Why Change Your Controller When You Can Change Your Planner: Drag-Aware Trajectory Generation for Quadrotor Systems

Motivated by the increasing use of quadrotors for payload delivery, weconsider a joint trajectory generation and feedback control design problem fora quadrotor experiencing aerodynamic wrenches. Unmodeled aerodynamic dragforces from carried payloads can lead to catastrophic outcomes. Prior workmodel aerodynamic effects as residual dynamics or external disturbances in thecontrol problem leading to a reactive policy that could be catastrophic.Moreover, redesigning controllers and tuning control gains on hardwareplatforms is a laborious effort. In this paper, we argue that adapting thetrajectory generation component keeping the controller fixed can improvetrajectory tracking for quadrotor systems experiencing drag forces. To achievethis, we formulate a drag-aware planning problem by applying a suitablerelaxation to an optimal quadrotor control problem, introducing a tracking costfunction which measures the ability of a controller to follow a referencetrajectory. This tracking cost function acts as a regularizer in trajectorygeneration and is learned from data obtained from simulation. Our experimentsin both simulation and on the Crazyflie hardware platform show that changingthe planner reduces tracking error by as much as 83%. Evaluation on hardwaredemonstrates that our planned path, as opposed to a baseline, avoids controllersaturation and catastrophic outcomes during aggressive maneuvers.