Unmanned Flight Data Acquisition And Model Identification

UAV(Unmanned Aerial Vehicle) plays a important role in national defense and civil life. As a member of UAV, Unmanned Helicopter has its own application domain. And researches on it has theoretic, engineering and economical value. Development of a reliable high-performance helicopter-based UAV requires an accurate and practical model of the vehicle dynamics. This article have designed a real-time data acquisition Equipment for the UAV flight and attained a state-space model for the helicopter's longitudinal channel.The data acquisition Equipment parts on the ground and at the plane gather the data at the same time, which is labeled by the GPS. The edges of PWM signal are captured by the CAP unit; Voltage from the integral circuit is gathered by the ADC module; the flight' state is gathered by ADS8364. The serial communication interface(SCI) for single and half-duplex communication based on First In First Out is adopted, and communication mode combined interrupt and query was used in system software.After getting the whole and useful data during the flight, A Savitzky-Golay (polynomial) FIR smoothing filter is designed to find the bad points. The method of Lagrange multipliers is used to correct the bad points found. The data selected for identifying model is normalized and tested for their compatibility. Regression method is adopted to get the ratio of the two parts of the longitudinal cyclic pitch. According to dynamic analysis of helicopter, the little perturbation equations are established, and the model structure is gained. For identifying the longitudinal difference equation, the basic recursive least squares (RLS) adaptive algorithm is adopted, the independent subsystem parameter identification and the subsystem parameter recursive identification are used. The comparison of simulation result and real data are carried out to prove the correctness of the model.The experimental result shows that the equipment can meet the demand of the time limit and the data's precision. The output of the state equation identified and the corresponding real data have almost the same change. So the model is valuable in some way.