A Novel Approach for Multiple Mobile Robot Path Planning in Dynamic Unknown Environment

Multiple mobile robot path planning in dynamic unknown environment is challenging. In this paper a novel artificial potential field (APF) based approach is proposed. In this approach, a robot was pulled to the goal position with the influence of the goal and static obstacles. When two or more robots meet at the range of safe distance, those robots was prioritized according to the priority of the task they undertake, velocity and their size. The robot with higher priority will pass the intersection first. If they have the same priority, they will generate a priority order randomly to each other. By this way, they will avoid collision between each other. Simulation results are provided to validate the concepts and analyses are made to show the effectiveness of the approach

I. Introduction

Multi robot path planning is one of the fundamental problems in multiple mobile robotic systems. Many researchers have studied this problem and proposed many different approaches. The approach based on genetic algorithm for multi robot path plan is proposed in [4] [6] [15]. A gated dipole based approach is proposed for path planning of a multi-robot system, where the cooperative control strategy is based on the multi-agent systems [7]. A decentralized approach to the conflict-free motion planning for multiple mobile robots is proposed in [8]. A distributed and optimal motion planning algorithm is proposed in [9], in which the D* search method is applied for path planning and velocity planning to optimize robot motion. Conte [10] proposed an approach using cell decomposition method for Cspace representation. Maren Bennewitz [11] proposed an approach using hill-climb algorithm to optimize the robot path. Yan et.al. [13] proposed a multi-robot real time path planning approach. In this approach, the robot just consider the collision problem with the robot who has higher priority. GAKUHARI [14] proposed a navigation approach for non-holonomic mobile robots in a dynamic environment. In the approach, the information of the environment and robots are fed back to the system in real time and global path planning is cyclically executed. The proposed system can be applied to dynamic environment and a reliable dead lock-free navigation of multiple robots. These proposed approaches are mainly used in known environment which means the robots have prior knowledge about the environment. Multi-robot path planning in unknown environment is rarely reported in literatures.