Q-Learning for Path Planning in Complex Environments: A YOLO and Vision-Based Approach

Abstract

In order for robots to navigate through complex and ever-changing environments, avoid obstacles, and arrive at their destination as fast as possible, they require effective path planning algorithms. This paper discusses the integration of three methods for figuring out the mobile robot's shortest path through a difficult environment. Three different algorithms' features are combined to identify the mobile robot, identify obstacles, and calculate the robot's shortest path. You Only Lock Once (YOLO) algorithm is used to locate and identify robots in the workplace. YOLO CNN was trained with images taken from the simulation program to detect and locate the robot in the workplace. The rest of the objects in the workplace are identified as obstacles by the computer vision algorithm. After that, the locations of the robot and obstacles were used to determine the shortest path to reach the target point. The Q learning algorithm was used to determine the shortest path for the robot. Q-Iearning is a technique used in reinforcement learning algorithms. It is commonly used in adaptive path planning for autonomous mobile robots. Using all three algorithms to find the best path for the robot in the shortest amount of time. It has been confirmed that the proposed system operates in complex and changing environments. Simulations were used to test the system and verify its proposed design. The system can also be used in practical applications, like warehouses.