Due to the complexity of huge, dynamic settings, mobile robots' path planning is a difficult issue to solve. This is because the robots must effectively attain their objectives while avoiding possible collisions with other robots or dynamic objects. Re-planning techniques, which recall a planning algorithm to look for an alternate route whenever the robot faces a conflict, are often used in situations with moving barriers in the way of conventional solutions. Nonetheless, route re-routing tactics often result in extra side trips. We present a learning-based method that uses spatial and temporal data from the surrounding environment to solve this problem. In this research, we survey the landscape of DRL techniques and navigational frameworks built on DRL. “Finally, we conduct a comprehensive comparison and analysis of the similarities and differences across four common use cases: local obstacle avoidance, indoor navigation, multi-robot navigation, and social navigation.” The evolution of DRL-based navigation is discussed next. At last, we talk about the problems with DRL-based navigation and some potential answers to those problems.

Mobile Robot Navigation; Obstacle Avoidance; Deep Reinforcement Learning”