Automatic driving automobile track planning method and system in complex driving environment

Abstract

The invention discloses an automatic driving automobile track planning method and system in a complex driving environment, and relates to the technical field of track planning, and the method comprises the steps: obtaining automobile sensor data and automobile state information, and carrying out the fusion of the data and information to obtain multi-source information; an automobile kinematic model is constructed, the input of the automobile kinematic model is the longitudinal acceleration and the front wheel turning angle in the multi-source information, and an MPC controller based on NNPID feedback is constructed through the MPC controller and the NNPID controller based on the automobile kinematic model; and updating the automobile state through the MPC controller based on NNPID feedback, obtaining the lateral deviation of T time domains in the future through the updated automobile state, correcting the lateral deviation in real time, and obtaining the trajectory planning of the automatic driving automobile. According to the invention, through fine road and traffic environment simulation and an advanced trajectory planning algorithm, a trajectory which better accords with the actual traffic condition can be generated, and the safety and the driving efficiency of the autonomous vehicle are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of trajectory planning, and particularly to a method and system for trajectory planning of an autonomous vehicle in a complex driving environment. Background Art

[0002] In a complex driving environment, an autonomous vehicle needs to accurately perceive the surrounding environment and predict the behaviors of other traffic participants. Wang et al. proposed a multi-sensor fusion perception method based on deep learning, which can significantly improve the perception ability of an autonomous vehicle for the surrounding environment. At the same time, Lee et al. proposed a vehicle trajectory prediction algorithm based on a behavior model, which can accurately predict the driving trajectories of other vehicles.

[0003] Based on perception and prediction, an autonomous vehicle needs to plan a safe and efficient driving trajectory. Currently, researchers have proposed various trajectory planning methods. Among them, the trajectory planning method based on optimization has received extensive attention due to its high efficiency and accuracy. Chen et al. proposed a trajectory planning method based on dynamic programming, which can achieve real-time and accurate trajectory planning in a complex driving environment. In addition, the trajectory planning method based on learning has also attracted much attention due to its powerful learning ability. Xu et al. realized the trajectory planning of an autonomous vehicle using deep reinforcement learning technology. With the continuous development of deep learning technology, its application in the trajectory planning of autonomous vehicles is becoming more and more extensive. Researchers have begun to attempt to combine deep learning technology with traditional trajectory planning methods to improve the accuracy and robustness of trajectory planning. Zhang et al. proposed an end-to-end trajectory planning method based on deep learning, which can directly output the driving actions of a vehicle according to sensor data, realizing the real-time planning and control of the driving trajectory of the vehicle.

[0004] Existing MPC-based trajectory planning methods mainly focus on the kinematic and dynamic constraints of the vehicle itself and the road geometry constraints. These systems usually rely on relatively basic road modeling, simple traffic participant behavior prediction, and basic trajectory planning algorithms for path planning, and are insufficient in robustness when dealing with emergencies and complex scenarios, resulting in limitations in the overall performance and safety of the autonomous driving system. Summary of the Invention

[0005] The purpose of the present invention is to provide a method and system for trajectory planning of an autonomous vehicle in a complex driving environment in view of the above-mentioned deficiencies of the prior art, so as to solve the problems in the prior art.

[0006] The present invention specifically provides the following technical solutions:

[0007] A method for trajectory planning of an autonomous vehicle in a complex driving environment, comprising:

[0008] Obtain different sensor data of the vehicle and the vehicle state information, fuse the different sensor data of the vehicle, and construct a unified environment model;

[0009] Based on the environment model, construct a vehicle kinematic model according to the relationship between vehicle states, and construct an MPC controller based on NNPID feedback in the vehicle kinematic model through an MPC controller and an NNPID controller;

[0010] Input the longitudinal acceleration and the front wheel steering angle in the vehicle state information into the vehicle kinematic model, update the vehicle state and obtain a longitudinal control quantity for adjusting the torque of the vehicle drive motor. Based on the updated vehicle state, obtain the lateral deviation in the next T time domains through the MPC controller based on NNPID feedback, and correct the lateral deviation by adjusting the front wheel steering angle to obtain a lateral control quantity. Use the lateral control quantity and the longitudinal control quantity to perform vehicle trajectory planning.

[0011] Preferably, the vehicle kinematic model is constructed according to the relationship between vehicle states, and the specific expression of the vehicle kinematic model is:

[0012]

[0013] where, φ is the heading angle, β is the sideslip angle of the center of mass, is the yaw angle, v x is the longitudinal vehicle speed, a x is the longitudinal acceleration, δ f is the front wheel steering angle, l r and l f are the distances from the center of mass to the front and rear axles, X g and Y g are the positions of the vehicle in the earth coordinate system, and are the first-order derivatives of v x , X g and Y g respectively.

[0014] Preferably, the longitudinal acceleration and the front wheel steering angle in the vehicle state information are input into the vehicle kinematic model to update the vehicle state, where the vehicle state includes the vehicle position, the vehicle heading angle and the vehicle speed, specifically:

[0015] Update the vehicle position, and the specific expression is:

[0016] x t+Δt =x t +v t ×cos(θ t )×Δt;

[0017] yt+Δt == y t + v t × sin(θ t ) × Δt;

[0018] Wherein, x t and y t are the x and y coordinates of the vehicle at the current moment, x t+Δt and y t+Δt are the x and y coordinates of the vehicle at the next moment, v t is the vehicle speed at the current moment, θ t is the vehicle heading angle at the current moment, and Δt is the time length of one step;

[0019] Update the vehicle heading angle, and the specific expression is:

[0020] θ t+Δt = θ t + v t / l × tan(δ) × Δt;

[0021] Wherein, θ t+Δt is the vehicle heading angle at the next moment, δ is the front wheel steering angle, and l is the vehicle wheelbase;

[0022] Update the vehicle speed, and the specific expression is:

[0023] v t+Δt = max(0, v t + a * Δt);

[0024] Wherein, v t+Δt is the vehicle speed at the next moment, and a is the longitudinal acceleration.

[0025] Preferably, obtaining the lateral deviation in the next T time domains by the MPC controller based on NNPID feedback includes:

[0026] Based on the MPC controller, obtain the lateral deviation e l according to the line P connecting the current trajectory matching point and the center point of the vehicle rear axle, and the specific expression is:

[0027] e l = P · n = d x · (-sinθ) + d y · cosθ;

[0028] Wherein, the vector P = (x - x m , y - y m ) = (d x , d y ), x is the abscissa of the vehicle position, y is the ordinate of the vehicle position, x m is the reference abscissa, ym is the reference vertical coordinate, d x is the abscissa difference, d y is the ordinate difference, θ is the heading angle of the vehicle position, and n is the normal vector at the vehicle position;

[0029] n = (-sinθ, cosθ);

[0030] Take the mean value of the T time domains after each prediction of the MPC as the final feedback error input, and use the final feedback error input as the lateral deviation e in the future T time domains le , and the specific expression is:

[0031]

[0032] where i is the i-th time domain.

[0033] Preferably, when correcting the lateral deviation by adjusting the front wheel angle, it further includes:

[0034] Use a PID controller to correct the output of the MPC, and adaptively adjust the parameters of the PID controller through a neural network. The specific expression is:

[0035]

[0036] where is the input layer - output layer weight matrix, y i is the vector after PID calculation. In the controller, z is the desired error, x' is the current actual lateral deviation, and u is the lateral steering angle control amount δ output by the NNPID N_PID , and k is the independent variable.

[0037] Preferably, the environmental model includes road boundaries, obstacle positions, traffic signals, and weather conditions.

[0038] The present invention provides an automatic driving vehicle trajectory planning system under a complex driving environment, including:

[0039] An acquisition module for obtaining different sensor data and vehicle state information of the vehicle, fusing the different sensor data of the vehicle, and constructing a unified environmental model;

[0040] A model construction module for constructing a vehicle kinematic model based on the relationship between vehicle states based on the environmental model, and constructing an MPC controller based on NNPID feedback in the vehicle kinematic model through an MPC controller and an NNPID controller;

[0041] A planning module is configured to input the longitudinal acceleration and the front wheel steering angle in the vehicle state information into a vehicle kinematic model, update the vehicle state, and obtain a longitudinal control quantity for adjusting the torque of the vehicle's power motor. Based on the updated vehicle state, a lateral deviation in the next T time domains is obtained through an MPC controller based on NNPID feedback, and the lateral deviation is corrected by adjusting the front wheel steering angle to obtain a lateral control quantity. The vehicle trajectory is planned using the lateral control quantity and the longitudinal control quantity.

[0042] The present invention provides a computer device, including a memory and a processor. A program is stored in the memory, and when the program is executed by the processor, the processor is caused to execute the steps of the above-mentioned method for planning the trajectory of an autonomous vehicle in a complex driving environment.

[0043] The present invention provides a storage medium, on which a computer program is stored. When the computer program is executed by a processor, the steps of the above-mentioned method for planning the trajectory of an autonomous vehicle in a complex driving environment are implemented.

[0044] Compared with the prior art, the present invention has the following remarkable advantages:

[0045] In the present invention, a vehicle kinematic model is constructed based on the relationship of multi-source information collected, and an MPC controller based on NNPID feedback is constructed. The longitudinal acceleration and the front wheel steering angle are input into the vehicle kinematic model to simulate and update the vehicle state. The lateral deviation in the next T time domains is obtained through the MPC controller based on NNPID feedback, and the lateral deviation is corrected. The lateral control quantity of the autonomous vehicle is fed back to eliminate the inherent hysteresis of the vehicle kinematic model, enabling the autonomous vehicle to maintain stable operation in the face of complex environments and emergencies, and the generated trajectory is more in line with the actual traffic conditions, improving the safety and driving efficiency of the autonomous vehicle. Description of the Drawings

[0046] Figure 1 It is a vehicle kinematic model diagram of the present invention;

[0047] Figure 2 It is an MPC controller diagram based on NNPID feedback of the present invention;

[0048] Figure 3 It is a structural schematic diagram of NNPID of the present invention;

[0049] Figure 4 It is a visual playback diagram of the output file of the present invention; where Figure 4 (a) is of the first type, Figure 4 (b) is of the second type;

[0050] Figure 5This is a flowchart of a trajectory planning method for an autonomous vehicle in a complex driving environment according to the present invention. Detailed implementation manners

[0051] Combined with the accompanying drawings in the present invention below, the technical solutions of the embodiments of the present invention are clearly and completely described. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0052] The main purpose of the present invention is to construct a trajectory planning algorithm model for autonomous vehicles suitable for complex driving environments by deeply researching and developing efficient and reliable trajectory planning technologies, improve the real-time performance and robustness of the trajectory planning algorithm, ensure stable and reliable operation in complex environments, explore the application of multi-source information fusion technology in trajectory planning, improve the accuracy and reliability of trajectory planning, and improve the driving safety and decision-making efficiency of autonomous vehicles in complex driving environments. Specifically, the objectives of the present invention include:

[0053] Enhance driving safety: By finely simulating and comprehensively analyzing complex road conditions and traffic participant behaviors, the present invention aims to develop trajectory planning algorithms that can handle various complex scenarios. These algorithms will help autonomous vehicles make more accurate and safe driving decisions, thereby reducing the risk of traffic accidents.

[0054] Optimize decision-making efficiency: Efficient trajectory planning algorithms can not only ensure driving safety but also improve the decision-making efficiency of autonomous vehicles. In changing road conditions, quick and accurate decisions can help autonomous vehicles better handle emergencies and improve driving smoothness and comfort.

[0055] Promote the practical application of autonomous driving technology: It will provide solid theoretical support and powerful technical support for the practical application of autonomous driving technology. By solving the key problems faced by autonomous vehicles in actual road environments, the present invention will promote the transformation of autonomous driving technology from theory to practice and provide strong support for the continuous progress and development of intelligent transportation systems.

[0056] Based on this, this embodiment provides a trajectory planning method for an autonomous vehicle in a complex driving environment, including:

[0057] Step S1: Obtain different sensor data of the vehicle and vehicle state information, perform data fusion on different sensor data of the vehicle, and construct a unified environment model.

[0058] Sensor data fusion: Use sensors such as lidar, cameras, and radars to obtain information about the surrounding environment, such as road boundaries, obstacle positions, traffic signals, and weather conditions. Through data fusion technology, the information from different sensors is fused into a unified environmental model.

[0059] Vehicle state information: Includes the vehicle's position, speed, acceleration, direction, etc. This information will be used in the vehicle dynamics model in the trajectory planning algorithm.

[0060] Obtain traffic rules, such as lane lines, traffic signals, intersection restrictions, etc., through map data or real-time traffic information as auxiliary information for explanation.

[0061] Researchers have improved and innovated the MPC algorithm, making great contributions to improving the robustness and real-time performance of the algorithm. However, few people have paid attention to the tracking accuracy of large-curvature curves. After comparing the performance of various algorithms, Chu D et al. added the PID algorithm to the feedback correction process of the MPC algorithm on the basis of selecting a simplified kinematic model to improve the trajectory tracking accuracy. However, after adding the PID feedback control, this part returns to the data-based control algorithm, which makes it fall into the limitations of data-driven. To improve the high-speed control effect of autonomous vehicles under a simplified model, the present invention adds lateral deviation prediction on the basis of the existing MPC algorithm and designs a trajectory tracking controller by adding NNPID negative feedback to propose the MPC-NNPID algorithm to achieve high-precision tracking control.

[0062] Step S2: Based on the environmental model, construct a vehicle kinematic model based on the relationship between vehicle states, and construct an MPC controller based on NNPID feedback in the vehicle kinematic model through an MPC controller and an NNPID controller.

[0063] The specific expression of the vehicle kinematic model is:

[0064]

[0065] where φ is the heading angle, β is the centroid side slip angle, is the yaw angle, v x is the longitudinal vehicle speed, a x is the longitudinal acceleration, δ f is the front wheel steering angle, l r and l f are the distances from the centroid to the front and rear axles, X g and Y g are the positions of the vehicle in the earth coordinate system, and are v x 、 X g and Y gThe first derivative.

[0066] Step S3: Input the longitudinal acceleration and the front wheel steering angle in the vehicle state information into the vehicle kinematic model, update the vehicle state and obtain the longitudinal control quantity for adjusting the torque of the vehicle power motor. Based on the updated vehicle state, obtain the lateral deviation in the next T time domains through the MPC controller based on NNPID feedback, and correct the lateral deviation by adjusting the front wheel steering angle to obtain the lateral control quantity. Use the lateral control quantity and the longitudinal control quantity for vehicle trajectory planning.

[0067] The MPC (model predictive control) controller based on NNPID (Neural Network Proportional-Integral-Derivative) feedback is as Figure 1 shown. This controller consists of two parts: a feedforward control rate and a feedback control rate. The feedforward control rate uses an MPC controller based on the kinematic model, and the feedback control rate is an NNPID controller (N-PID). The controller obtains the lateral deviation e le (k) in the next T time domains and corrects the lateral deviation in real time. The lateral control quantity is achieved by outputting the wheel steering angle δout(k), and the effects of the MPC and NNPID controllers are superimposed; the longitudinal control quantity is completed by adjusting the torque T(k) of the vehicle power motor.

[0068] Vehicle state update:

[0069] Update the vehicle position. The specific expression is:

[0070] x t+Δt = x t + v t × cos(θ t ) × Δt;

[0071] y t+Δt == y t + v t × sin(θ t ) × Δt;

[0072] where, x t and y t are the x and y coordinates of the vehicle at the current moment, x t+Δt and y t+Δt are the x and y coordinates of the vehicle at the next moment, v t is the vehicle speed at the current moment, θ t is the vehicle heading angle at the current moment, and Δt is the time length of one step, the environmental update period.

[0073] Update the vehicle heading angle. The specific expression is:

[0074] θ t+Δt = θ t + v t / l × tan(δ) × Δt;

[0075] Where θ t+Δt is the vehicle heading angle at the next moment, δ is the front wheel steering angle, and l is the vehicle wheelbase. Here, it is agreed that its value is: the vehicle length divided by 1.7.

[0076] Update the vehicle speed. The specific expression is:

[0077] v t+Δt = max(0, v t + a * Δt);

[0078] Where v t+Δt is the vehicle speed at the next moment, and a is the longitudinal acceleration.

[0079] Obtain the lateral deviation e le (k) in the future T time domains through the updated vehicle state, including:

[0080] Based on the MPC controller, referring to the abscissa x m , the ordinate y m , the heading angle θ m and the curvature k m , the tangent vector τ m and the normal vector n m at the matching point can be obtained according to the heading angle. The specific expression is:

[0081] τ m = (cosθ m , sinθ m );

[0082] n m = (-sinθ m , cosθ m );

[0083] Obtain the tangent vector τ and the normal vector n according to the heading angle θ of the vehicle position. The specific expression is:

[0084] τ = (cosθ, sinθ);

[0085] n = (-sinθ, cosθ);

[0086] Obtain the lateral deviation e l according to the line P connecting the current trajectory matching point and the center point of the vehicle rear axle. The specific expression is:

[0087] e l = P · n = dx ·(-sinθ) + d y ·cosθ;

[0088] e l = P·n = d x ·(-sinθ) + d y ·cosθ;

[0089] where the vector P = (x - x m , y - y m ) = (d x , d y ), x is the abscissa of the vehicle position, y is the ordinate of the vehicle position, d x is the abscissa difference, and d y is the ordinate difference.

[0090] Taking the mean value of the T time domains after each model prediction of MPC as the final feedback error input, and using the final feedback error input as the lateral deviation e le in the future T time domains, the specific expression is:

[0091]

[0092] where i is the i-th time domain.

[0093] When using a simplified kinematic model in trajectory tracking, the problem of steady-state error will inevitably occur. Especially when the vehicle is driving on a large-curvature curve, the deviation between the actual trajectory and the reference trajectory may increase significantly. This error mainly comes from the simplification and inaccuracy of the vehicle internal model in MPC and usually has a consistent trend. Therefore, it is particularly important to compensate the lateral deviation through NNPID feedback. This method can not only use the PID controller to correct the output of MPC, but also realize the adaptive adjustment of PID parameters through a neural network (NN). It should be noted that this neural network does not require the traditional training process, thus effectively improving the lateral trajectory tracking performance of the controller without significantly increasing the burden of parameter adjustment. The structural schematic diagram of NNPID is shown in the figure.

[0094] When performing real-time correction of the lateral deviation, it also includes:

[0095] Using the PID controller to correct the output of MPC and realizing the adaptive adjustment of the parameters of the PID controller through a neural network (NN), the specific expression is:

[0096]

[0097] where is the input layer - output layer weight matrix, y iIt is the vector after PID calculation. z, x, and u are the expected value, actual value, and network output respectively. In the controller, z is the expected error, which is set to 0, x' is the current actual lateral deviation, and u is the lateral steering angle control quantity δ output by NNPID. N_PID , where k is the independent variable.

[0098] After the algorithm design is completed, the present invention will enter the stage of algorithm implementation and preliminary testing. The present invention will use the Onsite open-source autonomous driving intelligent vehicle simulation test environment to implement the algorithm model and conduct preliminary testing and verification. In this stage, the present invention will focus on the real-time performance and robustness of the algorithm to ensure that it can quickly respond and handle various emergencies in actual applications. At the same time, the present invention will conduct sufficient unit tests and functional tests on the algorithm to ensure that it meets the design requirements. As shown in Table 1.

[0099] Table 1 Preliminary Test Table

[0100]

[0101] The Onsite open-source natural driving intelligent vehicle simulation test environment is a highly simulated virtual test platform that focuses on simulating complex traffic scenarios and natural driving behaviors in reality and provides support for the research and evaluation of autonomous driving technologies. This environment can accurately reproduce road structures, dynamic traffic participants (such as cars and pedestrians), weather conditions, and sensor feedback (such as lidar, cameras, and radars). It adopts an open architecture, supports users to customize test scenarios, vehicle dynamics models, and sensor configurations, and integrates with mainstream autonomous driving frameworks (such as ROS and Apollo) to provide flexible development and testing capabilities. Through real-time optimization and detailed performance analysis, this environment is suitable for path planning, obstacle avoidance, behavior prediction, and interaction research, while ensuring the repeatability and reliability of experimental results.

[0102] Column 1: Simulation timestamp. Columns 2-3: Control quantities returned by the planning and control module. Columns 4-11: Vehicle state information, which are the abscissa, ordinate, speed, acceleration, heading angle, front wheel angle, width, and length of the vehicle respectively. The acceleration and front wheel angle of the vehicle are values corrected by actuator dynamics.

[0103] The last column: Simulation running status.

[0104] Generally speaking, except for the last row, the rest of the values are -1. The meanings of the values in the end column are as follows: -1 indicates that the simulation is in progress. 1 indicates that the vehicle has completed the test task, that is, it has successfully reached the target area. 2 indicates that the simulation has timed out, that is, it has exceeded the longest recording time of OpenSCENARIO. 3 indicates that the vehicle has collided with background elements. 4 indicates that the vehicle has driven out of the map boundary, that is, it has exceeded the range of the static map bounding box. The other columns: represent the status information of other background elements in the simulation environment.

[0105] Each group of 7 columns represents the abscissa, ordinate, speed, acceleration, heading angle, width, and length of the corresponding background element respectively. The name of the background element is the part after the underscore of the column name. Visualize and playback the output file.

[0106] Based on the above method, the present invention provides an automatic driving vehicle trajectory planning system in a complex driving environment, including: a collection module, a model construction module, and a planning module.

[0107] Among them, the collection module is used to obtain different sensor data of the vehicle and vehicle state information, fuse the different sensor data of the vehicle, and construct a unified environment model; the model construction module is used to construct a vehicle kinematic model based on the relationship between vehicle states based on the environment model, and construct an MPC controller based on NNPID feedback in the vehicle kinematic model through an MPC controller and an NNPID controller; the planning module is used to input the longitudinal acceleration and front wheel angle in the vehicle state information into the vehicle kinematic model, update the vehicle state and obtain a longitudinal control quantity for adjusting the torque of the vehicle's power motor, based on the updated vehicle state, obtain the lateral deviation in the next T time domains through the MPC controller based on NNPID feedback, and correct the lateral deviation by adjusting the front wheel angle to obtain a lateral control quantity, and perform vehicle trajectory planning with the lateral control quantity and the longitudinal control quantity.

[0108] The present invention also provides a computer device, including a memory and a processor. When a program stored in the memory is executed by the processor, the processor executes the steps of an automatic driving vehicle trajectory planning method in a complex driving environment.

[0109] According to the disclosed embodiments, the computer device can communicate with one or more external devices (such as a keyboard, a pointing device, Bluetooth communication, etc.), or communicate with any device (such as a router, a demodulator, etc.) that enables the computing device to communicate with one or more other computing devices.

[0110] The present invention also provides a storage medium, on which a computer program is stored. When the computer program is executed by a processor, the steps of an automatic driving vehicle trajectory planning method in a complex driving environment are realized.

[0111] According to the disclosed embodiments, the storage medium may be a non-volatile computer-readable storage medium, which may include, for example, but not limited to: portable computer disks, hard disks, random access memories (RAMs), read-only memories (ROMs), erasable programmable read-only memories (EPROMs or flash memories), portable compact disk read-only memories (CD-ROMs), optical storage devices, magnetic storage devices, or any suitable combination of the above. In the present invention, the storage medium may be any tangible medium that contains or stores a program that can be used by or in conjunction with an instruction execution system, apparatus, or device.

[0112] The above content is a further detailed description of the present invention in combination with specific preferred embodiments. For those skilled in the technical field of the present invention, without departing from the concept of the present invention, several simple deductions or substitutions can be made, which should all be regarded as belonging to the protection scope of the present invention.

Claims

1. A trajectory planning method for an autonomous vehicle in a complex driving environment, characterized in that, Including: Obtain different sensor data of the vehicle and vehicle state information, fuse the different sensor data of the vehicle, and construct a unified environment model; Based on the environment model, construct a vehicle kinematic model according to the relationship between vehicle states, and construct an MPC controller based on NNPID feedback in the vehicle kinematic model through an MPC controller and an NNPID controller; Input the longitudinal acceleration and front wheel angle in the vehicle state information into the vehicle kinematic model, update the vehicle state and obtain a longitudinal control quantity for adjusting the torque of the vehicle power motor. Based on the updated vehicle state, obtain the lateral deviation in the next T time domains through the MPC controller based on NNPID feedback, and correct the lateral deviation by adjusting the front wheel angle to obtain a lateral control quantity, and perform vehicle trajectory planning with the lateral control quantity and the longitudinal control quantity.

2. The trajectory planning method for an autonomous vehicle in a complex driving environment according to claim 1, characterized in that, The construction of the vehicle kinematic model according to the relationship between vehicle states, and the specific expression of the vehicle kinematic model is: Where, φ is the heading angle, β is the sideslip angle at the center of mass, is the yaw angle, v x is the longitudinal vehicle speed, a x is the longitudinal acceleration, δ f is the front wheel steering angle, l r and l f are the distances from the center of mass to the front and rear axles, X g and Y g are the positions of the vehicle in the earth coordinate system, and are the first derivatives of v x , X g and Y g respectively.

3. A trajectory planning method for an autonomous vehicle in a complex driving environment according to claim 1, characterized in that, Input the longitudinal acceleration and front wheel angle in the vehicle state information into the vehicle kinematic model to update the vehicle state, where the vehicle state includes vehicle position, vehicle heading angle and vehicle speed, specifically: Update the vehicle position, and the specific expression is: x t+Δt = x t + v t × cos(θ t ) × Δt; y t+Δt == y t + v t × sin(θ t ) × Δt; where x t and y t are the x and y coordinates of the vehicle at the current moment, x t+Δt and y t+Δt are the x and y coordinates of the vehicle at the next moment, v t is the vehicle speed at the current moment, θ t is the vehicle heading angle at the current moment, and Δt is the time length of one time step; Update the vehicle heading angle, and the specific expression is: θ t+Δt = θ t + v t / l × tan(δ) × Δt; where θ t+Δt is the vehicle heading angle at the next moment, δ is the front wheel steering angle, and l is the vehicle wheelbase; Update the vehicle speed, and the specific expression is: v t+Δt = max(0, v t + a * Δt); Among them, v t+Δt is the vehicle speed at the next moment, and a is the longitudinal acceleration.

4. A trajectory planning method for an autonomous vehicle in a complex driving environment according to claim 1, characterized in that, Obtaining the lateral deviation in the next T time domains through the MPC controller based on NNPID feedback includes: Based on the MPC controller, the lateral deviation e is obtained according to the line P connecting the current trajectory matching point and the center point of the rear axle of the vehicle l , and the specific expression is as follows: e l = P·n = d x ·(-sinθ) + d y ·cosθ; Among them, the vector P = (x - x m , y - y m ) = (d x , d y ), where x is the abscissa of the vehicle position, y is the ordinate of the vehicle position, x m is the reference abscissa, y m is the reference ordinate, d x is the abscissa difference, d y is the ordinate difference, θ is the heading angle of the vehicle position, and n is the normal vector at the vehicle position; n = (-sinθ, cosθ); Take the mean value of the T time domains after each prediction of the MPC as the final feedback error input, and use the final feedback error input as the lateral deviation e in the next T time domains. le , and the specific expression is: where i is the i-th time domain.

5. A trajectory planning method for an autonomous vehicle in a complex driving environment according to claim 1, characterized in that, When correcting the lateral deviation by adjusting the front wheel angle, it also includes: Use a PID controller to correct the MPC output, and adaptively adjust the parameters of the PID controller through a neural network, and the specific expression is: Among them, is the input layer - output layer weight matrix, and y i is the vector after PID calculation. In the controller, z is the desired error, x' is the current actual lateral deviation, and u is the lateral steering angle control quantity δ output by NNPID N_PID , and k is the independent variable.

6. The trajectory planning method for an autonomous vehicle in a complex driving environment according to claim 1, wherein, The environment model includes road boundaries, obstacle positions, traffic signals and weather conditions.

7. An automatic driving vehicle trajectory planning system under a complex driving environment, characterized in that, Including: A collection module for obtaining different sensor data of the vehicle and vehicle state information, fusing the different sensor data of the vehicle, and constructing a unified environment model; A model construction module for constructing a vehicle kinematic model based on the environment model according to the relationship between vehicle states, and constructing an MPC controller based on NNPID feedback in the vehicle kinematic model through an MPC controller and an NNPID controller; A planning module for inputting the longitudinal acceleration and front wheel angle in the vehicle state information into the vehicle kinematic model, updating the vehicle state and obtaining a longitudinal control quantity for adjusting the torque of the vehicle power motor. Based on the updated vehicle state, obtaining the lateral deviation in the next T time domains through the MPC controller based on NNPID feedback, and correcting the lateral deviation by adjusting the front wheel angle to obtain a lateral control quantity, and performing vehicle trajectory planning with the lateral control quantity and the longitudinal control quantity.

8. A computer device, characterized in that, Including a memory and a processor. When the program stored in the memory is executed by the processor, the processor executes the steps of the method for autonomous vehicle trajectory planning in a complex driving environment according to any one of claims 1 to 6.

9. A storage medium, on which a computer program is stored, characterized in that, When the computer program is executed by a processor, it implements the steps of a method for trajectory planning of an autonomous vehicle in a complex driving environment according to any one of claims 1 to 6.

Images