TY - GEN
T1 - RAO
T2 - 30th AAAI Conference on Artificial Intelligence, AAAI 2016
AU - Santana, Pedro
AU - Thíebaux, Sylvie
AU - Williams, Brian
N1 - DBLP License: DBLP's bibliographic metadata records provided through http://dblp.org/ are distributed under a Creative Commons CC0 1.0 Universal Public Domain Dedication. Although the bibliographic metadata records are provided consistent with CC0 1.0 Dedication, the content described by the metadata records is not. Content may be subject to copyright, rights of privacy, rights of publicity and other restrictions.
PY - 2016
Y1 - 2016
N2 - Autonomous agents operating in partially observable stochastic environments often face the problem of optimizing expected performance while bounding the risk of violating safety constraints. Such problems can be modeled as chance-constrained POMDP's (CCPOMDP's). Our first contribution is a systematic derivation of execution risk in POMDP domains, which improves upon how chance constraints are handled in the constrained POMDP literature. Second, we present RAO, a heuristic forward search algorithm producing optimal, deterministic, finite-horizon policies for CCPOMDP's. In addition to the utility heuristic, RAO leverages an admissible execution risk heuristic to quickly detect and prune overly-risky policy branches. Third, we demonstrate the usefulness of RAO in two challenging domains of practical interest: power supply restoration and autonomous science agents.
AB - Autonomous agents operating in partially observable stochastic environments often face the problem of optimizing expected performance while bounding the risk of violating safety constraints. Such problems can be modeled as chance-constrained POMDP's (CCPOMDP's). Our first contribution is a systematic derivation of execution risk in POMDP domains, which improves upon how chance constraints are handled in the constrained POMDP literature. Second, we present RAO, a heuristic forward search algorithm producing optimal, deterministic, finite-horizon policies for CCPOMDP's. In addition to the utility heuristic, RAO leverages an admissible execution risk heuristic to quickly detect and prune overly-risky policy branches. Third, we demonstrate the usefulness of RAO in two challenging domains of practical interest: power supply restoration and autonomous science agents.
UR - https://www.scopus.com/pages/publications/85007233462
U2 - 10.1609/aaai.v30i1.10423
DO - 10.1609/aaai.v30i1.10423
M3 - Conference Paper
T3 - 30th AAAI Conference on Artificial Intelligence, AAAI 2016
SP - 3308
EP - 3314
BT - 30th AAAI Conference on Artificial Intelligence, AAAI 2016
PB - AAAI Press
Y2 - 12 February 2016 through 17 February 2016
ER -