Navigating the Unseen: Agile Path Planning in the Presence of Dynamic Obstacles for Mobile Robotics

Abstract

Path planning in dynamic environments is an important capability for autonomous robot navigation. As obstacles move unpredictably, robots must plan non-colliding paths in real time while considering factors like optimality, computability, and safety. This paper presents a methodology for robot path planning among moving obstacles towards a predefined goal location. A current state estimator tracked the robot and obstacle positions and velocities. Then a model predictive path planner leveraged the state estimate to generate a collision-free trajectory by solving a constrained finite-horizon optimal control problem. Trajectory optimization objectives included minimizing path length, smoothness, and avoidance distance from obstacles. Finally a model predictive tracking controller issued wheel velocity commands to closely track the trajectory while satisfying system dynamics and input constraints. Hardware in the loop simulation demonstrated reliable navigation around randomly moving spherical obstacles to goals up to thirty meters away. Remaining challenges for real world implementation include state estimation noise, unpredictable movements, computational expense, and dynamic constraints. The paper contributes an adaptable framework for safe robot navigation in complex, uncertain environments.