Expressive Optimal Temporal Planning via Optimization Modulo Theory
Authors: Stefan Panjkovic, Andrea Micheli
AAAI 2023 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We implemented the planning algorithm via OMT in a C++ tool called ANMLOMT, which can read either ANML (Smith, Frank, and Cushing 2008) or PDDL 2.1 specifications and can use either OPTIMATHSAT (Sebastiani and Trentin 2018) or Z3 (Bjørner, Phan, and Fleckenstein 2015) as OMT solvers. ... We compare our tool in four different configurations: with or without incrementality (incremental solvers are indicated as ANMLOMT INC ) and with either OPTIMATHSAT or Z3 as OMT engines. As a competitor, we compare against the OPTIC (Benton, Coles, and Coles 2012) planner as it is the only planner publicly available that is capable of dealing with our benchmarks. We ran the experiments on an Intel Core i9-9900KS workstation with a 1800s time limit and 20GB of memory limit. ... The coverage results are reported in Table 1 while in Figure 3 we plot the performance of the solvers. |
| Researcher Affiliation | Collaboration | Stefan Panjkovic1, 2, Andrea Micheli1 1 Fondazione Bruno Kessler, Trento, Italy 2 University of Trento, Trento, Italy {spanjkovic, amicheli}@fbk.eu |
| Pseudocode | Yes | Algorithm 1 Optimal Planning via OMT |
| Open Source Code | Yes | We implemented the planning algorithm via OMT in a C++ tool called ANMLOMT... The benchmarks and the tools used for the experimental evaluation are available in (Panjkovic and Micheli 2023). ... Panjkovic, S.; and Micheli, A. 2023. Expressive Optimal Temporal Planning via Optimization Modulo Theory: Additional Material. https://es-static.fbk.eu/people/panjkovic/aaai23. Accessed: 2022-11-28. |
| Open Datasets | Yes | The benchmarks and the tools used for the experimental evaluation are available in (Panjkovic and Micheli 2023). ... Panjkovic, S.; and Micheli, A. 2023. Expressive Optimal Temporal Planning via Optimization Modulo Theory: Additional Material. https://es-static.fbk.eu/people/panjkovic/aaai23. Accessed: 2022-11-28. |
| Dataset Splits | No | The paper describes solving planning problems and evaluating solvers on benchmarks, but it does not specify explicit training, validation, or test dataset splits with percentages or sample counts for model development or evaluation. |
| Hardware Specification | Yes | We ran the experiments on an Intel Core i9-9900KS workstation with a 1800s time limit and 20GB of memory limit. |
| Software Dependencies | Yes | We implemented the planning algorithm via OMT in a C++ tool called ANMLOMT, which can read either ANML (Smith, Frank, and Cushing 2008) or PDDL 2.1 specifications and can use either OPTIMATHSAT (Sebastiani and Trentin 2018) or Z3 (Bjørner, Phan, and Fleckenstein 2015) as OMT solvers. |
| Experiment Setup | Yes | The overall formula Πopt h is made of several components which encode different aspects of the problem: φINIT (initial state); φACTIONS h (actions encoding); φTIME h (time monotonicity); φFA h (frame axiom); φNSO h (action non-selfoverlapping); φMOD h (modf variables); φDENS h (density axiom); φGOAL h (abstract goals). ... if we want to minimize the makespan, we specify th 1 as the minimization objective, while for minimizing the total action cost the objective is P a A Ph i=1 ITE(ai, c(a), 0), where c(a) is the cost of action a. ... We compare our tool in four different configurations: with or without incrementality (incremental solvers are indicated as ANMLOMT INC ) and with either OPTIMATHSAT or Z3 as OMT engines. |