When both Grounding and not Grounding are Bad -- A Partially Grounded Encoding of Planning into SAT (Extended Version)
arXiv:2603.19429v1 Announce Type: new Abstract: Classical planning problems are typically defined using lifted first-order representations, which offer compactness and generality. While most planners ground these representations to simplify reasoning, this can cause an exponential blowup in size. Recent approaches instead operate directly on the lifted level to avoid full grounding. We explore a middle ground between fully lifted and fully grounded planning by introducing three SAT encodings that keep actions lifted while partially grounding predicates. Unlike previous SAT encodings, which scale quadratically with plan length, our approach scales linearly, enabling better performance on longer plans. Empirically, our best encoding outperforms the state of the art in length-optimal planning on hard-to-ground domains.
arXiv:2603.19429v1 Announce Type: new Abstract: Classical planning problems are typically defined using lifted first-order representations, which offer compactness and generality. While most planners ground these representations to simplify reasoning, this can cause an exponential blowup in size. Recent approaches instead operate directly on the lifted level to avoid full grounding. We explore a middle ground between fully lifted and fully grounded planning by introducing three SAT encodings that keep actions lifted while partially grounding predicates. Unlike previous SAT encodings, which scale quadratically with plan length, our approach scales linearly, enabling better performance on longer plans. Empirically, our best encoding outperforms the state of the art in length-optimal planning on hard-to-ground domains.
Executive Summary
This article explores a middle ground approach between fully lifted and fully grounded planning in classical planning problems. The authors introduce three SAT encodings that keep actions lifted while partially grounding predicates, achieving linear scalability and better performance on longer plans. Empirical results demonstrate the superiority of this approach in length-optimal planning on hard-to-ground domains. The proposed method offers a promising solution to the trade-off between compactness and reasoning efficiency in planning.
Key Points
- ▸ The article proposes a partially grounded encoding of planning into SAT to balance compactness and reasoning efficiency.
- ▸ The approach scales linearly with plan length, unlike previous SAT encodings that scale quadratically.
- ▸ Empirical results show that the proposed method outperforms the state of the art in length-optimal planning on hard-to-ground domains.
Merits
Strength in Addressing Scalability
The partially grounded encoding scales linearly with plan length, making it a significant improvement over previous SAT encodings that scale quadratically.
Improved Performance on Hard-to-Ground Domains
The proposed method outperforms the state of the art in length-optimal planning on hard-to-ground domains, demonstrating its effectiveness in practical applications.
Demerits
Limited Generalizability to Other Planning Problems
The article focuses on classical planning problems and may not be directly applicable to other types of planning problems or domains.
Expert Commentary
The article presents a novel approach to planning that balances compactness and reasoning efficiency. The partially grounded encoding is a promising solution to the trade-off between these two desirable properties. However, its generalizability to other planning problems and domains requires further investigation. The empirical results demonstrate the superiority of the proposed method in length-optimal planning, but its applicability to other planning scenarios is unclear. Nevertheless, the article's findings contribute to the ongoing discussion on lifted planning and abstraction, and its implications for planning systems and policy decisions are significant.
Recommendations
- ✓ Future research should explore the generalizability of the proposed method to other planning problems and domains.
- ✓ The authors should investigate the applicability of the partially grounded encoding to other planning scenarios, such as planning under uncertainty or planning with incomplete information.
Sources
Original: arXiv - cs.AI
