Abstract
This work addresses the velocity and attitude estimation of a quadrotor using motor speeds and inertial measurement unit (IMU) data. This result is obtained through a novel filter that employs a first order drag model which allows the velocity to be observed through the drag forces acting on the quadrotor. These forces are measured with the IMU. Compared to our previous contribution [1], decoupling the attitude into a yaw-tilt convention results in the tilt and velocity in the body-fixed frame being independent of the yaw. Thus, the velocity and attitude estimates are no longer affected by yaw drifts obtained from dead-reckoning procedures. The camera available on the vehicle can optionally be employed to recover the yaw in a computationally efficient way without requiring RANSAC, instead using the estimated velocity and attitude from the filter. The IMU is concurrently used to estimate the thrust, linear drag coefficients, and accelerometer biases. Experimental results on a quadrotor platform show the feasibility of the proposed approach.
Funding Information
  • Qualcomm Research
  • Army Research Laboratory (W911NF-17-2-0181)
  • Office of Naval Research (N00014-07-1-0829, N00014-14-1-0510)
  • Army Research Office (W911NF-13-1-0350)
  • National Science Foundation (IIS-1426840, IIS-1138847)

This publication has 12 references indexed in Scilit: