Data-Driven Robust Control Design: Unfalsified Control

摘要

Aerospace applications require precise control despite uncertain operating conditions and unanticipated circumstances such as battle damage. These systems must be designed to perform robustly, despite uncertain design models and difficult to analyze nonlinear effects. They must be capable of learning and adapting when accumulating data indicates that previous models must be abandoned and that existing control strategies must be changed. Data-driven design methods, collectively known as un-falsified control theory, facilitate the creation of robust control systems that learn, discover and evolve in real time in order to rapidly switch controller gains to compensate for the effects of battle, equipment failures, and other changing circumstances. Applications studies will be presented that include adaptive robot arm control and missile control. 'I have devised seven separate explanations, each of which would cover the facts as far as we know them. But which of these is correct can only be determined by fresh information which we shall no doubt find waiting for us.'