Adaptive filtering of tracking camera data and onboard sensors for a small helicopter autopilot

Abstract

A navigation system was designed for an autopilot of a small helicopter. It contains two remote web cameras and an onboard sensor with three gyroscopes and three accelerometers. The camera positions are unknown. The gyroscope and accelerometer measurements have unknown random bias with a slow trend. For estimation of these parameters an adaptive filtering system with the extended Kalman filter is developed. Tuning of the filter parameters is made online by minimization of the high frequency part of the helicopter velocity estimate. A new method is also developed for smoothing of the big oscillations forced by rotation of the main rotor that appear in the gyroscope measurements of the angular velocity. The method is based on equations from the helicopter motion model that give a relation between linear accelerations and the Euler angles estimated by integration of the angular velocity measurements.