Multi-Agent Path Finding for Large Agents

Authors: Jiaoyang Li,Pavel Surynek,Ariel Felner,Hang Ma,T. K. Satish Kumar,Sven Koenig7627-7634

AAAI 2019 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Experimental results show that all MC-CBS variants outperform CBS by up to three orders of magnitude in terms of runtime.
Researcher Affiliation Academia Jiaoyang Li CS Department Univ. of Southern California jiaoyanl@usc.edu; Pavel Surynek Faculty of Information Technology Czech Technical University pavel.surynek@fit.cvut.cz; Ariel Felner SISE Department Ben-Gurion University felner@bgu.ac.il; Hang Ma T. K. Satish Kumar Sven Koenig CS Department Univ. of Southern California
Pseudocode Yes Algorithm 1: A template for Max Weight-d.
Open Source Code No The paper does not provide any concrete access information (e.g., repository link, explicit statement of code release) for the source code of the methodology described.
Open Datasets Yes We used a large 4-neighbor 194 x 194 2D grid with 51.3% blocked cells, namely the benchmark game map lak503d from (Sturtevant 2012).
Dataset Splits No The paper does not provide specific dataset split information (exact percentages, sample counts, citations to predefined splits, or detailed splitting methodology) needed to reproduce the data partitioning into train/validation/test sets.
Hardware Specification Yes All algorithms were written in C++ and ran on a 2.80 GHz Intel Core i7-7700 laptop with 8 GB RAM and a runtime limit of 5 minutes.
Software Dependencies No The paper mentions that algorithms were written in C++ and that an off-the-shelf ILP solver was used, but it does not provide specific version numbers for any software components, libraries, or solvers.
Experiment Setup Yes Each agent is a 2.5 2.5 square whose reference point is its top-left corner. All algorithms use Equation (1) to detect conflicts. All MC-CBS variants use the rectangle constraints discussed in Section 4.3. For MAX, we tested lookahead depths d from 0 to 4. We used 50 instances with randomly generated start vertices and goal vertices for each number of agents.