Enhancing Mobile Robot Navigation in TurtleBot3 Burger: A ROS-Enabled Approach Focusing on Obstacle Avoidance in Real-World Scenario

Abstract

In robotics, reducing human interaction in industrial environments can be achieved through intelligent mechanisms that enable effective manoeuvring and processing of various tasks. This research aims to improve the framework that allows diverse robots to interact and cooperate to complete tasks in a heterogeneous robotic environment. The robotic system's compatibility with ROS (Robot Operating System) and Python libraries makes it highly versatile for mobile robot development. The main goal is to enable effective obstacle avoidance by the robotic system when an obstacle is introduced to the environment, enhancing the robot's navigation precision. This work focuses on hardware implementation alongside machine learning algorithms. Neural network-based approaches train the system to detect and avoid obstacles effectively. This includes training the robot to navigate in a known environment when an obstacle is introduced to the system. A laser range finder in this system can scan its surroundings to identify obstacles, which is crucial for obstacle avoidance. Deployed in real-world scenarios, the system demonstrates its autonomous obstacle scanning using Light Detection and Ranging (LiDAR) and the ability to execute manoeuvres and avoid collisions. Further, the robot is trained to move in an environment with dynamic obstacles using Q-learning, which enables the system to learn obstacle avoidance behaviour through feedback from its interactions with the environment. The integration of these machine learning approaches has shown promising results, with the robotic system successfully identifying obstacles and executing manoeuvres to avoid collisions, highlighting the potential of machine learning in enhancing the obstacle avoidance capabilities of mobile robots.