Swarm Robotics for Area Exploration Using Q-Learning

Swarm robotics is an approach inspired by natural systems, where tasks are efficiently accomplished by distributing work among multiple agents. One of the key tasks addressed in this work is area exploration. By leveraging decentralized coordination principles and a reinforcement learning technique called Q-learning, this study focuses on overcoming challenges associated with area exploration. A swarm robotics simulation is developed to map unknown environments using Simultaneous Localization and Mapping (SLAM). The TurtleBot3 Burger is initialized in a simulated environment using ROS Noetic and Gazebo. The trained robots aim to collaboratively map unknown environments while efficiently a voiding o bstacles. A c ustom Q- learning algorithm is implemented for obstacle avoidance tasks. By employing GMapping, a SLAM algorithm, the robots gen- erate individual 2D maps of their surroundings and share this information without conflicts. T hese m aps a re t hen m erged as accurately as possible to create a comprehensive map of the environment. Real-time mapping by the robots is visualized in Rviz, demonstrating the potential of swarm systems. This work effectively provides a collaborative map of the environment, con- tributing to advancements in swarm intelligence for cooperative robotic systems.