Online parameter identification study on a small fixed-wing UAV

The time-domain modified recursive least squares (MRLS) online parameter identification algorithms' accuracy and robustness have been verified through a small fixed-wing UAV flight test. The frequency domain least-squares (FLS) online parameter identification algorithm has also been tested, and is compared with the time-domain method. The flight test results show that with sufficient active excitation, these two types of online parameter identification algorithms are able to perform the system parameters identification accurately, even using low-cost sensors (MEMS Inertial Measurement Unit, and noisy wind vanes). A single iteration of the running time of the time-domain modified recursive least-squares (MRLS) on a low-cost arm chip takes less than 0.5ms, meeting the real-time requirements. The frequency-domain least-squares (FLS) algorithm takes slightly larger amount of time, but the convergence course is similar to time-domain algorithm. The verification of the online parameter identification algorithm via flight test provides technical foundations of the fault tolerance control system.