Trained FLS and Reduced GLVD for Motion Planning in Dynamic Unknown Environment
msra
摘要
Motion planning of robots in unknown environments is being more and more studied last years. We consider in this paper the problem of polygonal robots navigation in unknown environment composed of arbitrary fixed and moving obstacles. We propose a new combined navigation technique: the Fuzzy Logic (FL) method, and a Reduced Generalized Local Voronoi Diagram (RGLVD) method. We designed our own fuzzy controller and we trained it using MAGAD-FS method (Multi-Agent Genetic Algorithm for the Design of Fuzzy Systems). The obtained results are presented and are compared with the navigation using only our trained FLS and Reduced GLVD. We note that our combined approach is more efficient on terms of saving time and optimizing distance. I. INTRODUCTION Since few years, motion planning of mobile robots in dynamic and unknown environments represents the most complex problem of navigation in robotic issues. Several methods have been proposed to fill the problem of path planning. In fact a survey of techniques used for navigational planning is given in (2, 16, 19, 26, 22); classical ones, called global methods, require full environment knowledge. Some of those are: the potential field, the visibility graph (3, 4), the cellular decomposition (6) and the Voronoi diagram (5, 8, 14). The reactive methods, called local methods, are reactive and real time, because most of them take into account the local information extracted from robot sensors. They are based on intelligent techniques like neuronal networks (9), fuzzy logic system (FLS) (7, 15, 17, 22, 24, 25). The generalized local Voronoi diagram (GLVD) introduced by Makhovic (12) is a local method extended from the global one. As result of both techniques, global and local, the apparition of hybrid methods taking the advantages of each one. We can mention for example the combination of fuzzy logic with the electrostatic potential field (10, 13), fuzzy logic with the visibility graph (22), fuzzy logic with neural networks (23). In this paper we propose a new navigation method for polygonal robots in unknown environments using a reduced version of GLVD (RGLVD) and FLS, which are combined in order to have an efficient and a fast algorithm allowing motion planning and navigation avoiding any static and dynamic
更多查看译文
AI 理解论文
溯源树
样例
生成溯源树,研究论文发展脉络