PRRT: An Intelligent Planning Algorithm with Memory

Abstract

This paper proposes a practically executable path planning method, namely, Pheromones-RRT(PRRT), for robots with a large joint range in a complex environment. To inter-activate with the real world, the point cloud is utilized as the scene information, while for sampling, the pheromones approach is designed to describe the pheromone content carried by each sampling point. During the sampling process, random sampling nodes are performed with a probability of ε, or those nodes with the highest pheromone content in the current sampling tree are selected with a probability of 1-ε and sampled in their neighborhood. To avoid the local minimum problem, the concept of pheromone volatile factor (PVF) is proposed, while in the expansion, double trees are also generated by PRRT in both cartesian and configuration spaces to improve the speed of the algorithm. The pheromone accumulation enables PRRT to certain learning abilities, reducing the randomness of the sampling process. Simulation results show that the proposed method can effectively plan an optimal obstacle avoidance path with satisfactory performances as compared with the RRT-Connect.