Anti-locking brake control method and system based on reinforcement learning

Abstract

The invention discloses an anti-locking brake control method and system based on reinforcement learning. The method comprises the following steps of extracting key parameters; quantizing the extractedkey parameters, and defining a range of the parameters; establishing a reinforcement learning module including a reward function; defining a wheel speed change range reward function value and a braketime change range reward function value in the reward function, multiplying the wheel speed change range reward function value and the brake time change range reward function value and outputting a result; inputting the output result to the reinforcement learning module for training; and using the trained reinforcement learning module for controlling a vehicle for preventing brake locking. The defect of unreasonably defining the reward function in existing reinforcement learning is overcome, the reinforcement learning algorithm with the new defined reward function is applied to the anti-locking brake control system, and the performance more excellent than that of a traditional algorithm is achieved.

Description

Technical field [0001] The invention relates to the technical field of automobile automatic driving, in particular to an anti-lock braking control method and system based on reinforcement learning. Background technique [0002] In modern automobile products, the anti-lock brake system can ensure that the car prevents the wheels from locking during emergency braking, thereby stabilizing the body and shortening the braking distance. It is the standard configuration of the car. With the development of artificial intelligence, it has become possible to use artificial intelligence technology to achieve anti-lock brakes, which theoretically can achieve better performance than traditional algorithms. [0003] As an important direction of artificial intelligence, reinforcement learning is more suitable for dealing with serialization problems. The anti-lock braking operation of a car during emergency braking meets this feature of reinforcement learning, so it is feasible to use reinforcemen...

AI Technical Summary

Problems solved by technology

The basic idea of ​​reinforcement learning is: the agent makes an action in a specific environment, the environment gives feedback rewards according to the action, the agent adjusts the action according to the reward, and expects to obtain a higher reward. From this idea, it can be seen that the reward It is very important to the learning of the agent, and in the algorithm, the reward is realized by the defined reward function, but the current development level of deep reinforcement learning technology shows that it is suitable for the task of dealing with simple scenes. On the one hand, the development of the algorithm has no breakthrough Progress, on the other hand, the definition of the reward function is too simple, the learning content of the agent is very limited, and the complex definition of the reward function will make it difficult for the agent to distinguish the reward difference in different situations, so how to define the function is very important

Images