Composite Local Path Planning for Multi-Robot Obstacle Avoidance and Formation Navigation Sung-Gil Wee 1 , Yoon-Gu Kim 1# , Jinung An 1 , Dong-Ha Lee 1 and Suk-Gyu Lee 2 1 Wellness Convergence Research Center, DGIST Daegu, 711-873, South Korea # ryankim9@dgist.ac.kr 2 Department of Electrical Engineering, Yeungnam University Gyeongsan, 712-749, South Korea Abstract This paper proposes a composite local path planning method for multi-robot formation navigation with path deviation prevention using a repulsive function, A-star algorithm, and unscented Kalman filter (UKF). The repulsive function in the potential field method is used to avoid collisions among robots and obstacles. The A-star algorithm helps the robots to find an optimal local path. In addition, error estimation based on UKF guarantees small path deviation of each robot during navigation. The proposed method of composite local path planning is verified by the simulation results of the collective robot navigation because the system maintains a designated formation and performs a successful return to the assigned formation with effective obstacle avoidance under various experimental conditions. Keywords: Collision-free, Composite local path planning, Path deviation prevention, Path re-planning. 1. Introduction Recent advances in the fields of computation and intelligence are accelerating the development and practicality of multi-robot technologies, which include flexible multi-robot formations, collision-free path planning, and cooperative robots. In previous studies, most service robots were focused on enhancing their own intelligence and performing individual missions. However, recently, in order for special-purpose robots to effectively carry out complex and special assignments such as exploration, surveillance, rescue, manipulation, and other field applications, it is necessary that a multi-robot-based application scenario and strategy should be well- implemented through various types of technology integration, combining decentralized control, various sensors, robot intelligence, and so on [1, 2, 3, 4, 5, 6]. Autonomous navigation of robot intelligence has been studied in many previous works in the literature. With regard to path planning, the potential field has been employed to apply the virtual forces generated on a robot by using the energy magnitude working on the system. The robot finds a path to avoid obstacles using the potential field just like the reciprocal action of magnets [7, 8]. One of the graph search algorithms, the A-star algorithm, is also widely used in path planning [9]. However, most of the local path planning algorithms have inevitable limitations such as local minima problems. Koditschek [10] et al. developed a local minima free potential field method to overcome this type of problem; Chang [11] et al. suggested a path planning algorithm based on the potential field and the Voronoi diagram for a hybrid path planner that fulfills both map building and driving simultaneously. In addition, Carpin [12] et al. proposed a dynamic obstacle avoidance algorithm in which robots following a leader robot avoid dynamic obstacles using a decentralized control system. In this paper, we discuss the effective movement of a cluster of multiple robots as they attempt to find the shortest path and to avoid obstacles or other robots without any collisions. We propose a method of composite local path planning for multi-robot formation navigation with path deviation prevention achieved by using a repulsive function, the A-star algorithm, and an unscented Kalman filter (UKF). The repulsive function in the potential field method is used to avoid collisions among robots and obstacles. The A-star algorithm helps the robots find an optimal local path. In addition, next step position estimation based on UKF reduces deviation from the tracking path of each robot during navigation. The proposed method of composite local path planning is verified because the system maintains a designated formation and performs a successful return to the assigned formation with effective obstacle avoidance under various experimental conditions. 2. Functional Algorithms for Composite Local Path Planning This chapter describes the functional algorithms used for composite local path planning: a repulsive function of the ACSIJ Advances in Computer Science: an International Journal, Vol. 4, Issue 5, No.17 , September 2015 ISSN : 2322-5157 www.ACSIJ.org 61 Copyright (c) 2015 Advances in Computer Science: an International Journal. All Rights Reserved.