Generalized Potential Heuristics for Classical Planning

Francès, Guillem; Corrêa, Augusto B.; Geissmann, Cedric; Pommerening, Florian

doi:10.24963/ijcai.2019/771

Cited by 17 publications

(30 citation statements)

References 9 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Where they report a number of clauses for Q clear , Q on , Q grip and Q rew of, respectively, 767K, 3.3M, 358K and 1.2M, the number of clauses in our encoding is 242.3K, 281.5K, 100.8K and 98.9K, that is up to one order of magnitude smaller, which allows D2L to scale up to several other domains. Our approach is also more efficient than the one in (Francès et al 2019), which requires several hours to solve a domain such as Gripper, which D2L can tackle in a few seconds.…”

Section: Resultsmentioning

confidence: 99%

“…Deep reinforcement learning methods (Mnih et al 2015) have also been used to generate general policies from images without assuming prior symbolic knowledge (Groshev et al 2018;Chevalier-Boisvert et al 2019), in certain cases accounting for objects and relations through the use of suitable architectures (Garnelo and Shanahan 2019). Our work is closest to the works of Bonet, Francès, and Geffner (2019) and Francès et al (2019). The first provides a modelbased approach to generalized planning where an abstract QNP model is learned from the domain representation and sample instances and plans, which is then solved by a QNP planner (Bonet and Geffner 2020).…”

Section: Related Workmentioning

confidence: 91%

See 1 more Smart Citation

Learning General Policies from Small Examples Without Supervision

Francès,

Bonet,

Geffner

2021

Preprint

Self Cite

View full text Add to dashboard Cite

Generalized planning is concerned with the computation of general policies that solve multiple instances of a planning domain all at once. It has been recently shown that these policies can be computed in two steps: first, a suitable abstraction in the form of a qualitative numerical planning problem (QNP) is learned from sample plans, then the general policies are obtained from the learned QNP using a planner. In this work, we introduce an alternative approach for computing more expressive general policies which does not require sample plans or a QNP planner. The new formulation is very simple and can be cast in terms that are more standard in machine learning: a large but finite pool of features is defined from the predicates in the planning examples using a general grammar, and a small subset of features is sought for separating "good" from "bad" state transitions, and goals from non-goals. The problems of finding such a "separating surface" while labeling the transitions as "good" or "bad" are jointly addressed as a single combinatorial optimization problem expressed as a Weighted Max-SAT problem. The advantage of looking for the simplest policy in the given feature space that solves the given examples, possibly non-optimally, is that many domains have no general, compact policies that are optimal. The approach yields general policies for a number of benchmark domains.

show abstract

Section: Resultsmentioning

confidence: 99%

Section: Related Workmentioning

confidence: 91%

Learning General Policies from Small Examples Without Supervision

Francès,

Bonet,

Geffner

2021

Preprint

Self Cite

View full text Add to dashboard Cite

show abstract

“…The sketch language is the language of general policies (Bonet and Geffner 2018) that has been used for learning as well (Martín and Geffner 2004;Bonet, Francès, and Geffner 2019;Francès, Bonet, and Geffner 2021). The description logic features have also been used to learn linear value functions that can be used to solve problems greedily (Francès et al 2019;de Graaff, Corrêa, and Pommerening 2021) and dead-end classifiers (Ståhlberg, Francès, and Seipp 2021). The use of numerical features that can be incremented and decremented qualitatively is inspired by QNPs (Srivastava et al 2011;Bonet and Geffner 2020).…”

Section: Related Workmentioning

confidence: 99%

Learning Sketches for Decomposing Planning Problems into Subproblems of Bounded Width: Extended Version

Drexler¹,

Seipp²,

Geffner³

2022

Preprint

View full text Add to dashboard Cite

Recently, sketches have been introduced as a general language for representing the subgoal structure of instances drawn from the same domain. Sketches are collections of rules of the form C → E over a given set of features where C expresses Boolean conditions and E expresses qualitative changes. Each sketch rule defines a subproblem: going from a state that satisfies C to a state that achieves the change expressed by E or a goal state. Sketches can encode simple goal serializations, general policies, or decompositions of bounded width that can be solved greedily, in polynomial time, by the SIWR variant of the SIW algorithm. Previous work has shown the computational value of sketches over benchmark domains that, while tractable, are challenging for domain-independent planners. In this work, we address the problem of learning sketches automatically given a planning domain, some instances of the target class of problems, and the desired bound on the sketch width. We present a logical formulation of the problem, an implementation using the ASP solver Clingo, and experimental results. The sketch learner and the SIWR planner yield a domain-independent planner that learns and exploits domain structure in a crisp and explicit form.

show abstract

“…Rather than using ML to learn an optimal plan directly, it can be used to learn general heuristics that simplify the problem and thus help to solve the problem more efficiently. By automatically labeling states on whether they are alive or unsolvable, a mixed integer linear program can be developed [FCGP19]. As the method automatically labels the states by brute force, it only works on problems with a small number of reachable states.…”

Section: Unsupervised Learningmentioning

confidence: 99%

A Research Agenda for AI Planning in the Field of Flexible Production Systems

Köcher,

Heesch,

Widulle

et al. 2021

Preprint

View full text Add to dashboard Cite

Production companies face problems when it comes to quickly adapting their production control to fluctuating demands or changing requirements. Control approaches aiming to encapsulate production functions in the sense of services have shown to be promising in order to increase flexibility of Cyber-Physical Production Systems. But an existing challenge of such approaches is finding production plans based on provided functionalities for a set of requirements, especially when there is no direct (i.e., syntactic) match between demanded and provided functions. In such cases it can become complicated to find those provided functions that can be arranged into a plan satisfying the demand. While there is a variety of different approaches to production planning, flexible production poses specific requirements that are not covered by existing research. In this contribution, we first capture these requirements for flexible production environments. Afterwards, an overview of current Artificial Intelligence approaches that can be utilized in order to overcome the aforementioned challenges is given. Approaches from both symbolic AI planning as well as approaches based on Machine Learning are discussed and eventually compared against the requirements. Based on this comparison, a research agenda is derived.

show abstract

Generalized Potential Heuristics for Classical Planning

Cited by 17 publications

References 9 publications

Learning General Policies from Small Examples Without Supervision

Learning General Policies from Small Examples Without Supervision

Learning Sketches for Decomposing Planning Problems into Subproblems of Bounded Width: Extended Version

A Research Agenda for AI Planning in the Field of Flexible Production Systems

Contact Info

Product

Resources

About