RL-LSTM Enhanced Hybrid Navigation: An Intelligent APF Approach for Differential Wheel Drive

This study introduces a hybrid path-planning method for a two-wheel mobile robot by integrating artificial potential field guidance, memory-based directional smoothing, and reinforcement learning with an interactive graphical interface. Users can define initial and target positions and obstacles in a two-dimensional space. The robot navigates safely toward the target by combining force-based attraction and repulsion, past motion data, and adaptive decision-making through learning. The system enables real-time training, path visualization, and curve smoothing. This approach improves movement through complex environments by combining classical navigation techniques, memory-driven behavior, and intelligent learning for better flexibility and obstacle avoidance.