Hybrid Autonomous Navigation using Rapid Exploring Random Trees and Reinforcement Learning for Obstacle Avoidance

The planning of paths is a significant issue for applications in a variety of fields, including robotics and video games. For autonomous vehicles to be implemented in practice, navigation systems are essential. Although both traditional and contemporary control techniques have been investigated, Simultaneous Localization and Mapping (SLAM) has constraints brought on by its intricacy and sensor needs. Consequently, the use of Artificial Intelligence (AI) algorithms is becoming more and more popular. This study suggests combining Rapid Exploring Random Trees (RERT) for global path planning with an algorithm for Reinforcement Learning (RL) agent acting as a local planner for avoiding obstacles. In contrast to other techniques that just use point of interest or reinforcement learning for universal path planning, this method combines random trees with reinforcement learning to guarantee effective navigation in challenging environments. The suggested approach makes navigation more effective by mapping the surroundings and creating ideal global routes that are split up into intermediate routes, toward the objective. Results from experiments show that the method works well, especially when navigating through complex landscapes, and it offers encouraging developments for autonomous navigation systems.