EHS: Hybrid Estimation and Control with Bounded Probabilities

Cornell University, Ithaca NY

Investigators

Abstract

Mark E. Campbell, Cornell University, Hybird Estimation and Control with Bounded Probabilities This project is developing a novel approach for hybrid estimation for applications envisioned in hybrid systems research. Three areas of hybrid estimation research are developed, with an integrated evaluation process using known examples and real time testbeds. First, a bounded probability hybrid estimator is developed, where multiple continuous estimators, each at a different level of probability, are implemented in parallel. The system (user or hybrid controller) can switch estimators to the level of probability desired. Level sets are used to define each probability level, and relate it back to the original probability density function. Second, connections between hybrid estimation and control and planning are made, with a specific focus on switching hybrid vehicle control, including transition of bounded model uncertainties. Third, a computational infrastructure for the hybrid estimators is developed, including real time implementations. Theoretical work is systematically validated using simple, known examples first, followed by more complex examples that can be numerically evaluated, to ultimately a real time, embedded hybrid systems evaluation on a Cornell testbed. The Cornell SeaScan UAV testbed is a suite of up to four small UAV's, and a similar lab-based, hardware-in-the-loop testbed. During the summers, undergraduate students are teamed with PhD students and post-docs in an effort to transition and verify the technologies experimentally. This work is expected to advance hybrid systems theory by enabling a general class of hybrid theory (such as control and planning) to be implemented in a realistic setting. The hybrid estimator allows a level of probability (such as risk) to be "dialed up." While vehicle control and planning are used as motivation, the hybrid estimator is developed to allow general implementation for hybrid systems, from biological to air traffic control systems.