Verifiable, Control-Oriented Learning On The Fly
A collaborative project supported by the Department of Defense and coordinated by Air Force Office of Scientific Research.
- University of Texas at Austin: Arie Israel (Math), Ufuk Topcu (Aero) and Rachel Ward (Math)
- Princeton University: Amir Ali Ahmadi (OR), Charles Fefferman (Math) and Clancey Rowley (MAE)
- Northeastern University: Mario Sznaier (ECE)
Objective: The objective of this project is to develop a theoretical and algorithmic foundation for run-time learning and control of physical, autonomous systems. The resulting algorithms will
- adapt to unforeseen, possibly abrupt changes in the system and its environment;
- establish verifiable guarantees---in a sense to be precisely defined---with respect to high-level safety and performance specifications; and
- obey and leverage the laws of physics and contextual knowledge.
Approach: At a high level, the proposed approach treats the (limited) data that the system generates and the existing side information as the first-class objects of control-oriented learning. Such side information includes the physical laws that the system obeys, the context in which it serves, and the structure in its mathematical representation.
The approach also embraces the fact that developing truly autonomous systems—and, in particular, control-oriented learning on the fly—is beyond the reach of any single discipline. It distills ideas from (and in many cases discovers novel extensions of) a number of conventionally disparate disciplines, and composes them in a unified foundation.
Research thrusts: The proposed research plan relies on a composition of learning, verification and synthesis (its three main thrusts respectively). The outcomes of these thrusts complement each other toward developing verifiable functionality for control-oriented learning on the fly. The back-end Computation Engine will provide cross-cutting, efficient optimization algorithms addressing the needs of all thrusts. All three thrusts build on a common working principle supported by the front-end Constraint Engine, which systematically integrates data and side information for the utilization of our control-oriented learning algorithms.