Perception and Navigation for Autonomous Rotorcraft

To enable a quadrotor to navigate autonomously, a set of interdependent requirements must be addressed, including control, state estimation, map learning, and mission planning. Autonomous navigation requires stabilization of the vehicle. To stabilize the vehicle, the state of the vehicle must be known. In GPS-denied environments, state estimation is calculated using simultaneous localization and mapping algorithms. Once the state of the vehicle is known, depending on the planned mission, the vehicle can move from one point to another using a path planning algorithm. This paper proposes an integrated solution for all these requirements. The paper studies the possibility of using state-of-the-art sensing technologies to perform perception in GPS-denied environments and achieve autonomy with minimum human intervention. Multiple tests, performed in simulated and real-world environments, show the effectiveness of the proposed solution.