Active Perception and Safe Planning for Quadrotor Navigation in Unknown Environments

This paper is concerned with the quadrotor navigation problem by proposing an integrated framework for active perception and safe planning, enabling quadrotors to navigate in unknown environments while autonomously avoiding areas with higher obstacle density during the planning process to generate collision-free trajectories. The active perception component leverages the quadrotor's onboard sensors to collect environmental data and dynamically adjusts its heading and path. By integrating real time sensor feedback with an active perception path planning algorithm, it maximizes the visibility and sensing coverage of regions of interest, enhancing the perception of key environmental features. Concurrently, a risk-aware safe flight corridor (SFC) is generated by constructing convex polyhedron around accessible areas using environmental data, with the corridors dynamically adapted based on obstacle density and proximity. This framework combines the optimization of model predictive control (MPC) to ensure that the planned trajectory is stable and dynamically feasible, adhering to the quadrotor dynamics. The motion of yaw angle is also planned in this framework to improve the quadrotor's perception efficiency and path optimization quality. Simulation results demonstrate the system's ability to navigate in unknown environments efficiently while maintaining a high safety margin and adaptability to different scenarios.