A Reinforcement Learning based Path Planning Approach in 3D Environment

Optimal motion planning involves obstacles avoidance whereas path is the key to success in optimal planning. Due computational demands, most of algorithms can not be employed for real-time-based applications. Model-based reinforcement learning approaches have received particular recent past. Yet, such do deterministic output due randomness. In this paper, we investigate existing learning-based and propose an approach 3D environment. One a tree-based approach, other two are based on Q-learning approximate policy gradient, respectively. We tested preceding different simulators, each which consists set random that changed or moved dynamically. After analysing result computation time, concluded tree search provides highly stable results. However, time considerably higher than approaches. Finally, comparative results provided terms accuracy time.