Efficient and Lightweight Multi-Constraint Online Planning for Aerial Tracking of Random Targets in Complex Environments

Tracking randomly moving targets autonomously by quadrotors in complex environments has always been a significant challenge. Most existing research focuses on unobstructed scenarios and fails to provide comprehensive and computationally efficient solutions. In this paper, we present an online trajectory planning framework. By using polynomial trajectory representations, we formulate various constraints to achieve robust tracking of dynamic objects in intricate environmental situations. First, we developed a simple and efficient object detection and motion prediction method. Then, we deeply analyze the analytical requirements for effective tracking. Next, we propose an intuitive augmented-awareness path finding approach. Finally, by considering trajectory smoothness, obstacle avoidance, and dynamic feasibility, we establish a spatio-temporal joint optimization framework that can operate efficiently at the millisecond level. Notably, the obstacle avoidance strategy we adopt is designed to reduce computational cost. Experimental results show that our method can handle more complex scenarios well while maintaining lower computational requirements.