Global optical flow-based estimation of velocity for multicopters using monocular vision in GPS-denied environments

Abstract Velocity estimation is a requisite for multicopters to guarantee flight stability and maneuverability. In this paper, a global optical flow-based approach is proposed for estimation of multicopter velocity using the off-the-shelf onboard sensors, including a downward-looking monocular camera, an Inertial Measurement Unit (IMU) and a sonar facing downwards in GPS-denied environments. The proposed method does not require any prior information or artificial landmark of the environment. First, a feature-based estimation method is proposed for richly textured surfaces while an image pyramids-based method extended for low-textured backgrounds to obtain the pixel motion. Then, combined with the estimated pixel motion, vision-based velocity estimation is developed, offering the visual observations. Furthermore, a classical linear Kalman filter for velocity estimation by fusing all the available information from sensors is derived. Finally, the proposed method is validated and demonstrated with a collection of synthetic data based on a Microsoft AirSim high-fidelity visual simulation platform as well as flight experiments with real data from a commercial DJI Matrice 100 quadcopter equipped with a Guidance visual sensing module. Simulation and experimental results indicate that the proposed method is capable of providing an accurate estimate of multicopter velocity in GPS-denied or confined environments.