Event
M.S. Thesis Defense: Saleh Kemal
Tuesday, December 1, 2026
1:30 p.m.
AVW 1146
Announcement: M.S. Thesis Defense
Name: Saleh Kemal
Committee:
Professor John Baras (Chair)
Professor P.S. Krishnaprasad
Professor Calin Belta
Date/time: Tuesday, December 1, 2026 at 1:30 PM
Committee:
Professor John Baras (Chair)
Professor P.S. Krishnaprasad
Professor Calin Belta
Date/time: Tuesday, December 1, 2026 at 1:30 PM
Location: AVW 1146
Title: Belief-Informed Hybrid Control With Target-Set Convergence
Abstract: Controlling a hybrid dynamical system toward a target set under parameter uncertainty raises a dual challenge. Each control action affects both the system state and what the controller learns about the parameter. An informative action may make little immediate progress or even move the system away from the target, while an action that appears favorable under the current belief may perform poorly under the true parameter.
We propose a belief-informed dual-control algorithm that selects feedback actions from a catalog subject to a conditional expected-decrease constraint on a target-set progress function. A slack budget bounds cumulative deviations from the required decrease. With a two-step lookahead, the algorithm uses predicted observations to update the parameter belief before evaluating which subsequent actions would satisfy the constraint. Under correct conditional prediction, distance-comparison bounds, and recursive feasibility, we prove almost-sure convergence to the target set at decision times. We also bound the expected cumulative progress-function value and the expected time to enter a neighborhood of the target set.
We validate the algorithm numerically on a planar regulation problem with an unknown control direction and on two robotic tasks: bimanual assembly under friction uncertainty and bimanual handoff under grasp uncertainty.
