Optimal autonomous quadrotor navigation in an obstructed space

This paper presents an ambitious methodology of autonomous navigation for multirotor UAVs in obstructed environments. The strategy was formulated to provide the multirotor vehicles the capability to produce autonomously quasi-optimal and safe trajectories, although generally they have at their disposal limited computational resources on board. The problem is formulated in a model predictive control (MPC) architecture in which motion planning and trajectory tracking processes are solved separately as if they were stored in two different devices. The first process uses a spline-based motion planning approach to generate smooth and safe trajectories. At this step also a multirotor's simpified dynamic model and environment information are taken into account. The second process uses trajectory inputs, which are total thrust and attitude angle rates, to steer the multirotor during the flight. Both adequate time horizon and update frequency are chosen in order to account for disturbances and dynamics model mismatch. The methodology is validated by simulations and future work will include experimental tests in outdoor environment.