Adaptive Multi-stage Sampling for Risk Sensitive Control

Abstract: Most applications of stochastic control involve optimization over expected or mean outcomes. Unfortunately, this does not account for a decision maker’s desire to reduce variability or the occurrence of rare, disastrous events. Risk sensitive control introduces dynamic, time-consistent risk measures to create policies that minimize risk in sequential decision making under uncertainty. In this talk we discuss an adaptive sampling based algorithm for risk sensitive control, aimed at mitigating the computational cost involved. We present the results of simulations of risk sensitive benchmarks to demonstrate the feasibility of this approach. Speaker(s): Brian Rybicki, Agenda: 6:30pm-7:00pm: Light dinner and refreshments 7:00pm-8:00pm: Talk with Q&A Room: Ground Floor – Jajodia Auditorium, Bldg: Nguyen Engineering , George Mason University, 4511 Patriot Circle, Fairfax, Virginia, United States, 22030, Virtual: https://events.vtools.ieee.org/m/364800