Hybrid rule-based motion planner for mobile robot in cluttered workspace

Abstract

Motion planning problem is an active field in robotics. It is concerned with converting high-level task specifications into low-level descriptions of how to move and provides a feasible sequence of movements that avoid obstacles while respecting kinematic and dynamic equations. In this work, new planners are designed with the aim of developing an efficient motion planner in a heterogeneous, cluttered, and dynamic workspace. The planners are composed of two layers, and they use a rule-based system as a guidance. The first layer uses exact cell decomposition method, which divides the workspace into manageable regions and finds the adjacency information for them. The second layer utilizes rapidly exploring random tree algorithm RRT that finds a solution in a cluttered workspace. The adjacency information of the free cells and the exploration information that is provided by RRT are combined and utilized to help the planners classifying the free regions and guiding the growth of RRT trees efficiently toward the most important areas. Two types of the planners are proposed, the first one uses adviser that pulls the trees' growth toward the boundary areas between explored and unexplored regions, while the adviser of the second planner uses the collision information and fuzzy rules to guide the trees' growth toward areas that have low collision rate around the boundaries of explored regions. The planners are tested in stationary as well as in changed workspace. The proposed methods have been compared to other approaches and the simulation results show that they yield better results in terms of completeness and efficiency.