Estimating and reacting to forces and torques resulting from common aerodynamic disturbances acting on quadrotors

Abstract Quadrotors are increasingly expected to perform a wide variety of tasks that put them in close proximity to other objects and surfaces in the environment (including other quadrotors), where they are often subject to significant external forces and torques resulting from aerodynamic effects. We present an algorithm – based on an Unscented Kalman Filter – that estimates such forces and torques without making assumptions about their source, allowing us to bypass much of the complexity involved in modeling how wind currents interact with quadrotor dynamics. Furthermore, our algorithm does not rely on special sensors, making it suitable for commercial systems where payload and add-on capabilities are limited. Via experiment we show that the estimation algorithm can be used in conjunction with controls and machine learning for detecting and avoiding downwash and walls, and for tracking wind from a fan. We also show that the algorithm is sensitive enough to measure even small changes in force and torque.