Mobile robot collision avoidance planning method based on deep reinforcement learning in static environment

The invention belongs to the technical field of mobile robot navigation, and particularly relates to a mobile robot collision avoidance planning method based on deep reinforcement learning in a staticenvironment. The method comprises the following steps of: acquiring original data by using a laser range finder, taking the processed data as a state S of an A3C algorithm, constructing an A3C-LSTM neural network, taking the state S as network input, outputting corresponding parameters by the neural network through the A3C algorithm, and selecting actions executed by a mobile robot in each step through normal distribution by utilizing the parameters. According to the method, the environment does not need to be modeled, and the mobile robot successfully avoids obstacles in a complex static obstacle environment through a deep reinforcement learning algorithm. According to the method, a continuous action space model with bow turning constraint is designed, multi-thread asynchronous learningis adopted, and compared w