Path planning using deep q-networks for mobile robot navigation [document multimédia] / -

Abstract

Efficient navigation of mobile robots in static environments remains a key challenge in robotics. In this study, we use a **tracked mobile robot** model to test an intelligent path planning approach based on **Deep Q-Learning (DQN)**. Unlike classical methods like A and Dijkstra that depend on predefined maps, DQN allows the robot to learn how to navigate by interacting with its environment. This learning capability helps the robot avoid obstacles and find optimal paths without prior knowledge. By combining DQN with traditional algorithms, we enhance adaptability and decision-making. Simulations in MATLAB showed that the proposed method improved navigation efficiency in grid- based static environments.