LAO∗: A heuristic search algorithm that finds solutions with loops

Hansen, Eric A.; Zilberstein, Shlomo

doi:10.1016/s0004-3702(01)00106-0

Cited by 248 publications

(220 citation statements)

References 9 publications

Supporting

Mentioning

210

Contrasting

Unclassified

Order By: Relevance

“…This is because V L and V U are initialized with functions trivially satisfying these properties, and these properties are invariant over Bellman updates on non-absorbing states (given monotonicity, V L can only grow, while V U can only decrease). Thanks to monotonicity, with the same arguments as given for LAO * (Hansen & Zilberstein, 2001), we get that V U converges to V * in finite time on the π U -greedy graph.…”

Section: Ao *mentioning

confidence: 73%

“…We design variants of AO * and LRTDP, as well as a family of depth-first oriented heuristic searches, systematizing algorithm parameters underlying improved LAO * (here: ILAO * ) (Hansen & Zilberstein, 2001), heuristic dynamic programming (Bonet & Geffner, 2003a), and learning depth-first search (Bonet & Geffner, 2006). We furthermore design a variant of FRET better suited to problems with uninformative initial upper bounds.…”

Section: Search Algorithmsmentioning

confidence: 99%

“…For AO * , we restrict ourselves to the acyclic case, where the overhead for repeated value iteration fixed points, inherent in LAO * (Hansen & Zilberstein, 2001), disappears. (The ILAO * variant, where that issue has been addressed through a depth-first orientation, is covered as part of our depth-first oriented heuristic search family introduced in Section 3.4 below.)…”

Section: Ao *mentioning

confidence: 99%

“…We refer to such algorithms as Depth-First Heuristic Search (DFHS). Known instances are ILAO * (Hansen & Zilberstein, 2001), 8 heuristic dynamic programming (HDP) (Bonet & Geffner, 2003a), and learning depth-first search (LDFS) (Bonet & Geffner, 2006). Their commonality lies in conducting depth-first searches (DFS) on the state-space subgraph defined by actions greedy on a current upper bound V U , which is being updated backwards in DFS, until a termination criterion applies.…”

Section: Depth-first Heuristic Searchmentioning

confidence: 99%

“…MDP heuristic search (Barto, Bradtke, & Singh, 1995;Hansen & Zilberstein, 2001;Bonet & Geffner, 2003b;McMahan, Likhachev, & Gordon, 2005;Smith & Simmons, 2006;Bonet & Geffner, 2006) has the potential to find optimal policies without building the entire state space, but Kolobov et al (2011) are the only authors addressing optimal MaxProb through heuristic search. Part of the reason for this lack of research on heuristic search for MaxProb are the following two major obstacles.…”

Section: Introductionmentioning

confidence: 99%

See 4 more Smart Citations

Goal Probability Analysis in Probabilistic Planning: Exploring and Enhancing the State of the Art

Steinmetz¹,

Hoffmann²,

Buffet

2016

jair

View full text Add to dashboard Cite

Unavoidable dead-ends are common in many probabilistic planning problems, e.g. when actions may fail or when operating under resource constraints. An important objective in such settings is MaxProb, determining the maximal probability with which the goal can be reached, and a policy achieving that probability. Yet algorithms for MaxProb probabilistic planning are severely underexplored, to the extent that there is scant evidence of what the empirical state of the art actually is. We close this gap with a comprehensive empirical analysis. We design and explore a large space of heuristic search algorithms, systematizing known algorithms and contributing several new algorithm variants. We consider MaxProb, as well as weaker objectives that we baptize AtLeastProb (requiring to achieve a given goal probabilty threshold) and ApproxProb (requiring to compute the maximum goal probability up to a given accuracy). We explore both the general case where there may be 0-reward cycles, and the practically relevant special case of acyclic planning, such as planning with a limited action-cost budget. We design suitable termination criteria, search algorithm variants, dead-end pruning methods using classical planning heuristics, and node selection strategies. We design a benchmark suite comprising more than 1000 instances adapted from the IPPC, resource-constrained planning, and simulated penetration testing. Our evaluation clarifies the state of the art, characterizes the behavior of a wide range of heuristic search algorithms, and demonstrates significant benefits of our new algorithm variants.

show abstract

Section: Ao *mentioning

confidence: 73%

Section: Search Algorithmsmentioning

confidence: 99%

Section: Ao *mentioning

confidence: 99%

Section: Depth-first Heuristic Searchmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations