sudden deceleration or acceleration during the lane-changing process, affecting the comfort of passengers and the safety of the overall driving.

[0004] Technical Solution: A lane-changing method for autonomous driving vehicles based on multi-objective optimization of the present invention includes the following steps:

[0005] S1. Real-time collect the surrounding environment information during the driving of the autonomous driving vehicle, and construct an environment model for providing real-time lane-changing decisions and providing constraint conditions for trajectory planning;

[0006] S2. When the environment model provides a lane-changing decision, use a fifth-order polynomial to model the lane-changing trajectory to obtain a trajectory model;

[0007] S3. Taking safety, efficiency and comfort as optimization objectives, use an adaptive particle swarm optimization algorithm to calculate the optimal parameters of the trajectory model to obtain an optimized trajectory; the safety objective is the reciprocal of the collision time with adjacent vehicles during the lane-changing process, the efficiency objective is the average speed during the lane-changing process, and the comfort objective is the jerk;

[0008] S4. Trajectory execution: The optimized trajectory is executed by the vehicle-by-wire system to ensure smooth lane-changing.

[0009] Further, in the step S1, the environment information is obtained through a millimeter-wave radar, a lidar and a camera.

[0010] Further, the environment information in the step S1 includes the surrounding vehicle state information, road information and dynamic traffic condition information.

[0011] Further, in the step S1, the surrounding vehicle state includes the position, speed, and acceleration of the surrounding vehicles; the road information includes lane lines, speed limits, and traffic signs; and the dynamic traffic condition information includes congestion, pedestrians, and accidents.

[0012] Further, in step S2, the expression of the trajectory model is as follows:

[0013] y(t) = b0 + b1t + b2t 2 + b3t 3 + b4t 4 + b5t 5 ;

[0014] x(t) = a0 + a1t + a2t 2 + a3t 3 + a4t 4 + a5t 5 ;

[0015] Among them, the coefficients y(t) and x(t) respectively represent the trajectories of the vehicle's position changing with time t in the lateral and longitudinal lane-changing directions; the coefficients b and a are the coefficients of the trajectory equation, which determine the shape of the trajectory.

[0016] According to the starting state and target state of the vehicle, the following boundary conditions are set:

[0017] Starting point (t = 0): Position y(0) = y0, speed Acceleration

[0018] Starting point (t = 0): Position x(0) = x0, speed Acceleration

[0019] End point (t = T): Position y(T) = y f , speed Acceleration

[0020] End point (t = T): Position x(T) = x f , speed Acceleration

[0021] Among them, y f , y0 are respectively the starting and target lateral positions of the lane change; v y0 , v yf are respectively the starting and target lateral speeds; a y0 , a yf are respectively the starting and target lateral accelerations; x f , x0 are respectively the starting and target longitudinal positions of the lane change; vx0 and v xf are the starting and target longitudinal speeds respectively; a x0 and a xf are the starting and target lateral accelerations respectively; T is the total lane-changing time.

[0022] Furthermore, the construction method with safety, efficiency, and comfort as the optimization objectives in step S3 includes:

[0023] S3.1. Safety objective: The TTC value is the collision time with the adjacent vehicle during the lane-changing process, S is the reciprocal of the TTC value, and minimizing S results in a safer lane-changing process. The expression is:

[0024]

[0025] where d and d a are the current positions of the host vehicle and the adjacent vehicle respectively; v and v a are the speeds of the host vehicle and the adjacent vehicle respectively;

[0026] S3.2. Comfort objective: J is the jerk, and minimizing J reduces the vehicle's swaying and improves passenger comfort. The expression is:

[0027]

[0028] where T is the time required for lane-changing; y(t) is the lateral displacement of the vehicle; represents the change rate of the lateral acceleration, and the larger the value, the greater the impact on the vehicle;

[0029] S3.3. Efficiency objective: E is the average speed during the lane-changing process, and maximizing E can improve the lane-changing efficiency. The expression is:

[0030]

[0031] where T is the total time required for lane-changing; v(t) is the instantaneous longitudinal speed of the vehicle at time t;

[0032] S3.4. Combining the above objectives, the weighted summation method is used to construct the comprehensive objective function:

[0033] F(X) = ω1 * S(X) + ω2J(X) + ω3 * E(X);

[0034]

[0035] Among them, ω1, ω2, and ω3 are weight coefficients, satisfying ω1 + ω2 + ω3 = 1; S measures safety, and a smaller value is better; J measures comfort, and a smaller value is better; E measures efficiency, and a larger value is better; X represents the lane-changing trajectory parameter vector; F represents the comprehensive objective function of the lane-changing process, and the more excellent the value, the closer it is to the ideal lane-changing scheme.

[0036] Further, the calculation method of the optimal parameters in step S3 is specifically as follows:

[0037] S3.5. Each particle X i represents a lane-changing scheme, and its solution vector is defined as:

[0038] X i = (x0, y0, x f , y f , T, b0, b1, b2, b3, b4, b5, a0, a1, a2, a3, a4, a5);

[0039] Among them, x0 and y0 are the starting points of lane-changing; x f , y f are the ending points of lane-changing; T is the lane-changing time; b0, b1, b2, b3, b4, b5 are the parameters of the lateral trajectory model; a0, a1, a2, a3, a4, a5 are the parameters of the longitudinal trajectory model; during initialization, each particle randomly generates a solution vector within a preset range;

[0040] S3.6. The inertia weight ω is adaptively adjusted with iteration, and the expression is:

[0041] Among them, t is the current iteration number, and N is the maximum iteration number; ω max and ω min are the maximum and minimum inertia weights respectively;

[0042] S3.7. Set the particle swarm size, and randomly initialize the trajectory parameters of each particle; set the particle velocity, and initialize the individual optimal solution and the global optimal solution;

[0043] S3.8. Calculate the fitness: evaluate the safety, comfort, and efficiency objective function values of each trajectory; calculate the comprehensive fitness function, and update the individual optimal solution and the global optimal solution of each particle;

[0044] Update the particle position and velocity: dynamically adjust the trajectory parameters of the particle according to the inertia weight, the individual optimal solution, and the global optimal solution;

[0045]

[0046] Among them: c1 and c2 are learning factors; r1 and r2 are random numbers between [0, 1]; is the position of the particle in the t-th generation, i.e., the trajectory parameter; is the velocity of the particle; P i t is the historical optimal position of the particle; G t The global optimal position, i.e., the optimal trajectory parameter found currently;

[0047] S3.9. Set the termination condition. When the fitness value change of the global optimal solution G t in several consecutive generations is less than the preset error threshold ε, it is considered that the optimization process has converged. The expression is:

[0048] |F(G t+1 ) - F(G t )| < ε,

[0049] If the termination condition is satisfied, output the current optimal trajectory; otherwise, continue to iterate until the maximum number of iterations N is reached.

[0050] Furthermore, the specific steps of step S4 include the following steps:

[0051] S4.1. Trajectory tracking control: Adopt PID control to achieve steering angle, acceleration, and braking control;

[0052] S4.2. Trajectory execution constraints: Lateral error constraint: Ensure that the vehicle stays on the planned trajectory; Longitudinal speed constraint: The lane-changing speed should be within the safe range; Acceleration constraint: Prevent sudden acceleration or deceleration;

[0053] S4.3. Abnormal handling mechanism: If a sudden situation is detected during the execution process, including sudden braking of the vehicle in front, approaching of a vehicle on the side, the lane change will be immediately terminated, the vehicle will return to the original lane, and the lane-changing opportunity will be re-evaluated.

[0054] The present invention provides a lane-changing device for an autonomous vehicle based on multi-objective optimization, including:

[0055] At least one processor;

[0056] At least one memory for storing at least one program;

[0057] When the at least one program is executed by the at least one processor, the at least one processor implements the above-mentioned lane-changing method for an autonomous vehicle based on multi-objective optimization.

[0058] The present invention provides a computer-readable storage medium, in which a program executable by a processor is stored. The program executable by the processor is used to execute the above-mentioned lane-changing method for an autonomous vehicle based on multi-objective optimization when executed by the processor.

[0059] Beneficial effects: Compared with the prior art, the present invention has the following remarkable advantages: 1. Multi-objective optimization: Simultaneously optimize safety, comfort, and efficiency to ensure the quality of lane change.

[0060] 2. Adaptive optimization: Dynamically adjust the optimization strategy for different scenarios to improve adaptability.

[0061] 3. Efficient solution: Adopt Adaptive Particle Swarm Optimization (APSO) to improve the calculation efficiency and meet the real-time requirements.

[0062] 4. Precise control; It can be widely applied to automatic driving scenarios such as highways, urban roads, and driverless logistics vehicles, improve the intelligent level of lane change decision-making and execution, and enhance the overall traffic efficiency and safety. Brief description of the drawings

[0063] Figure 1 It is a flow chart of the present invention.

[0064] Figure 2 It is a comparison chart of the comprehensive objective function values of two algorithms, PSO and APSO, applied to lane change trajectory optimization. Detailed implementation manners

[0065] The technical solution of the present invention will be further described below with reference to the drawings.

[0066] As Figure 1 shown, the automatic driving lane change method based on multi-objective optimization in this embodiment senses the surrounding environment through lidar, continuously collects environmental data, generates a lane change intention, and uses the methods of multi-objective optimization algorithm and vehicle dynamics model to establish a lane change mechanism with three major objectives of comfort, safety, and efficiency, so that the automatic driving vehicle can successfully change lanes. The method includes the following steps:

[0067] Step 1: Environmental perception and data processing

[0068] The automatic driving vehicle obtains real-time surrounding environment information through sensors such as millimeter-wave radar, lidar, camera, GPS / IMU, etc., and performs data fusion to construct a traffic environment model, provide real-time lane change decision support, and provide constraint conditions for trajectory planning.

[0069] The surrounding environment information includes the surrounding vehicle status information, road information, and dynamic traffic situation information. The surrounding vehicle status: such as the position, speed, acceleration, etc. of the surrounding vehicles; road information: such as lane lines, speed limits, traffic signs; dynamic traffic information: such as congestion, pedestrians, accidents, etc.

[0070] The specific sensor data acquisition methods are as follows: Millimeter-wave radar: Detect the positions, speeds, and accelerations of the vehicles in front and behind. LiDAR: Obtain lane line and obstacle information to ensure the safety of lane changes. Camera: Identify traffic signs, signal lights, and lane lines. GPS / IMU: Provide high-precision position information and vehicle motion states.

[0071] The specific sensor data fusion method is as follows: Use Kalman filtering for data fusion to improve accuracy. Generate a local environment map for trajectory planning and optimization calculations.

[0072] Step 2: Trajectory planning

[0073] When the environmental model provides a lane change decision, use a fifth-degree polynomial to model the lane change trajectory to obtain a trajectory model, ensuring the smoothness and executability of the lane change trajectory;

[0074] 2.1 Trajectory model establishment The trajectory equations are as follows:

[0075] y(t) = b0 + b1t + b2t 2 + b3t 3 + b4t 4 + b5t 5 ;

[0076] x(t) = a0 + a1t + a2t 2 + a3t 3 + a4t 4 + a5t 5 ;

[0077] Among them, the coefficients y(t) and x(t) represent the trajectories of the vehicle's position changing with time t in the lateral and longitudinal lane change directions respectively; the coefficients b and a are the coefficients of the trajectory equations, which determine the shape of the trajectory.

[0078] 2.2 Trajectory boundary conditions

[0079] According to the vehicle's starting state and target state, set the following boundary conditions:

[0080] Starting point (t = 0): Position y(0) = y0, speed Acceleration

[0081] Starting point (t = 0): Position x(0) = x0, speed Acceleration

[0082] End point (t = T): Position y(T) = y f , speed Acceleration

[0083] Termination point (t = T): Position x(T) = x f , velocity acceleration

[0084] where y f , y0 are the starting and target lateral positions of lane change respectively; v y0 , v yf are the starting and target lateral velocities respectively; a y0 , a yf are the starting and target lateral accelerations respectively; x f , x0 are the starting and target longitudinal positions of lane change respectively; v x0 , v xf are the starting and target longitudinal velocities respectively; a x0 , a xf are the starting and target lateral accelerations respectively; T is the total lane change time.

[0085] Step 3: Multi-objective optimization modeling and solution

[0086] Taking safety, efficiency and comfort as optimization objectives, the optimal parameters of the trajectory model are calculated using the adaptive particle swarm optimization algorithm to obtain the optimized trajectory:

[0087] 3.1 Safety objective

[0088]

[0089] where d, d a are the current positions of the host vehicle and the adjacent vehicle respectively; v, v a are the velocities of the host vehicle and the adjacent vehicle respectively.

[0090] Generally, we hope to maximize TTC to ensure safety. However, to construct a minimization optimization objective function, we usually use its reciprocal for minimization.

[0091]

[0092] A lower value of S means a safer lane change process.

[0093] 3.2 Comfort objective

[0094] Minimize the jerk:

[0095]

[0096] where J represents the squared integral of jerk, measuring the smoothness of the trajectory; T is the time required for lane change; y(t) is the lateral displacement of the vehicle; Represents the rate of change of lateral acceleration. The larger the value, the greater the impact on the vehicle.

[0097] Minimizing J can reduce vehicle jitter and improve passenger comfort.

[0098] 3.3 Efficiency objective

[0099]

[0100] Where, E is the average speed during the lane change process; T is the total time required for the lane change; v(t) is the instantaneous longitudinal speed of the vehicle at time t.

[0101] Maximizing E can improve the lane change efficiency.

[0102] 3.4 Combining the above objectives, a weighted summation method is used to construct a comprehensive objective function:

[0103] F(X) = ω1 * S(X) + ω2J(X) + ω3 * E(X);

[0104]

[0105]

[0106] Where, ω1, ω2, ω3 are weight coefficients, satisfying (ω1 + ω2 + ω3 = 1); S measures safety, and a smaller value is better; J measures comfort, and a smaller value is better; E measures efficiency, and a larger value is better; X represents the lane change trajectory parameter vector; F represents the comprehensive objective function of the lane change process, and the closer the value is to the ideal lane change scheme, the better.

[0107] 3.5 An adaptive particle swarm optimization algorithm (APSO) is used to calculate the optimal parameters of the vehicle lane change trajectory:

[0108] Each particle X i represents a lane change scheme, and its solution vector is defined as:

[0109] X i = (x0, y0, x f , y f , T, b0, b1, b2, b3, b4, b5, a0, a1, a2, a3, a4, a5);

[0110] Where, x0, y0 are the starting points of the lane change; x f , y f are the ending points of the lane change; T is the lane change time; b0, b1, b2, b3, b4, b5 are the parameters of the lateral trajectory model; a0, a1, a2, a3, a4, a5 are the parameters of the longitudinal trajectory model; at initialization, each particle randomly generates a solution vector within a preset range;

[0111] 3.6 To improve the convergence, the inertia weight ω is adaptively adjusted with the iteration, and the expression is:

[0112]

[0113] where t is the current iteration number and N is the maximum iteration number; ω max and ω min are the maximum and minimum inertia weights respectively;

[0114] 3.7 Initialization

[0115] Set the particle swarm size, and randomly initialize the trajectory parameters of each particle; set the particle velocity, and initialize the individual best solution and the global best solution.

[0116] 3.8 Particle update rule

[0117] Calculate the fitness: evaluate the objective function values of safety, comfort, and efficiency of each trajectory; calculate the comprehensive fitness function, and update the individual best solution and the global best solution of each particle

[0118] Update the particle position and velocity: dynamically adjust the trajectory parameters of the particle according to the inertia factor, the individual best solution, and the global best solution:

[0119]

[0120] where: c1 and c2 are learning factors; r1 and r2 are random numbers between [0, 1]; is the position of the particle in the t-th generation, i.e., the trajectory parameter; is the velocity of the particle; P i t is the historical best position of the particle; G t The global best position is the currently found optimal trajectory parameter.

[0121] 3.9 Set the termination condition. When the change in the fitness value of the global best solution G t is less than the preset error threshold ε for several consecutive generations, it is considered that the optimization process has converged, and the expression is:

[0122] |F(G t+1 ) - F(G t )| < ε;

[0123] If the above termination condition is satisfied, output the current optimal trajectory; otherwise, continue the iteration until the maximum iteration number N is reached.

[0124] Such as Figure 2As shown in the figure, the comprehensive objective function value of the APSO algorithm is smaller than that of the PSO algorithm. According to the definition of the weight coefficient, the smaller the comprehensive objective function value, the better the lane-changing plan. This indicates that during the lane-changing process, the APSO is a better lane-changing plan.

[0125] Step 4: Trajectory execution

[0126] The optimized trajectory is executed by the vehicle's by-wire system to ensure smooth lane change.

[0127] 4.1 Trajectory tracking control

[0128] PID control is adopted to achieve steering angle, acceleration, and braking control.

[0129] 4.2 Trajectory execution constraints

[0130] Lateral error constraint: Ensure that the vehicle stays on the planned trajectory; acceleration constraint: Prevent sudden acceleration or deceleration.

[0131] 4.3 Abnormal handling mechanism

[0132] If sudden situations (such as the vehicle in front suddenly braking or a vehicle approaching from the side) are detected during the execution process, the lane change will be immediately terminated, the vehicle will return to the original lane, and the lane change opportunity will be re-evaluated.

[0133] The embodiments of the present invention aim to elaborate its technical principles and application methods so that those skilled in the art can understand and implement them. However, the protection scope of the present invention is not limited to the above specific embodiments. Any equivalent changes, improvements, or extensions made on the basis of the core idea of the present invention should be covered within the protection scope of the present invention.

Claims

1. A lane-changing method for an autonomous vehicle based on multi-objective optimization, characterized in that, It includes the following steps: S1. Collect the surrounding environment information during the driving of the autonomous vehicle in real time, construct an environment model, which is used to provide real-time lane-changing decisions and provide constraint conditions for trajectory planning; S2. When the environment model provides a lane-changing decision, use a fifth-degree polynomial to model the lane-changing trajectory to obtain a trajectory model; S3. Taking safety, efficiency, and comfort as optimization objectives, use an adaptive particle swarm optimization algorithm to calculate the optimal parameters of the trajectory model to obtain an optimized trajectory; the safety objective is the reciprocal of the time to collision with adjacent vehicles during the lane-changing process, the efficiency objective is the average speed during the lane-changing process, and the comfort objective is the jerk; S4. Trajectory execution: The optimized trajectory is executed by the vehicle-by-wire system to ensure a smooth lane change.

2. The lane-changing method for an autonomous driving vehicle based on multi-objective optimization according to claim 1, wherein In step S1, the environment information is obtained through millimeter-wave radar, lidar, and cameras.

3. The lane-changing method for an autonomous driving vehicle based on multi-objective optimization according to claim 1, wherein The environment information in step S1 includes the surrounding vehicle state information, road information, and dynamic traffic condition information.

4. The lane-changing method for an autonomous vehicle based on multi-objective optimization according to claim 3, characterized in that In step S1, the surrounding vehicle state includes the position, speed, and acceleration of the surrounding vehicles; the road information includes lane lines, speed limits, and traffic signs; the dynamic traffic condition information includes congestion, pedestrians, and accidents.

5. The lane-changing method for an autonomous driving vehicle based on multi-objective optimization according to claim 1, wherein In step S2, the expression of the trajectory model is as follows: y(t) = b0 + b1t + b2t 2 + b3t 3 + b4t 4 + b5t 5 ; x(t) = a0 + a1t + a2t 2 + a3t 3 + a4t 4 + a5t 5 ; Among them, the coefficients y(t) and x(t) respectively represent the trajectories of the vehicle's position changing with time t in the lateral and longitudinal lane-changing directions; the coefficients b and a are the coefficients of the trajectory equation, which determine the shape of the trajectory; According to the starting state and target state of the vehicle, set the following boundary conditions: Starting point (t = 0): position y(0) = y0, velocity Acceleration Starting point (t = 0): position x(0) = x0, velocity acceleration Termination point (t = T): Position y(T) = y f , velocity acceleration End point (t = T): position x(T) = x f , velocity acceleration Among them, y f , y0 are the starting and target lateral positions for lane change respectively; v y0 , v yf are the starting and target lateral speeds respectively; a y0 , a yf are the starting and target lateral accelerations respectively; x f , x0 are the starting and target longitudinal positions for lane change respectively; v x0 , v xf are the starting and target longitudinal speeds respectively; a x0 , a xf are the starting and target longitudinal accelerations respectively; T is the total lane change time.

6. The lane-changing method for an autonomous driving vehicle based on multi-objective optimization according to claim 1, wherein The construction method taking safety, efficiency, and comfort as optimization objectives in step S3 includes: S3.

1. Safety objective, the TTC value is the time to collision with adjacent vehicles during the lane-changing process, S is the reciprocal of the TTC value, and minimize S to obtain a safer lane-changing process. The expression is: where d and d a are the current positions of the host vehicle and the adjacent vehicle respectively; v and v a are the speeds of the host vehicle and the adjacent vehicle respectively. S3.

2. Comfort objective, J is the jerk, and minimize J to reduce the vehicle's shaking and improve passenger comfort. The expression is: Among them, T is the time required for lane change; y(t) is the lateral displacement of the vehicle; represents the change rate of lateral acceleration. The larger the value, the greater the impact on the vehicle; S3.

3. Efficiency objective, E is the average speed during the lane-changing process, and maximize E to improve the lane-changing efficiency. The expression is: Among them, T is the total time required for lane change; v(t) is the instantaneous longitudinal speed of the vehicle at time t; S3.

4. Combining the above objectives, use the weighted summation method to construct a comprehensive objective function: F(X) = ω1*S(X) + ω2J(X) + ω3*E(X); Among them, ω1, ω2, and ω3 are weight coefficients, satisfying ω1 + ω2 + ω3 = 1; S measures safety, and a smaller value is better; J measures comfort, and a smaller value is better; E measures efficiency, and a larger value is better; X represents the lane-changing trajectory parameter vector; F represents the comprehensive objective function of the lane-changing process, and the smaller the value, the closer it is to the ideal lane-changing scheme.

7. The lane-changing method for an autonomous driving vehicle based on multi-objective optimization according to claim 1, wherein The calculation method of the optimal parameters in step S3 is specifically as follows: S3.5, each particle X i represents a lane-changing plan, and its solution vector is defined as: X i =(x0,y0,x f ,y f ,T,b0,b1,b2,b3,b4,b5,a0,a1,a2,a3,a4,a5); Among them, x0 and y0 are the starting points of lane change; x f , y f are the ending points of lane change; T is the lane change time; b0, b1, b2, b3, b4, b5 are the parameters of the lateral trajectory model; a0, a1, a2, a3, a4, a5 are the parameters of the longitudinal trajectory model; during initialization, each particle randomly generates a solution vector within a preset range; S3.

6. The inertial weight ω is adaptively adjusted with iteration, and the expression is: where t is the current iteration number and N is the maximum number of iterations; ω max and ω min are the maximum and minimum inertia weights, respectively; S3.

7. Set the particle swarm size, and randomly initialize the trajectory parameters of each particle; set the particle velocity, and initialize the individual optimal solution and the global optimal solution; S3.

8. Calculate fitness: Evaluate the objective function values of safety, comfort, and efficiency for each trajectory; calculate the comprehensive fitness function, and update the individual optimal solution and the global optimal solution of each particle. Update particle position and velocity: Dynamically adjust the trajectory parameters of the particle according to the inertia weight, individual optimal solution, and global optimal solution. Where: c1 and c2 are learning factors; r1 and r2 are random numbers between [0, 1]; is the position of the particle at the t-th generation, i.e., the trajectory parameter; is the velocity of the particle; is the historical optimal position of the particle; G t The global optimal position, i.e., the optimal trajectory parameter found currently; S3.

9. Set the termination condition. When the change in the fitness value of the global optimal solution G for several consecutive generations is less than the preset error threshold ε, it is considered that the optimization process has converged. The expression is as follows: t When the change in the fitness value of the global optimal solution G for several consecutive generations is less than the preset error threshold ε, it is considered that the optimization process has converged. The expression is as follows: |F(G t+1 ) - F(G t )| < ε, If the termination condition is satisfied, output the current optimal trajectory; otherwise, continue to iterate until the maximum number of iterations N is reached.

8. The lane-changing method for an autonomous driving vehicle based on multi-objective optimization according to claim 1, wherein The specific steps of step S4 include the following steps: S4.

1. Trajectory tracking control: Use PID control to achieve steering angle, acceleration, and braking control. S4.

2. Trajectory execution constraints: Lateral error constraint: Ensure that the vehicle stays on the planned trajectory; longitudinal speed constraint: The lane-changing speed should be within the safe range; acceleration constraint: Prevent sudden acceleration or deceleration. S4.

3. Abnormal handling mechanism: If sudden situations are detected during the execution process, including sudden braking of the vehicle ahead and approaching of vehicles on the side, the lane change will be immediately terminated, the vehicle will return to the original lane, and the lane-changing opportunity will be re-evaluated.

9. A lane-changing device for an autonomous vehicle based on multi-objective optimization, characterized in that, Including: At least one processor; At least one memory for storing at least one program; When the at least one program is executed by the at least one processor, the at least one processor implements a lane-changing method for an autonomous driving vehicle based on multi-objective optimization according to any one of claims 1-8.

10. A computer-readable storage medium storing a program executable by a processor, characterized in that, The program executable by the processor is used to execute the lane-changing method for an autonomous driving vehicle based on multi-objective optimization according to any one of claims 1-8 when executed by the processor.

Images