Perception-Aware Trajectory Planning of Quadrotor in SE(3)
Shibo Xu, Juntong Qi, Mingming Wang, Chong Hu Wu · 2023
Recent achievements in trajectory planning enable quadrotors to navigate autonomously in unknown complex environments. However, active adjustment of the limited field of view(FOV) remains a major challenge in trajectory planning, and existing methods rarely consider quadrotor visibility into unexplored spaces. This paper proposes a Perception-aware local trajectory planning framework considering limited FOV constraints, which enhances the quadrotor's observation of waypoints on the trajectory. In contrast to existing work, this paper designs a path search with an optimistic FOV constraint and matches the observed relations between waypoints. Trajectory optimization is performed in SE(3) and takes into account the metric of limited FOV constraints and avoidance of obstacle occlusions between waypoints with observation relationships. These metrics are formulated as differentiable cost functions to actively observe future trajectories. Physical simulation and actual flight experiments verify the effectiveness of the proposed method.