Modeling of unmanned small scale rotorcraft based on Neural Network identification

Design and development of unmanned aerial vehicles has attracted increased interest in the recent past. Rotorcraft UAVs, in particular have more challenges than its fixed wing counterparts. More research and experiments have been conducted to study the stability and control of RUAVs. A model-based control system design is particularly of our interest since it avoids a tedious trial and error process. To be able to successfully stabilize and control the RUAVs therefore a sufficiently accurate model is necessary. There are many methods in modeling small-scale rotorcraft. Using a standard first-principle based modeling approach, considerable knowledge about rotorcraft flight dynamics is required to derive the governing equation. Another method is system identification from flight data. This paper presents a method for system identification using Neural Networks. Input-output data are provided from nonlinear simulation of X-Cell 60 small scale helicopter. The data is used to train the multi-layer perceptron combined with NNARXM time regression input vector to learn nonlinear behavior of the vehicle.