Real-time parameter estimation for mini aerial vehicles using low-cost hardware

摘要

This work describes the design and implementation of a real-time aerodynamic parameter estimation algorithm on a small remotely piloted aircraft. The Extended Kalman Filter (EKF) is adapted for aerodynamic parameter estimation. A formulation is given which is similar to the recursive least squares (RLS) algorithm but uses noise covariances instead of a forgetting factor for tuning. Optimization for low computing power hardware is discussed. A demonstrator aircraft based on an R/C model is equipped with the required hardware (air and inertial data sensors, onboard processor, telemetry). Wind tunnel tests and calculations produce a reference data set for aerodynamics and propulsion of this aircraft, which is then used for simulation. This simulation allows to prove the performance of the parameter identification (PID) algorithm and predict the set of parameters which is identifiable with the given hardware. The main influence on identifiability is the relative contribution of a derivative to a coefficient in relation to the output noise level. Correlation issues are identified which arise because of the very fast rolling motion (and somewhat less the pitching motion) in comparison to the achievable update rates. Two sorts of flight test results are presented: post-flight analyses of logged flight data and identified parameters from the working real-time algorithm. Although some minor derivatives are not identifiable, the results prove the general feasibility of the approach.