Planning Aggressive Drone Manoeuvres

Abstract

This paper addresses the problem of performing aggressive manoeuvres by using multirotor vehicles that include passing through any specific point within the full state space of the vehicle. To this end, the design of optimal trajectories considers the dynamical model of the vehicles by numerically integrating it backwards in time, in the manifold where the dynamics evolve, and dividing the manoeuvres into three distinct phases to accommodate any combination of initial, desired, and final states. In the first phase, the vehicles fly from an initial to a launch configuration to achieve the necessary momenta to reach the desired one in the second phase. To ensure the feasibility of executing the second phase, the relation between snap and body torques is exploited by commanding the vehicles to track geodesic curves on SO(3) during the backwards integration. The vehicles are then driven to a final configuration in the third phase. Most existing solutions to execute aggressive and precise manoeuvres with these rotorcraft focus either on the attitude control problem, leaving the position in open-loop, or use different controllers for different sections of the manoeuvre. In this work, a single tracking controller is considered to validate the proposed trajectory planning strategy in a realistic simulation environment, which involves the PX4 firmware, and in a controlled experimental setup. The results demonstrate that accurate tracking of the designed trajectories enables the vehicles to perform 360-degree loops at great speed and manoeuvres that facilitate the exchange of a parcel between two multirotor vehicles during flight.