Modeling And Identification Of A Micro Co-Axial Helicopter Based On Position And Body Angles Measurements

Author: Ory Schnitzer
Supervisor: Profs. Moshe Idan & Tal Shima

Experimental project

The Cooperative Autonomous Systems Laboratory (CASY) was recently established at the Technion. The laboratory is designed to operate as a flexible test bed for demonstrating and validating results related to cooperating multi-UAV. A motion capture system estimates the position and orientation (6DOF) of each vehicle at a rate of 100HZ. These estimates are processed on a PC, which calculates control commands based on this feedback.

The computed commands are then converted to an analog signal which
is transmitted through modified hobby radio transmitters. The primary aerial vehicle currently in use is the CX-3 co-axial micro electrical helicopter, manufactured by e-flight. Simple control algorithms, which were tuned heuristically are able to stabilize the helicopter in a hover and also allow slow position transitions, even though no velocity or acceleration estimates are available. An accurate dynamic model of the helicopter could enable better controller design and model based state estimation. This report summarizes
initial modeling, flight data recording and identification attempts.

A detailed parametric dynamic model is derived from first principles, including specific modeling of all major components of the vehicle. Even after linearizing the model about hovering conditions it remains highly coupled. Furthermore, it is seen that the stabilizing bar (different in many ways than the stabilizing bar of standard helicopters)
actuating the upper rotor acts as mechanical feedback which is likely to affect the observability of the coupled upper & lower rotor, stabilizing bar and fuselage dynamics.

Reviewing the available publications on identification attempts of aerial vehicles (large and small) quickly reviled that attempting to identify an aerial vehicle based on position and body angles measurements is unusual, and explicitly does not pertain to the minimal standards required for accurate identification stated by leading researchers. For these reasons, the parameters of the proposed dynamic model were not attempted to be identified. Instead,
identification of the low order dynamics was set as an initial goal.

Five experimental flights were designed
and performed. The different experiments differ in length and the choice of excitation inputs, amount of feedback and frequency range of the excitation. SISO frequency responses and the coherence functions were calculated. In order to identify a MIMO state space model, the MIMO frequency response should be calculated.

Although calculating the MIMO response after calculating the SISO responses is simple, this was not yet done (appropriately) but the relevant theory of MIMO frequency responses calculation
and important practical considerations are reviewed from the literature. In this project, only low order SISO transfer function (process models) were identified. Both frequency domain and time domain methods were tried. The attempts in the frequency domain were not successful. Time domain identification results show good prediction. A simple controller is presented based on an identified transfer function. Conclusions and propositions for future work are offered as well.