MODELLING AND SIMULATION OF A PATH-PLANNING MOBILE ROBOT FOR SUSTAINABLE AGRICULTURAL APPLICATIONS

This study presents a simulation of a path-planning mobile robot navigation system for agricultural applications, integrating the A* algorithm and reinforcement learning techniques. The system aims to create a robust and efficient system capable of autonomously navigating complex agrarian environments. The A* algorithm is used for initial path planning, while reinforcement learning refines the path in real-time to handle dynamic obstacles and environmental changes. The simulation environment, developed using ROS and Gazebo, replicates a realistic agricultural field with varying terrains and obstacles. The results show that the path efficiency of the A* Algorithm is 120m, Reinforcement Learning is 115m, and the Hybrid Approach is 110, with time taking 45, 42, and 40 seconds respectively. Also, the obstacle avoidance metric shows that the hybrid approach has fewer collisions during the simulation, with a 98% obstacle avoidance rate compared with the reinforcement learning of 95% and the A* algorithm of 90%. For the adaptability to changing conditions metric for A* Algorithm, Reinforcement Learning, and Hybrid Approach, the performance under varied terrain is 80%, 85%, and 90%, respectively. The hybrid approach of A* and reinforcement learning significantly enhances the robot's navigation capabilities, demonstrating superior performance across all evaluated metrics. This research contributes to the advancement of autonomous agricultural robots, providing a foundation for future development and field trials. Integrating advanced path-planning algorithms in agricultural robotics holds significant potential for improving productivity and efficiency in farming operations.