Pattern Selection for Optimal Classical Planning with Saturated Cost Partitioning
Authors: Jendrik Seipp
IJCAI 2019 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We implemented the SYS-SCP pattern selection algorithm in the Fast Downward planning system [Helmert, 2006] and conducted experiments with the Downward Lab toolkit [Seipp et al., 2017c] on Intel Xeon Silver 4114 processors. Our benchmark set consists of all 1827 tasks without conditional effects from the optimization tracks of the 1998 2018 IPCs. The tasks belong to 48 different domains. We limit time by 30 minutes and memory by 3.5 Gi B. All benchmarks1, code2 and experimental data3 have been published online. |
| Researcher Affiliation | Academia | Jendrik Seipp University of Basel, Switzerland jendrik.seipp@unibas.ch |
| Pseudocode | Yes | Algorithm 1 shows pseudo-code for the procedure, which we call SYS-SCP. |
| Open Source Code | Yes | All benchmarks1, code2 and experimental data3 have been published online. |
| Open Datasets | Yes | Our benchmark set consists of all 1827 tasks without conditional effects from the optimization tracks of the 1998 2018 IPCs. |
| Dataset Splits | No | The paper does not provide specific train/validation/test dataset splits in terms of percentages or counts for the benchmark tasks. It mentions using '1000 sample states obtained with random walks' for heuristic diversification, but this is not a dataset split for model training. |
| Hardware Specification | Yes | We implemented the SYS-SCP pattern selection algorithm in the Fast Downward planning system [Helmert, 2006] and conducted experiments with the Downward Lab toolkit [Seipp et al., 2017c] on Intel Xeon Silver 4114 processors. |
| Software Dependencies | No | The paper mentions 'Fast Downward planning system [Helmert, 2006]' and 'Downward Lab toolkit [Seipp et al., 2017c]' but does not provide specific version numbers for these software dependencies or any other libraries. |
| Experiment Setup | Yes | We use at most 2M states per PDB and 20M states in the PDB collection for all SYS-SCP runs. We limit time by 30 minutes and memory by 3.5 Gi B. The combination that solves the highest number of tasks is 10s for the inner and 100s for the outer loop. We use these values in all other experiments. |