The FF Heuristic for Lifted Classical Planning
Authors: Augusto B. Corrêa, Florian Pommerening, Malte Helmert, Guillem Francès9716-9723
AAAI 2022 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | In our experiments, we show that a planner using the lifted FF implementation produces state-of-the-art results for lifted planners. It also reduces the gap to state-of-the-art ground planners in domains where grounding is feasible. |
| Researcher Affiliation | Academia | 1University of Basel, Switzerland 2Universitat Pompeu Fabra, Spain |
| Pseudocode | Yes | Algorithm 1: Executing an annotated Datalog program. |
| Open Source Code | Yes | The source code of our implementation is available online (Corrˆea et al. 2022). Corrˆea, A. B.; Pommerening, F.; Helmert, M.; and Franc es, G. 2022. Code from the AAAI 2022 paper The FF Heuristic for Lifted Classical Planning . https://doi.org/10.5281/ zenodo.6373793. Accessed: 2022-04-11. |
| Open Datasets | Yes | We benchmark our algorithms on two sets. The first set contains 1001 IPC tasks from 29 STRIPS domains. This is the same set used by Corrˆea et al. (2021). The second set contains 862 hard-to-ground (HTG) tasks over 11 different domains. The HTG set is the same used by Lauer et al. (2021). |
| Dataset Splits | No | The paper evaluates on benchmark sets but does not explicitly mention training, validation, and test splits with specific percentages or counts. It does not refer to a validation dataset. |
| Hardware Specification | Yes | Our experiments were run on an Intel Xeon Silver 4114 processor running at 2.2 GHz using a run time limit of 30 minutes and a memory limit of 16 Gi B per task. |
| Software Dependencies | Yes | We implemented the Datalog-based FF heuristic using the Powerlifted (PWL) planner by Corrˆea et al. (2020). Fast Downward (FD), release 20.06 (Helmert 2006). |
| Experiment Setup | Yes | We always consider action schemas as unit-cost. We tested hadd and h FF (both lifted and ground versions) using a simple Eager Greedy Best-First Search (Pohl 1970), and also using the Lazy GBFS with preferred operators and a boosted dual-queue (Richter and Helmert 2009; Corrˆea et al. 2021). |