Counterexample-Guided Cartesian Abstraction Refinement for Classical Planning

  title={Counterexample-Guided Cartesian Abstraction Refinement for Classical Planning},
  author={Jendrik Seipp and Malte Helmert},
  journal={J. Artif. Intell. Res.},
Counterexample-guided abstraction refinement (CEGAR) is a method for incrementally computing abstractions of transition systems. We propose a CEGAR algorithm for computing abstraction heuristics for optimal classical planning. Starting from a coarse abstraction of the planning task, we iteratively compute an optimal abstract solution, check if and why it fails for the concrete planning task and refine the abstraction so that the same failure cannot occur in future iterations. A key… 

New Refinement Strategies for Cartesian Abstractions

It is demonstrated that it is possible to execute all optimal abstract plans in the concrete task simultaneously, and methods to minimize the time and number of refinement steps until the authors find a concrete solution are presented.

Incremental Search for Counterexample-Guided Cartesian Abstraction Refinement

This work casts the refinements of Cartesian abstraction refinement as an incremental search problem and shows that this drastically reduces the time for computing abstractions.

Counterexample-Guided Abstraction Refinement for Pattern Selection in Optimal Classical Planning

Using multiple CEGAR algorithm runs, this work can compute much larger pattern collections, still in shorter time than existing approaches, which leads to a planner that outperforms the state-of-the-art pattern selection methods by a significant margin.

Refining Abstraction Heuristics during Real-Time Planning

Results on various benchmarks, including the sliding tile puzzle and several IPC domains, indicate that the use of abstraction-based heuristics can improve performance compared to traditional heuristic updating.

Interleaving Search and Heuristic Improvement

This work interleave the computation of abstraction heuristics with search, avoiding a long precomputation phase and allowing information from the search to be used for guiding the abstraction selection.

Online Relaxation Refinement for Satisficing Planning: On Partial Delete Relaxation, Complete Hill-Climbing, and Novelty Pruning

This work introduces several online-refinement search algorithms, based on hill-climbing and greedy best-first search, and evaluates them with the partial delete relaxation heuristic hCFF, which can be refined by treating additional conjunctions of facts as atomic, and whose refinement operation satisfies the convergence property required for completeness.

Online Saturated Cost Partitioning for Classical Planning

By incrementally extending the set of cost-partitioned heuristics online during the search, this work drastically speed up the planning process and even often solve more tasks.

Endomorphisms of Classical Planning Tasks

This paper employs structure-preserving maps on labeled transition systems (LTSs), namely endomorphisms well known from model theory, in order to detect redundancy, and forms the computation of endomorphicisms as a constraint satisfaction problem (CSP) that can be solved by an off-the-shelf CSP solver.

Bridging the Gap between Abstractions and Critical-Path Heuristics via Hypergraphs

This paper devise a method to construct hyperabstractions, using either a set of abstractions or a critical-path heuristic as a seed, in a way that guarantees that the resulting distance estimations dominate those of the input heuristics, sometimes even strictly.



Counterexample-guided Abstraction Refinement for Classical Planning Master's Thesis Erklärung (declaration) Zusammenfassung

This work proposes a CEGAR algorithm for computing abstraction heuristics for optimal classical planning and shows how refining the abstraction online when the search errs reduces the number of expanded nodes.

Online Refinement of Cartesian Abstraction Heuristics

It is described how online refinement can effectively be used in A* and the approach is evaluated on the IPC benchmarks, where it outperforms offline-generated abstractions in many domains.

Merge-and-shrink abstractions for classical planning : theory, strategies, and implementation

This thesis provides a comprehensive description of the merge-and-shrink framework in terms of transformations of transition systems, and describes an optimized implementation of the merged heuristics framework that substantially improves the efficiency compared to previous implementations.

Flexible Abstraction Heuristics for Optimal Sequential Planning

An approach to deriving consistent heuristics for automated planning, based on explicit search in abstract state spaces, which subsumes planning with pattern databases as a special case and shows that the approach is competitive with the state of the art.

Relaxation Refinement: A New Method to Generate Heuristic Functions

This work pioneer the application of (predicate) abstraction refinement for the generation of heuristic functions that are intelligently adapted to the problem at hand and investigates how an abstraction refinement process for generating heuristic function should differ from the process used in the verification context.

Counterexample-guided Planning

This work extends the counterexample-guided abstraction refinement paradigm to probabilistic models (namely, perfect information games and, as a special case, MDPs), which allows it to apply to the AI planning problem.

Optimal admissible composition of abstraction heuristics

The Fast Downward Planning System

  • M. Helmert
  • Computer Science
    J. Artif. Intell. Res.
  • 2006
A full account of Fast Downward's approach to solving multivalued planning tasks is given and a new non-heuristic search algorithm called focused iterative-broadening search, which utilizes the information encoded in causal graphs in a novel way is presented.

Narrowing the Gap Between Saturated and Optimal Cost Partitioning for Classical Planning

It is shown that searching in the space of orders leading to significantly better heuristic estimates than with previously considered orders and using multiple orders leads to a heuristic that is significantly better informed than any single-order heuristic.

Concise finite-domain representations for PDDL planning tasks