Robot Navigation With MCTS for Human-Aware Waypoint Planning

The patent implements dynamic path planning by continuously updating the costmap and re-evaluating candidate waypoints based on real-time pedestrian movement patterns. The system transitions from static obstacle treatment to dynamic prediction of human motion trajectories, allowing the robot to adapt its navigation strategy as pedestrians move through the environment.

The patent performs preliminary evaluation of candidate waypoints by predicting future pedestrian positions and interactions before the robot commits to a path. The Monte Carlo Tree Search algorithm simulates multiple possible future scenarios to pre-assess the safety and optimality of each candidate waypoint, enabling proactive rather than reactive navigation decisions.

The patent divides the computational workload into distinct segments handled by different processors: one processor dedicated to neural network computation for pedestrian behavior prediction, another for Monte Carlo Tree Search algorithm execution, and a third for path optimization. This segmentation allows parallel processing of independent computational tasks, reducing overall computation time while maintaining modular system architecture.

The patent employs multiple processors that each perform specialized functions but can be integrated into a unified navigation system. The neural network processor handles behavior prediction, the MCTS processor handles path evaluation, and the optimization processor handles final path selection, with all processors working together through shared data structures and communication protocols to achieve comprehensive navigation functionality.