Robot Navigation With Reinforcement Learned Path Generation and Fine-Tuned Motion Control

In this letter, we propose a novel reinforcement learning (RL) based path generation (RL-PG) approach for mobile robot navigation without a prior exploration of an unknown environment. Multiple predictive path points are dynamically generated by a deep Markov model optimized using an RL approach for the robot to track. To ensure safety when tracking the predictive points, the robot's motion is fine-tuned by a motion fine-tuning module. Such an approach, using a deep Markov model with RL algorithm for planning, focuses on the relationship between adjacent path points. We analyze the benefits of our proposed approach and show it is more effective and has higher success rates than the RL-based approach DWA-RL (Patel et al. 2021) and a traditional navigation approach APF (Chen et al. 2021). We deploy our model on both simulation and physical platforms and demonstrate our model performs robot navigation effectively and safely.