Please use this identifier to cite or link to this item:
Title: Modular Robotic Reinforcement Learning Platform for Object Manipulation
Author(s): Kurumbaparambil, Vishnudev
Rajanayagam, Subashkumar
Twieg, StefanLook up in the Integrated Authority File of the German National Library
Granting Institution: Hochschule Anhalt
Issue Date: 2024
Language: English
Abstract: The field of robotics and autonomous systems has witnessed significant advancements in recent years, with an increasing focus on enhancing the capabilities of robotic agents through RL. This project centres around applying Reinforcement Learning techniques to do object manipulation tasks with a specific focus on making a robot arm reach a target object. Using a combination of Gazebo and Robot Operating System (ROS) environments, the robot arm is trained using custom OpenAI Gym environments to simulate the task. The primary objective involves positioning the end-effector of the robot arm close to a designated object and overcoming challenges such as self-collisions during movements. Various iterations of RL training, including different reward logics, curriculum learning approaches, and fine-tuning parameters, are explored to refine the decision-making capabilities of the agent. The training process of Curriculum Learning involves a phased approach, starting with basic movements and progressing to more complex tasks, demonstrating improved performance. However, challenges such as prolonged training times and uncertainties in arm behaviour persist. The project highlights the complexities inherent in designing effective RL strategies for robotic systems and stresses the need for further research to enhance computational efficiency and reliability in realworld applications.
Open Access: Open access publication
License: (CC BY-SA 4.0) Creative Commons Attribution ShareAlike 4.0(CC BY-SA 4.0) Creative Commons Attribution ShareAlike 4.0
Appears in Collections:International Conference on Applied Innovations in IT (ICAIIT)

Files in This Item:
File Description SizeFormat 
4_2_ICAIIT_2024_Part_2_paper_14.pdf2.07 MBAdobe PDFThumbnail