Target following and dynamic obstacle avoidance control method for differential slip steered vehicles

Abstract

The invention belongs to the technical field of unmanned driving, and discloses a target following and dynamic obstacle avoidance control method for a speed difference slip steering vehicle. The deep deterministic strategy in reinforcement learning is used to establish four neural networks; The cost range determines the single-step reward function of the action; the continuous action output is determined by the actor-critic strategy, and the network parameters are continuously updated by gradient transfer; the network model for following and avoiding obstacles is trained according to the current state. The invention improves the intelligence of vehicle following and obstacle avoidance, and can better adapt to unknown environments and cope with other emergencies well. The complexity of establishing a simulation environment during reinforcement learning training is reduced. Using the neural network prediction model trained in advance, the position and posture of the target vehicle and the obstacle at each step can be obtained from the initial position and posture of the target and the obstacle and the action value of each step, which improves the accuracy and efficiency of the simulation.

Description

technical field [0001] The invention belongs to the technical field of unmanned driving, and in particular relates to a target following and dynamic obstacle avoidance control method of a speed difference slip steering vehicle. Background technique [0002] At present, the closest existing technology: the traditional method mainly adopts path planning and path tracking control methods when performing target following and obstacle avoidance. That is, a path is first planned from the current target state and obstacle information, and then the trajectory tracking control method is used to control the vehicle to travel along the planned path, and at the same time, the local path planning method is used to avoid obstacles. The trajectory tracking is usually in constant speed mode, which adopts a combination of path planning and trajectory tracking. Due to the complexity of the control of the speed difference slip steering vehicle, the controlled vehicle cannot be adjusted in real...

AI Technical Summary

Problems solved by technology

[0005] (1) When performing trajectory tracking, it is usually a constant speed mode, which adopts a combination of path planning and trajectory tracking, and due to the complexity of the control of the speed difference slip steering vehicle, it cannot be well timed according to the dynamic changes of targets and obstacles. Adjust the charged vehicle

[0006] (2) DDPG algorithm in reinforcement learning When the agent and the environment it interacts with are complex, the design of the reward function will also be difficult

At the same time, improper reward function settings will also make the actions output by the learning model unable to effectively solve the problem, greatly reducing the training effect

[0007] Difficulties in solving the above technical problems: (1) For wheeled differential slip steer vehicles, based on traditional path planning and trajectory tracking methods, there will be great uncertainty, and there is little room for improvement in control accuracy, making it difficult to cope with various an emergency

(2) When the DDPG algorithm in reinforcement learning solves complex environmental problems, it is difficult to design a reward function, which will lead to a large deviation between the training effect and the ideal situation

(3) During the training process, in order to obtain the real trajectory information, the target vehicle and obstacles must be loaded, which makes the calculation efficiency of the system slow down

Images