Sensor based Adaptive Neural Network Control for Small-scale Unmanned Helicopter

The advantage of neural network (NN) controllers lies on their capability of learning from training data to approximate highly nonlinear functions for nonlinear system control. In this paper, we present a MIMO neural network controller with two inputs and two outputs for flight control of a small-scale unmanned helicopter. The feedback is based on multiple on-board sensors including altitude meter, digital gyro, acceleration meter and GPS, etc. The neural network based controller has an adaptive NN PID configuration. According to the sensor measurement, it tunes the PID parameters k_p, k_b, and k_d in real time to control the helicopter flying in the required attitude. The control law and its training scheme are presented. The MATLAB based simulation shows that the outputs of the system can responds the input signals accurately.