Iterative parameter estimation and model prediction of a rotary unmanned aerial vehicle

This paper presents a method for identifying and predicting attitude dynamics in a rotary unmanned aerial vehicle UAV. A base model which uses a first principles model is simplified for use in real-time prediction. An integral-based parameter identification method is presented to identify the unknown intrinsic helicopter parameters outdoors with significant wind disturbance and on a test bench. The yaw axis is used as a proof-of-concept on both the tests. For the outdoor flight, the system identification is performed using open-loop commands from a test pilot. The test-bench experiments use proportional integral derivative PID control and thus provide a further validation of the methods in closed-loop as well as open-loop. The results show that all the major yaw dynamics were captured and good future state predictions of 0.1 to 0.3 seconds were obtained using slow time varying parameters and disturbance modelling.