Automatic driving decision planning method based on deep reinforcement learning and A-Star algorithm

Abstract

The invention discloses an automatic driving decision planning method based on deep reinforcement learning and an A-Star algorithm, and the method comprises the steps: carrying out the global path planning through the A-Star algorithm, processing local environment perception information through a deep reinforcement learning model, and making a real-time decision; a neural network model is trained by setting a reward function, so that a complex traffic environment can be effectively processed on the premise of ensuring safety, and good real-time performance and flexibility are achieved; a dynamic safe vehicle distance calculation mechanism based on the vehicle speed and a reward and punishment mechanism for behaviors such as deviating from a global path and approaching a front vehicle are introduced, so that the driving safety is improved; through simulation environment training, the intelligent agent can learn an optimal driving strategy, and more efficient and more intelligent automatic driving is realized. According to the method, the driving decision can be efficiently and accurately provided for the automatic driving vehicle, and the method has good generalization ability and adaptability so as to adapt to complex and changeable traffic environments.

Description

Technical Field

[0001] The present invention relates to the field of autonomous driving technology, and specifically to an autonomous driving decision planning method based on deep reinforcement learning and A-Star algorithm. Background Art

[0002] Autonomous driving technology has made significant progress in recent years, but it still faces many challenges, especially in decision-making and planning in complex dynamic environments. Autonomous driving vehicles need to perform safe and reliable path planning and real-time decision-making in various complex environments to ensure passenger safety, improve driving efficiency and reduce the occurrence of traffic accidents.

[0003] Path planning is one of the core tasks of autonomous driving, which usually involves determining the optimal path from a starting point to a target point in a known environment. Existing path planning technologies can be roughly divided into two categories: global path planning and local path planning.

[0004] Global path planning: often used to plan long-distance paths from the starting point to the destination. The classic A-Star algorithm is widely used due to its high efficiency and optimality. The A-Star algorithm uses a heuristic function to guide the search process and can find the shortest path from the starting point to the target in a known map environment. However, the A-Star algorithm assumes that the environment is static and has poor adaptability to dynamically changing environments. Once the environment changes, such as the appearance of new obstacles, the A-Star algorithm needs to replan the path, resulting in insufficient real-time performance.

[0005] Local path planning: It is used for local planning problems in dynamic environments during actual driving of autonomous vehicles, especially when avoiding obstacles and dealing with emergencies. Existing methods include those based on artificial potential field method, dynamic window method and rapidly expanding random tree (RRT). These methods have certain advantages in dealing with dynamic environments, but they often have limitations in local optimal problems and it is difficult to take into account the global optimal problem.

[0006] As an emerging machine learning method, deep reinforcement learning has demonstrated its excellent performance in many fields. In particular, in the field of autonomous driving, deep reinforcement learning has shown great potential in processing high-dimensional perceptual data and continuous action space by learning optimal strategies through interaction with the environment. For example, studies have used deep Q-networks (DQNs) to achieve end-to-end autonomous driving behavior learning in a simulated environment. Although reinforcement learning has shown great potential in autonomous driving decision planning, there are still some challenges and limitations. First, reinforcement learning algorithms usually require a large number of samples to learn effective strategies. In autonomous driving scenarios, they still face problems such as low sample efficiency, slow convergence, and difficult to explain strategies; second, in real-world driving environments, the safety and reliability requirements of reinforcement learning models are extremely high, and reinforcement learning methods alone may not be able to meet these requirements.

[0007] Although the A-Star algorithm performs well in path planning, its application in dynamic environments is limited; and although deep reinforcement learning performs well in dealing with dynamic decision-making problems, its global planning ability is weak and it is easy to fall into local optimality. At present, it is difficult for existing technologies to simultaneously take into account the optimality of global path planning and the real-time and flexibility of local decision-making. Summary of the invention

[0008] The purpose of the present invention is to provide an autonomous driving decision planning method based on deep reinforcement learning and A-Star algorithm to solve the above defects.

[0009] In order to achieve the above object, the present invention provides the following technical solutions:

[0010] An autonomous driving decision planning method based on deep reinforcement learning and A-Star algorithm includes the following steps:

[0011] S1, use A-Star algorithm to perform global path planning on a known map;

[0012] S2. Building a neural network model based on deep reinforcement learning, which is responsible for processing local environment perception information and making real-time decisions;

[0013] S3. Setting a reward function for the neural network model;

[0014] S4, the neural network model receives the perception information acquired and transmitted by the perception system;

[0015] S5, using the global path information of step S1, the reward function of step S3 and the perception information of step S4, training the neural network model of step S2;

[0016] S6. The reinforcement learning module in the neural network model optimizes the local path based on the global path planning provided by the perception information and the A-Star algorithm, uses the trained neural network model, and then outputs the decision.

[0017] Preferably, in step S5, the deep neural network model of step S2 is trained, including training of independent scenarios and training of mixed scenarios;

[0018] The independent scene is one of a straight road scene, a sharp bend scene, an intersection scene, a ramp scene, and a roundabout scene; the mixed scene is a combination of two or more of the straight road scene, the sharp bend scene, the intersection scene, the ramp scene, and the roundabout scene;

[0019] During training, the Q function used is: Q(s,a;θ)=W2 ·f(W 1 ·s+b 1 )+b 2 , where s represents the input state vector, i.e., the current state, a represents the action taken in the current state s, θ represents the parameters of the neural network, and W 1 is the weight matrix from the input layer to the hidden layer, W 2 is the weight matrix from the hidden layer to the output layer, b 1 is the bias vector of the hidden layer, b 2 is the bias vector of the output layer, and f(·) is the nonlinear activation function;

[0020] The loss function L(θ) is defined as the mean square error between the target Q value y and the Q value Q(s,a;θ) predicted by the current Q network. The loss function L(θ) formula is:

[0021] L(θ)=E[(yQ(s,a;θ)) 2 ],

[0022] Where E represents the expected value, which is usually the mean of a mini-batch of samples and is used to stabilize the gradient update.

[0023] The target Q value y is calculated based on the current reward and the maximum Q value of the next state, and is calculated as follows:

[0024] y=R+γ·a′maxQ(s′,a′;θ - ),

[0025] In the formula, R is the immediate reward obtained by the agent based on the neural network model after taking action a in state s; γ is the discount factor, which ranges from [0,1] and represents the discount rate of future rewards; s′ represents the next state reached by the agent after taking action a, and max a′ Q(s′,a′;θ - ) represents the one with the largest Q value among all possible actions a' in the next state s', θ - represents the parameters of the target Q network;

[0026] Substitute the target Q value y into the loss function L(θ) to obtain the specific loss function expression:

[0027]

[0028] By minimizing this loss function, the parameters θ in the DON network are gradually adjusted to enable the network to more accurately predict the Q value of each state-action pair, thereby improving the decision-making ability of the neural network model.

[0029] Preferably, in step S1, the path generated by global path planning is the optimal path from the starting point to the target point. The optimal path serves as a reference for the reinforcement learning module in the subsequent neural network model to guide the optimization process of local decision-making.

[0030] Preferably, in step S2, the constructed neural network model is based on the DQN algorithm; the neural network model includes an input layer, an output layer, and a hidden layer.

[0031] Preferably, in step S3, when setting the reward function, the reward and punishment indicators include path execution, collision avoidance, distance keeping, traffic rule compliance, and driving stability, which comprehensively evaluate the driving behavior of the model and continuously optimize the model's strategy through a reinforcement learning algorithm.

[0032] Preferably, in step S4, the perception information includes: the deviation D of the autonomous vehicle from the planned path m , the shortest longitudinal distance D between the autonomous vehicle and the vehicle ahead y , the lateral distance D between the autonomous vehicle and the nearest vehicle in the other lanes of the current road section x , the angle deviation D between the current heading angle of the automatic car and the line connecting the two path points on the navigation path ahead a , the current speed v of the automatic car, the current steering wheel angle a of the automatic car, the current longitudinal acceleration a of the automatic car y , the current lateral acceleration of the automatic car is a x , the longitudinal relative speed v between the vehicle in front of the automatic car and the automatic car in the current road section t .

[0033] Preferably, the deviation D of the autonomous vehicle from the planned path is m The calculation formula is: Where d represents the distance between the current position of the autonomous vehicle and the matching point on the planned path, and D represents the maximum width of the autonomous vehicle;

[0034] The shortest longitudinal distance D between the autonomous vehicle and the vehicle ahead y Ratio of dynamic safety distance to d The calculation formula is:

[0035] In the formula, v represents the current vehicle speed, β represents the dynamic safety distance coefficient at the current vehicle speed, and d base Indicates the basic safety distance.

[0036] Preferably, the global path planning and local path optimization are performed in real time during vehicle driving to dynamically adjust the driving strategy.

[0037] The beneficial effects of the present invention are:

[0038] (1) The present invention constructs a hierarchical system based on the autonomous driving decision-making planning method of deep reinforcement learning and the A-Star algorithm. The top layer uses the A-Star algorithm for global path planning, and the bottom layer uses reinforcement learning to handle local decisions, such as obstacle avoidance and speed control. By combining the path planning ability of the A-Star algorithm with the adaptive decision-making ability of reinforcement learning, it can effectively plan the optimal path in a complex and changeable road environment, and dynamically adjust the driving strategy to ensure the safety and efficient driving of the vehicle in various scenarios.

[0039] (2) In the method of the present invention, a dynamic safe vehicle distance calculation mechanism based on vehicle speed is introduced to ensure that the vehicle can maintain an appropriate safe distance at different vehicle speeds. Combined with the reward and punishment mechanism of reinforcement learning, the system can give gradually increasing penalties when the vehicle distance is insufficient, thereby effectively avoiding rear-end collisions and other collision accidents and improving driving safety.

[0040] (3) In the method of the present invention, deep reinforcement learning combined with the A-Star algorithm not only improves the decision-making planning efficiency of the autonomous driving system, but also reduces the computational overhead of the traditional rule-based decision-making system in complex environments. The intelligent agent based on the neural network model can be trained extensively in a simulation environment to learn the best driving strategy, thereby achieving more efficient and intelligent autonomous driving.

[0041] (4) The method of the present invention can provide driving decisions for autonomous driving vehicles efficiently and accurately, and has good generalization ability and adaptability to complex and changing traffic environments. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] Figure 1 : Schematic diagram of the process of the present invention;

[0043] Figure 2 : A schematic diagram of the perception information received by the Neutral Intelligence Neural Network model of the present invention;

[0044] Figure 3 : Schematic diagram of the relationship between the neural network model and the state and decision commands in the present invention. DETAILED DESCRIPTION

[0045] The present invention is further described below in conjunction with the embodiments. It should be noted that these are merely examples and illustrations of the concept of the present invention. Those skilled in the art may make various modifications or additions to the described specific embodiments or replace them in a similar manner. As long as they do not deviate from the concept of the invention or exceed the scope defined by the claims, they should be deemed to fall within the protection scope of the present invention.

[0046] Embodiment 1:

[0047] Figure 1 It is a schematic diagram of the process of the present invention, such as Figure 1 As shown in the figure, the autonomous driving decision planning method based on deep reinforcement learning and A-Star algorithm includes the following steps:

[0048] S1. Use the A-Star algorithm to perform global path planning on a known map.

[0049] The path generated by global path planning on a known road map using the A-Star algorithm is the optimal path from the starting point to the target point. This optimal path serves as a reference for the reinforcement learning module in the subsequent neural network model to guide the optimization process of local decisions. The A-Star algorithm is an algorithm based on heuristic search. By combining the actual cost and estimated cost of the path, it can find the path with the lowest cost in the graph search problem.

[0050] S2. Build a neural network model based on deep reinforcement learning.

[0051] The constructed neural network model is based on the DQN algorithm and includes a three-layer structure of input layer, output layer, and hidden layer. It is responsible for processing local environmental perception information and making real-time decisions.

[0052] S3. Set a reward function for the neural network model.

[0053] When setting the reward function, the reward and punishment indicators include path execution, collision avoidance, distance maintenance, traffic rule compliance, driving stability and other aspects, which comprehensively evaluate the driving behavior of the model and continuously optimize the model's strategy through reinforcement learning algorithms.

[0054] Events of reward and punishment include:

[0055] Deviation from the global path, punishing the behavior of the autonomous car deviating from the center line of the road. The farther away from the center, the heavier the penalty. The reward R is: R = -D m , D m is the deviation of the autonomous vehicle from the planned path;

[0056] Approaching the vehicle ahead, if the distance between the autonomous vehicle and the vehicle ahead in the same direction is less than the dynamic safety distance, it will enter an unsafe situation. The smaller the ratio of the distance between the autonomous vehicle and the vehicle ahead in the same direction to the dynamic safety distance, the heavier the penalty. The reward is: D y is the shortest longitudinal distance between the automatic car and the vehicle in front, v represents the current vehicle speed, β represents the dynamic safety distance coefficient at the current vehicle speed, and d base Indicates the basic safety distance;

[0057] Approaching lateral vehicles, if the autonomous car is less than 6 meters away from vehicles on both sides of the intersection, it will enter an unsafe situation and a small penalty will be given as a warning, with a reward of -3;

[0058] Too large longitudinal acceleration: If the longitudinal acceleration is too large (including sudden acceleration and sudden braking), it will cause serious pitching of the vehicle, affecting the comfort and safety of the vehicle. When the longitudinal acceleration exceeds 5m / s 2 When the value is higher, the punishment is heavier, and the reward is: R = 5-|a y |, a y is the current longitudinal acceleration of the automatic car;

[0059] Excessive lateral acceleration: If the lateral acceleration is too large, it may cause the vehicle to lose control or cause discomfort to the passengers. When the lateral acceleration exceeds 3m / s 2 When the value is higher, the punishment is heavier, and the reward is: R = 3-|a x |, a x is the current lateral acceleration of the automatic car;

[0060] Speeding: If the speed of an autonomous vehicle exceeds the speed limit by 80% of the road, a penalty will be imposed according to the speed difference;

[0061] If the speed of an automatic car is lower than the walking speed (1.5m / s), a slight penalty will be given. If the speed is too slow, the reward will be -1.

[0062] S4. The neural network model receives the perception information acquired and transmitted by the perception system.

[0063] Figure 2 This is a schematic diagram of the perception information received by the intelligent neural network model. Figure 2 As shown, the perception information includes: the deviation D of the automatic car from the planned path m , the shortest longitudinal distance D between the autonomous vehicle and the vehicle ahead y , the lateral distance D between the autonomous vehicle and the nearest vehicle in the other lanes of the current road section x , the angle deviation D between the current heading angle of the automatic car and the line connecting the two path points on the navigation path ahead a , the current speed v of the automatic car, the current steering wheel angle a of the automatic car, the current longitudinal acceleration a of the automatic car y , the current lateral acceleration of the automatic car is a x , the longitudinal relative speed v between the vehicle in front of the automatic car and the automatic car in the current road section t .

[0064] Deviation D between the autonomous vehicle and the planned path mThe calculation formula is: Where d represents the distance between the current position of the autonomous vehicle and the matching point on the planned path, and D represents the maximum width of the autonomous vehicle.

[0065] The shortest longitudinal distance D between the autonomous vehicle and the vehicle ahead y Ratio of dynamic safety distance to d The calculation formula is:

[0066] In the formula, v represents the current vehicle speed, β represents the dynamic safety distance coefficient at the current vehicle speed, and d base Indicates the basic safety distance.

[0067] All perception information is combined into a 13-dimensional vector as the input state, including: path deviation distance, speed, longitudinal acceleration, lateral acceleration, steering angle, relative speed of the nearest vehicle, lateral distance, and heading angle deviation in dimension 1; the relative position of the nearest lateral vehicle in dimension 2; and the distance to the preceding vehicle in dimension 3.

[0068] S5. Use the global path information of step S1, the reward function of step S3, and the perception information of step S4 to train the neural network model of step S2.

[0069] Among them, the neural network model is trained, including independent scenario training and mixed scenario training.

[0070] The independent scene is one of the straight road scene, sharp turn scene, intersection scene, ramp scene, and roundabout scene; the mixed scene is a combination of two or more of the straight road scene, sharp turn scene, intersection scene, ramp scene, and roundabout scene.

[0071] During training, the Q function used is: Q(s,a;θ)=W 2 ·f(W 1 ·s+b 1 )+b 2 , where s represents the input state vector, i.e., the current state, a represents the action taken in the current state s, θ represents the parameters of the neural network, and W 1 is the weight matrix from the input layer to the hidden layer, W 2 is the weight matrix from the hidden layer to the output layer, b 1 is the bias vector of the hidden layer, b 2 is the bias vector of the output layer, and f(·) is a nonlinear activation function, such as ReLU (Rectified Linear Unit): f(x) = max(0,x).

[0072] The loss function L(θ) is defined as the mean square error between the target Q value y and the Q value Q(s,a;θ) predicted by the current Q network. The loss function L(θ) formula is:

[0073] L(θ)=E[(yQ(s,a;θ)) 2 ],

[0074] Where E represents the expected value, which is usually the mean of a mini-batch of samples and is used to stabilize the gradient update.

[0075] The target Q value y is calculated based on the current reward and the maximum Q value of the next state, and is calculated as follows:

[0076] y=R+γ·a′maxQ(s′,a′;θ - ),

[0077] In the formula, R is the immediate reward obtained by the agent based on the neural network model after taking action a in state s; γ is the discount factor, which ranges from [0,1] and represents the discount rate of future rewards; s′ represents the next state reached by the agent after taking action a, and max a′ Q(s′,a′;θ - ) represents the one with the largest Q value among all possible actions a' in the next state s', θ - Represents the parameters of the target Q network.

[0078] Substitute the target Q value y into the loss function L(θ) to obtain the specific loss function expression:

[0079]

[0080] By minimizing this loss function, the parameters θ in the DON network are gradually adjusted to enable the network to more accurately predict the Q value of each state-action pair, thereby improving the decision-making ability of the neural network model.

[0081] S6, the reinforcement learning module in the neural network model optimizes the local path based on the global path planning provided by the perception information and A-Star algorithm, and then outputs the decision to be executed by the controller. The global path planning and local path optimization are carried out in real time during the vehicle driving process to dynamically adjust the driving strategy.

[0082] (1) The present invention constructs a hierarchical system based on the autonomous driving decision-making planning method of deep reinforcement learning and the A-Star algorithm. The top layer uses the A-Star algorithm for global path planning, and the bottom layer uses reinforcement learning to handle local decisions, such as obstacle avoidance and speed control. By combining the path planning ability of the A-Star algorithm with the adaptive decision-making ability of reinforcement learning, it can effectively plan the optimal path in a complex and changeable road environment, and dynamically adjust the driving strategy to ensure the safety and efficient driving of the vehicle in various scenarios.

[0083] (2) In the method of the present invention, a dynamic safe vehicle distance calculation mechanism based on vehicle speed is introduced to ensure that the vehicle can maintain an appropriate safe distance at different vehicle speeds. Combined with the reward and punishment mechanism of reinforcement learning, the system can give gradually increasing penalties when the vehicle distance is insufficient, thereby effectively avoiding rear-end collisions and other collision accidents and improving driving safety.

[0084] (3) In the method of the present invention, deep reinforcement learning combined with the A-Star algorithm not only improves the decision-making planning efficiency of the autonomous driving system, but also reduces the computational overhead of the traditional rule-based decision-making system in complex environments. The intelligent agent based on the neural network model can be trained extensively in a simulation environment to learn the best driving strategy, thereby achieving more efficient and intelligent autonomous driving.

[0085] (4) The method of the present invention can provide driving decisions for autonomous driving vehicles efficiently and accurately, and has good generalization ability and adaptability to complex and changing traffic environments.

[0086] The above is an exemplary description of the invention. Obviously, the specific implementation of the present invention is not limited to the above-mentioned method. As long as such non-substantial improvements are made by adopting the method concept and technical solution of the present invention, or the concept and technical solution of the invention are directly applied to other occasions without improvement, they are all within the protection scope of the present invention.

Claims

1. An autonomous driving decision-making planning method based on deep reinforcement learning and A-Star algorithm, characterized in that: The following steps are involved: S1, use A-Star algorithm to perform global path planning on a known map; S2. Building a neural network model based on deep reinforcement learning, which is responsible for processing local environment perception information and making real-time decisions; S3. Setting a reward function for the neural network model; S4, the neural network model receives the perception information acquired and transmitted by the perception system; S5, using the global path information of step S1, the reward function of step S3 and the perception information of step S4, training the neural network model of step S2; S6. The reinforcement learning module in the neural network model optimizes the local path based on the global path planning provided by the perception information and the A-Star algorithm, uses the trained neural network model, and then outputs the decision.

2. The autonomous driving decision-making planning method based on deep reinforcement learning and A-Star algorithm according to claim 1, characterized in that: In step S5, the neural network model of step S2 is trained, including independent scene training and mixed scene training; The independent scene is one of a straight road scene, a sharp bend scene, an intersection scene, a ramp scene, and a roundabout scene; the mixed scene is a combination of two or more of the straight road scene, the sharp bend scene, the intersection scene, the ramp scene, and the roundabout scene; During training, the Q function used is: Q(s,a;θ)=W2·f(W1·s+b1)+b2, where s represents the input state vector, i.e. the current state, a represents the action taken under the current state s, θ represents the parameters of the neural network, W1 is the weight matrix from the input layer to the hidden layer, W2 is the weight matrix from the hidden layer to the output layer, b1 is the bias vector of the hidden layer, b2 is the bias vector of the output layer, and f(·) is the nonlinear activation function; The loss function L(θ) is defined as the mean square error between the target Q value y and the Q value Q(s,a;θ) predicted by the current Q network. The loss function L(θ) formula is: L(θ)=E[(yQ(s,a;θ)) 2 ], Where E represents the expected value, which is usually the mean of a mini-batch of samples and is used to stabilize the gradient update. The target Q value y is calculated based on the current reward and the maximum Q value of the next state, and is calculated as follows: and=R+γ a′maxQ(s′,a′;θ - ), In the formula, R is the immediate reward obtained by the agent based on the neural network model after taking action a in state s; γ is the discount factor, which ranges from [0,1] and represents the discount rate of future rewards; s′ represents the next state reached by the agent after taking action a, and max a′ Q(s′,a′;θ - ) represents the action a′ corresponding to the maximum Q value among all possible actions in the next state s′, θ - represents the parameters of the target Q network; Substitute the target Q value y into the loss function L(θ) to obtain the specific loss function expression: By minimizing this loss function, the parameters θ in the DON network are gradually adjusted to enable the network to more accurately predict the Q value of each state-action pair, thereby improving the decision-making ability of the neural network model.

3. The autonomous driving decision-making planning method based on deep reinforcement learning and A-Star algorithm according to claim 1, characterized in that: In step S1, the path generated by global path planning is the optimal path from the starting point to the target point. The optimal path serves as a reference for the reinforcement learning module in the subsequent neural network model to guide the optimization process of local decision-making.

4. The autonomous driving decision-making planning method based on deep reinforcement learning and A-Star algorithm according to claim 1, characterized in that: In step S2, the constructed neural network model is based on the DQN algorithm; the neural network model includes an input layer, an output layer, and a hidden layer.

5. The autonomous driving decision-making planning method based on deep reinforcement learning and A-Star algorithm according to claim 1, characterized in that: In step S3, when setting the reward function, the reward and punishment indicators include path execution, collision avoidance, distance maintenance, traffic rule compliance, and driving stability, which comprehensively evaluate the driving behavior of the model and continuously optimize the model's strategy through the reinforcement learning algorithm.

6. The autonomous driving decision-making planning method based on deep reinforcement learning and A-Star algorithm according to claim 1, characterized in that: In step S4, the perception information includes: the deviation D of the autonomous vehicle from the planned path m , the shortest longitudinal distance D between the autonomous vehicle and the vehicle ahead y , the lateral distance D between the autonomous vehicle and the nearest vehicle in the other lanes of the current road section x , the angle deviation D between the current heading angle of the automatic car and the line connecting the two path points on the navigation path ahead a , the current speed v of the automatic car, the current steering wheel angle a of the automatic car, the current longitudinal acceleration a of the automatic car y , the current lateral acceleration of the automatic car is a x , the longitudinal relative speed v between the vehicle in front of the automatic car and the automatic car in the current road section t .

7. The autonomous driving decision-making planning method based on deep reinforcement learning and A-Star algorithm according to claim 6, characterized in that: The deviation D of the autonomous vehicle from the planned path m The calculation formula is: Where d represents the distance between the current position of the autonomous vehicle and the matching point on the planned path, and D represents the maximum width of the autonomous vehicle; The shortest longitudinal distance D between the autonomous vehicle and the vehicle ahead y Ratio of dynamic safety distance to d The calculation formula is: In the formula, v represents the current vehicle speed, β represents the dynamic safety distance coefficient at the current vehicle speed, and d base Indicates the basic safety distance.

8. The autonomous driving decision-making planning method based on deep reinforcement learning and A-Star algorithm according to claim 1, characterized in that: The global path planning and local path optimization processes are all performed in real time during vehicle driving to dynamically adjust the driving strategy.

Images