Mobile robot navigation method based on improved Q-learning algorithm

Abstract

In recent years, autonomous navigation of mobile robots has become one of the hot topics in research. An improved Q-learning method (IQL) combining action selection strategy and Q function update method is proposed. A new exploration strategy of staged $\varepsilon$ greedy exploration is used in IQL. In addition, the idea of backtracking is used to update the Q value to speed up the convergence speed. Experiments conducted in a grid environment show that compared with the classic algorithm, the algorithm converges faster and the program execution time is shorter.