Path planning of mobile robot in complex environment based on improved Q-learning algorithm

Abstract

Path planning is one of the key technologies of mobile robots. This paper applies the Q-learning algorithm to path planning of mobile robots. To solve the blindness of the mobile robot in exploring the environment, this paper combines RRT algorithm and improves it to increase the goal-orientated performance of the robot in exploring the environment. To overcome the slow convergence rate of the classical Q-learning, the exploration rate, discount rate, learning rate and other parameters in Q-learning are modified. A modified ant colony optimisation (ACO) algorithm is designed to make the robot consider the information of other robots and improve the precision of action decision. The hierarchical idea in the grey wolves optimisation algorithm is introduced into the modified ACO algorithm to realise the dynamic regulation of pheromone. Finally, the modified Dyna-2 algorithm is designed to enhance the generalisation of the Q-learning algorithm. The potential of the proposed hybrid intelligent algorithm is proved by the path planning experiments in two static complex environments.