Planning as heuristic search

By Blai Bonet, Hector Geffner. Artificial Intelligence, 2001

1,247 citations recorded by the source corpus.

Subfields: Artificial Intelligence, Computer Networks and Communications

View Paper: https://www.semanticscholar.org/paper/2387b8880e989197e5d2e6c5642fdde4f3561600

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Computer Networks and Communications

Citations in the corpus

ARA*: Anytime A* with Provable Bounds on Sub-Optimality (2003, background): “A* search with inflated heuristics (actual heuristic values are multiplied by an inflation factor > 1) is sub-optimal but proves to be fast for many domains [1, 5, 8] and also provides a bound on the sub-optimality, namely, the by which the heuristic is inflated [7].”
Landmarks, Critical Paths and Abstractions: What's the Difference Anyway? (2009, methodology): “Instead, we focus on the max heuristic h max (Bonet and Geffner 2001), which provides an admissible estimate for h + .”
Inventing Relational State and Action Abstractions for Effective and Efficient Bilevel Planning (2022, methodology): “First, we note that it is highly uncommon to use hAdd with A ∗ in practice with hand-deﬁned PDDL representations, because hAdd is inadmissible and suffers greatly from overestimation issues [94].”
Anytime Dynamic A*: An Anytime, Replanning Algorithm (2005, background): “…the fact that in many domains inflating the heuristic values used by A* (resulting in the weighted A* search) often provides substantial speed-ups (Bonet & Geffner 2001; Korf 1993; Zhou & Hansen 2002; Edelkamp 2001; Rabin 2000; Chakrabarti, Ghosh, & DeSarkar 1988) at the cost of solution…”
PDDLStream: Integrating Symbolic Planners and Blackbox Samplers via Optimistic Adaptive Planning (2018, methodology): “This idea is similar to the concept of layer for facts and actions in a relaxed planning graph (Bonet and Geffner 2001) with the distinction that a stream instance can be evaluated many times.”
Subdimensional expansion for multirobot path planning (2015, positive): “Inflating the heuristic by multiplying the heuristic by some > 1 is known to significantly decrease the time A* requires to find a solution in many cases [60, 61, 62, 63, 64].”
Black-Box Policy Search with Probabilistic Programs (2015, methodology): “er the policy, either represented explicitly or as an implicit property of the planning algorithm. Approximation schemes include value/policy iteration, Q-learning, methods based on heuristic search [Bonet and Geffner, 2001], Monte Carlo sampling such as Monte Carlo Tree Search [Kocsis and Szepesvar”
Integrated Task and Motion Planning (2020, background): “for the action to be executed. Furthermore, the AI planning community has developed a repertoire of very eective, domain-independent search algorithms that exploit this type of action representation (16, 17, 18). Research in task and motion planning (tamp) seeks to combine AI approaches to task plan”
Planning as satisfiability: Heuristics (2012, methodology): “We consider translations of hard combinatorial search problems into planning as proposed by Porco et al. [10], as well as hard instances generated according to problem parameters that have been empirically determined to be hard for existing algorithms [11].”
Robust planning with incomplete domain models (2017, methodology): “Our procedure employs an extension of the common relaxation technique (c.f., Bonet and Geffner (2001)) by ignoring both the known and possible delete effects of actions in constructing π̃.7 Relaxed planning graph construction We construct the relaxed planning graph G = 〈L1, A1, ..., LT−1, AT−1, LT…”
The LAMA Planner: Guiding Cost-Based Anytime Planning with Landmarks (2010, methodology): “The cost estimate for an operator in the original FF heuristic is its depth in the relaxed planning graph, which in the case of planning with unit-cost operators is equivalent (Fuentetaja, Borrajo, & Linares López, 2009) to propagating costs via the hmax criterion (Bonet & Geffner, 2001).”
Deterministic planning in the fifth international planning competition: PDDL3 and experimental evaluation of the planners (2009, background): “This situation can be particularly problematic for delete-relaxed plan heuristics [11]: in the delete relaxation, the optimal (and also the easiest) solution is to store all crates in area D32, which is clearly not very indicative of the real plan.”
Concise finite-domain representations for PDDL planning tasks (2009, methodology): “• Edelkamp and Helmert’s algorithm [15] proposed for the MIPS planner [16,17], • Scholz’s algorithm for finding c-constraints [44], • Gerevini and Schubert’s DISCOPLAN [25,26], • Rintanen’s invariant synthesis algorithm [42], • Bonet and Geffner’s algorithm for generating mutexes [5], and • Fox and ”
The Metric-FF Planning System: Translating ''Ignoring Delete Lists'' to Numeric State Variables (2003, background): “FF (Homann & Nebel, 2001) is a close relative of HSP ( Bonet & Gener, 2001 ).”
A Survey on Ambient Intelligence in Health Care (2013, neutral): “, forward and backward search [101]), graph-based techniques (e.”