Output Feedback Self-Tuning PID Using Wavelet Neural Network Identification for Real-Time UASs Trajectory Tracking

We propose an output feedback selftuning PID control approach that leverages Wavelet Neural Network (WNN) for online model identification and real-time trajectory tracking of Unmanned Aircraft Systems (UASs). In contrast to conventional neural network (NN)-based nonlinear dynamic modeling, which approximates nonlinearities using a superposition of sigmoidal functions, our method employs wavelet functions as universal approximators. The input-output dynamics identification is performed using discrete adaptive wavelets as activation functions. The WNN scheme preserves spatiotemporal information, a feature unavailable when adopting classic NN architectures. The contribution of the proposed control strategy has two major components: (i) the output identification of a fast and agile UAS, and (ii) the exploitation of such knowledge to implement the self-tuning of the model-free PID control gains. Experimental results on real-time trajectory tracking are presented to demonstrate the effectiveness of the proposed approach.