Path planning optimization of mobile robots using reinforcement learning

Abstract

Autonomous mobile robots are growing in everyday life. They can navigate in an uncontrolled environment without the requirement for electro-mechanical guidance devices. It has an array of sophisticated sensors that enable them to understand and infer their environment, which helps them to perform their task in the most efficient manner and navigate around obstacles. Navigation is very important in any mobile device. It requires a map of the environment, continuous localization of the robot within that environment using a multitude of sensors, and path finding algorithms that utilize knowledge of the map and the dynamics of the robot to generate paths that can be feasible traverse. newlineIn path planning, autonomous mobile robots should track the best collision-free path from a start position to a target position without colliding with any of the obstacles in the environment. A good path planning algorithm can lessen the wear and capital investment of the mobile robot. Several path planning algorithms are there to find the shortest path from the start location to the goal location. Each algorithm has its advantages and disadvantages based on the applications. Therefore, an effective framework is required to overcome these issues in path planning of mobile robots. The objective of the thesis is to accomplish safe and efficient travel from one location to another, without colliding with obstacles. newlineThis research compares the node-based and sampling-based algorithms like A*, Artificial Potential Field (APF), Rapidly Exploring Random Tree (RRT), Bidirectional Rapidly Exploring Random Tree (B-RRT), and Probabilistic Road Map (PRM), and analysed in MATLAB R2019a in various static environments. newline