Dynamic obstacle avoidance based on multi-sensor fusion and Q-learning algorithm

Owing to the shortness from the single-function obstacle avoidance sensors itself, and the low efficiency brought by obstacle uncertainty under the dynamic environment, a solution is proposed for the problem that mobile robot would automatically avoid in the static and dynamic environment. In this paper, the feature level fusion of laser sensor and sonar sensor is used to make up for the shortcomings of single laser or single sonar. Then it is more flexible and convenient to avoid obstacles by adding the action angle of state transition in learning algorithm. Then, by adding the action angle of the state transition to the Q learning algorithm, the obstacles are more flexible and convenient. The validity of the proposed method is verified by simulation, and an example of a robot is given to illustrate the effectiveness of the proposed method.