Autonomous Navigation for Mobile Robots in Dynamic Environments Based on Deep Reinforcement Learning

Achieving efficient and socially compliant autonomous navigation in dynamic environments is a significant challenge for mobile robots. Traditional algorithms lack sufficient dynamic adaptability and struggle to handle changes in complex scenarios. While deep reinforcement learning methods show certain advantages in dynamic environments, their effectiveness diminishes as crowd size increases. To address this, we propose a socially compliant deep reinforcement learning model with a well-designed negative exponential decay reward function. The model dynamically adjusts the robot’s speed and direction to achieve efficient obstacle avoidance and autonomous navigation. Experimental results demonstrate that the proposed approach exhibits good performance in autonomous navigation within dynamic environments and crowds.