Anytime Multi-Agent Path Finding via Machine Learning-Guided Large Neighborhood Search
Authors: Taoan Huang, Jiaoyang Li, Sven Koenig, Bistra Dilkina9368-9376
AAAI 2022 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We show experimentally that our solver, MAPF-ML-LNS, significantly outperforms MAPF-LNS on the standard MAPF benchmark set in terms of both the speed of improving the solution and the final solution quality. |
| Researcher Affiliation | Academia | University of Southern California {taoanhua, jiaoyanl, skoenig, dilkina}@usc.edu |
| Pseudocode | Yes | Algorithm 1: MAPF-LNS |
| Open Source Code | No | No concrete access information (e.g., a specific repository link or an explicit statement of code release for the paper's methodology) is provided. The paper only mentions using an existing 'open-source solver (Joachims 2006)' for a component. |
| Open Datasets | Yes | We compare against MAPF-LNS on five grid maps of different sizes and structures from the MAPF benchmark set (Stern et al. 2019) |
| Dataset Splits | No | The paper describes using '16 instances from 16 random scenarios' for training and '4 MAPF instances from another 4 random scenarios' for validation, but does not provide explicit dataset split percentages or absolute sample counts for train/test/validation on a single defined dataset. |
| Hardware Specification | Yes | We implement MAPF-ML-LNS in C++ and conduct our experiments on a 2.4 GHz Intel Core i7 CPU with 16 GB RAM. |
| Software Dependencies | No | The paper mentions using an 'open-source solver (Joachims 2006)' for learning the ranking function, but does not specify its version number or any other software dependencies with version numbers (e.g., programming language, libraries, or frameworks). |
| Experiment Setup | Yes | We use regularization parameter C = 0.1 and the default values for the other parameters in the SVMrank solver. During training, we run Algorithm 2 for T = 100 iterations. We sample S = 20 agent sets in each iteration of MAPF-ML-LNS. The runtime limit of PP per replan is set to 2 seconds for the warehouse map and 0.6 seconds for the other grid maps initially and then adaptively set to twice the average runtime of all successful replans so far after the first 30 successful replans. |