Experimental inverted pendulum unfalsified control

摘要

Unfalsified control is applied to an experimental inverted pendulum in the lab. A non-linear model for the system is developed. Using a set of six candidate controllers, of which two are destabilizing, a supervisory program is built. Experimental results show the best controllers are selected most of the time. When a destabilizing controller is put in the loop it is quickly removed and the performance does not degrade significantly. The application shows unfalsified control can be useful for control adaptation to process changes, even knowing it can some times select destabilizing controllers.