A Path Planning Framework for Autonomous Navigation with Uncertain Obstacles
Zhao He, RongBing Zhang, YiNi Yin, ZhongZhi Ma · 2025
In autonomous navigation, ensuring safe and efficient path planning is challenging due to uncertain obstacles. Uncertainty in obstacle positions and velocities arises inevitably from limitations in perception technologies, sensor noise, and localization errors. Traditional obstacle avoidance methods often assume precisely defined obstacles, but such assumptions overlook real-world uncertainties. In this paper, we propose a robust path planning algorithm that models obstacles as Gaussian-distributed random variables, with boundaries defined by 3-sigma uncertainty ellipsoids, capturing 99.7% of possible obstacle states. An approximate analytical method is employed to calculate the distance from a point to an ellipsoid, allowing for the generation of collision-free trajectories within a nonlinear Model Predictive Control (NMPC) framework. Simulation results show that the algorithm achieves an error of less than 0.1% when navigating near obstacles, ensuring precise and reliable performance. This makes the method particularly well-suited for handling uncertain, dynamic environments in real-time applications.