University of Southern California Department of Electrical Engineering

Publications, Reports and Other Links
Michael G. Safonov

Papers and Reports:

Unfalsified Control Theory:   Quantitative theory of learning and adaptation

·         Let the data speak...

·         Use evolving real-time data to unfalsify (validate) controllers against hard performance criteria:

o    Choose criteria expressible directly in terms of observed data (sensor outputs, actuator inputs).

o    Avoid criteria based on “noise models” or other prior beliefs.

·         Whenever the currently active controller is falsified by evolving I/O data, it is automatically replaced by an unfalsified controller.

·         Selected papers & software demos on Unfalsified Control.

SLIDE SHOW on Data-Driven Unfalsified Adaptive Control


Return to Main Page