Implementation and Validation of Gaussian Process Model Reference Adaptive Control for Fixed Wing Unmanned Aerial Systems

Abstract

Over the last couple of decades, rapid development of unmanned aerial systems (UAS) has been observed. UAS are becoming an integral part of various industries such as agriculture, communications, defense, first response and geophysical surveys. This wide range of applications over different industries demand a number of mission specific vehicle platforms. The platforms must be reliable in all environments as well as in the presence of various uncertainties. Presently, the UAS that are flown autonomously rely on extensive manual tuning of control parameters. The control parameters are platform specific, hence transferring the controllers from one platform to another, is time consuming and requires extensive testing. A detailed approach to the implementation of an adaptive, platform independent controller, which leverages Bayesian Non-parametric (GP-MRAC) approach towards the adaptive control is presented. A Hardware-in-the-loop (HITL) framework is integrated to test the developed autopilot system. Extensive testing in the HITL environment was done and results from HITL tests and real world flight test results are presented. The results indicates that GP-MRAC outperforms the baseline PID controller and the RBF-NN MRAC from the HITL.