Real Time Path Planning of Robot using Deep Reinforcement Learning

• Published in: IFAC-PapersOnLine Vol. 53; no. 2; pp. 15602 - 15607
• Main Authors: Raajan, Jeevan, Srihari, P V, Satya, Jayadev P, Bhikkaji, B, Pasumarthy, Ramkrishna
• Format: Journal Article
• Language: English
• Published: Elsevier Ltd 2020
• Subjects: Adaptation in Autonomous Vehicles; Autonomous Vehicles; Learning; Path Planning; Q-Learning; Reinforcement Learning; Trajectory; Learning; Autonomous Vehicles; Path Planning; Reinforcement Learning; Q-Learning; Trajectory; Adaptation in Autonomous Vehicles
• ISSN: 2405-8963; 2405-8963
• DOI: 10.1016/j.ifacol.2020.12.2494

This paper considers finding a path in real time for a robot from the given initial position to the goal position. The environment is assumed to be mapped (known completely) and the resulting path should avoid all the obstacles, both static and dynamic in the mapped environment. The robot’s (agent) dynamics is assumed to be discrete LTI with process noise and is controlled with a finite set of inputs. An MDP formulation and a solution based on Deep Reinforcement Learning framework are presented for the problem. Numerical experiments are performed for the proposed method using Deep Q-Network algorithm and the results are compared with the state of the art sampling based path planning algorithms for both static and dynamic environments. It is shown that even though the proposed algorithm provides a sub-optimal path, the computational time is shown to be significantly faster compared to the traditional methods of path planning.