Real-Time B-Spline Path Planning for Vision-Based Collision Avoidance

In this paper, we present a novel approach to real-time collision avoidance and path planning for unmanned aerial vehicles (UAVs) using basic camera inputs. Our system predicts the future trajectories of nearby flying objects and computes a feasible path to avoid collisions while maintaining progress toward a target destination. Unlike traditional radar- based methods, our solution requires only two parameters from the visual feed—bearing and pixel size—allowing for lightweight, real-time obstacle detection and avoidance. This approach benefits UAVs operating in shared airspaces with manned aircraft, where rapid, autonomous decision-making is essential for safety. Preliminary results show that our system can efficiently compute avoidance maneuvers and plan routes in real-time, even in dynamic environments.