Solving Sum-of-Costs Multi-Agent Pathfinding with Answer-Set Programming
Authors: Rodrigo N. Gómez, Carlos Hernández, Jorge A. Baier9867-9874
AAAI 2020 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | In our experimental evaluation, we show that our approach outperforms searchand SAT-based sum-of-costs MAPF solvers when grids are congested with agents. |
| Researcher Affiliation | Academia | 1Departamento de Ciencia de la Computaci on, Pontificia Universidad Cat olica de Chile, Chile 2Departamento de Ciencias de la Ingenier ıa, Universidad Andr es Bello, Chile 3Instituto Milenio Fundamentos de los Datos, Chile |
| Pseudocode | Yes | Indeed, most of the code needed to solve a MAPF instance is included in this paper. Examples of this code are the rules (1), (2), (3), (4), (6), (7), (8), (9), (10), (11), (12), (13), (14) provided throughout the paper. |
| Open Source Code | No | The paper states, 'The code used for our implementation was written in Python 3.7 using Clingo 5.3 (Gebser et al. 2014) for the ASP solver.' However, it does not provide any link or explicit statement about making their specific implementation open-source. |
| Open Datasets | Yes | First, we experimented on 8 8 and 20 20 randomly generated problems with 10% obstacles. We experimented with a warehouse grid used in the MAPF literature (e.g., Felner et al. 2018), shown in Figure 5. |
| Dataset Splits | No | The paper describes using randomly generated problems and a warehouse grid for evaluation, but does not specify training, validation, or test dataset splits as it applies to standard machine learning tasks. |
| Hardware Specification | Yes | All experiments were run on a 3.40GHz Intel Core i53570K with 8GB of memory running Linux. |
| Software Dependencies | Yes | The code used for our implementation was written in Python 3.7 using Clingo 5.3 (Gebser et al. 2014) for the ASP solver. |
| Experiment Setup | Yes | Clingo was run with 4 threads in parallel-mode, and using USC as the optimization strategy, unless otherwise stated. All experiments were run on a 3.40GHz Intel Core i53570K with 8GB of memory running Linux. We set a runtime limit of 5 minutes for all problems. |