This project is part of my MSc thesis research. I developed an OpenAI Gym–based simulation environment consisting of a robot, dynamic pedestrians, and a goal location. The objective of the robot is to reach the destination without colliding with pedestrians whose coordinates change over time. The model is trained using Deep Reinforcement Learning with a Hybrid Proximal Policy Optimization (H-PPO) algorithm.
The repository RLAutonomousRobotNavigation2 (https://github.com/ddharshan/RLAutonomousRobotNavigation2) contains the simulation environment, including the robot model, pedestrian models, and the goal representation.
The repository RLRobotTraining2 (https://github.com/ddharshan/RLRobotTraining2) contains the Deep Neural Network architecture used to train the robot within this environment.
For a detailed explanation of the methodology and results, please refer to my thesis publication: https://papers.ssrn.com/sol3/papers.cfm?abstract_id=5178305
Pls feel free to reach me, if u get any doubts - [email protected]
Developed by Dhivyadharshan Seetharaman
MSc in Industrial Automation
Original implementation and training by the author.
If you use this work in research, academic projects, or publications, please cite:
@misc{Ddharshan_RLAutonomousRobotNavigation2,
author = {Ddharshan},
title = {RLAutonomousRobotNavigation2: Reinforcement Learning-Based Autonomous Robot Navigation},
year = {2022},
publisher = {GitHub},
journal = {GitHub repository},
url = {https://github.com/ddharshan/RLAutonomousRobotNavigation2}
}