Symbolic Numeric Planning with Patterns

Authors: Matteo Cardellini, Enrico Giunchiglia, Marco Maratea

AAAI 2024 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental For the experimental analysis we considered all the domains and problems of the 2023 Numeric International Planning Competition (IPC) (Arxer and Scala 2023). We compared our planner PATTY with the three symbolic planners SPRINGROLL... The results show that, compared to the other symbolic planners, our planner PATTY has always better performance on every domain, while compared to all the other planners, PATTY has overall remarkably good performances, being the fastest system able to solve most problems on the largest number of domains. For lack of space, Table 1 presents the results...
Researcher Affiliation Academia Matteo Cardellini*1, 2, Enrico Giunchiglia*2, Marco Maratea3 1 DAUIN, Politecnico di Torino, Italy 2 DIBRIS, Universit a di Genova, Italy 3 De Ma CS, Universit a della Calabria, Italy
Pseudocode No The paper describes methods using prose and mathematical formulas, but no explicit pseudocode or algorithm blocks are provided.
Open Source Code Yes Our planner is available at https://pattyplan.com
Open Datasets Yes For the experimental analysis we considered all the domains and problems of the 2023 Numeric International Planning Competition (IPC) (Arxer and Scala 2023).
Dataset Splits No The paper mentions using domains and problems from the 2023 Numeric International Planning Competition (IPC) but does not specify any training, validation, or test data splits.
Hardware Specification Yes Analyses have been run on an Intel Xeon Platinum 8000 3.1GHz with 8 GB of RAM.
Software Dependencies Yes All the symbolic planners have been run using Z3 v4.12.2 (De Moura and Bjørner 2008) for checking the satisfiability of the formula (3), represented as a set of assertions in the SMT-LIB format (Barrett, Fontaine, and Tinelli 2016).
Experiment Setup Yes We then considered the same settings used in the Agile Track of the IPC, and thus with a time limit of 5 minutes. For the symbolic planners, the bound is increased starting from n = 1 until a plan is found or resources run out.