COMPARATIVE STUDY ON SAMPLING-BASED PATH PLANNING ALGORITHMS
Abstract - With growing research in the field of mobile robots under various domains, path planning of 2 wheeled differential-drive robot is constantly evolving with new methods of planning and executing the trajectory plan. The optimal path of a mobile robot is determined by considering path length, collision-free space, computation time, and computation memory size. This paper presents a comparison of various rapidly exploring random tree (RRT) algorithms for path planning in a ROS-based environment, using Rviz for the simulation of the path. The studying algorithms include RRT, RRT*, Informed RRT*, and RRT-Connect. For analysis of these algorithms, a global planner plugin is created for the move-base navigation stack. The performance of the algorithms will be evaluated based on the key factors mentioned earlier. The study will identify the most suitable algorithm for path planning in complex environments with obstacles and narrow passages. The study also analyzes the limitations of algorithms concerning their environment. The proposed global planner plugin can be integrated into the ROS navigation stack for the autonomous navigation of 2 wheeled differential-drive robot in Gazebo simulation. The study concluded on the analysis that RRT-Connect would be a better planning algorithm in comparison to the other three variants as it shows an 83% percent better result on a median concerning the key performance indicators.
Keywords - ROS, RRT, RRT*, Informed-RRT*, RRT - Connect, Path Planning.