Scalable Mechanism Design for Multi-Agent Path Finding

Authors: Paul Friedrich, Yulun Zhang, Michael Curry, Ludwig Dierks, Stephen McAleer, Jiaoyang Li, Tuomas Sandholm, Sven Seuken

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

Reproducibility Variable Result LLM Response
Research Type Experimental Our experiments verify that our mechanisms behavior follows our claims on optimality, scalability, and payments. Our benchmarks are strategyproof mechanisms whose allocation rule solves classical MAPF, i Bundle [Amir et al., 2015] being the most advanced competition. Figure 1 compares the success rate, runtime and welfare over FCFS of MCPP with different sample sizes m on the random-32-32-20 map.
Researcher Affiliation Collaboration 1ETH AI Center 2University of Zurich 3Carnegie Mellon University 4Harvard University 5University of Illinois at Chicago 6Strategy Robot, Inc., Strategic Machine, Inc., Optimized Markets, Inc.
Pseudocode Yes Mechanism 1: MCPP Input: Agent reports ( 1, . . . , n), No. of samples m Output: Conflict-free paths ( 1, . . . , n) and payments (p1, . . . , pn) Allocation rule: Prioritized Planning
Open Source Code Yes Code at https://github.com/lunjohnzhang/MAPF-Mechanism.
Open Datasets Yes We evaluate payment-CBS (PCBS), exhaustive-PBS (EPBS) and Monte-Carlo prioritized planning (MCPP) on four 2D maps selected from the commonly used MAPF benchmarks by Stern et al. [2019].
Dataset Splits No The paper evaluates mechanisms on instances but does not provide specific train/validation/test dataset splits (percentages or counts), nor does it describe cross-validation setups or predefined split files. It mentions '100 instances' for each run but not how those instances are split for training, validation, or testing.
Hardware Specification Yes Our code was implemented in C++ and ran on a 2x12-core Intel Xeon E5-2650v4 machine with 128 GB of RAM.
Software Dependencies No Our code was implemented in C++ and ran on a 2x12-core Intel Xeon E5-2650v4 machine with 128 GB of RAM. The paper only mentions 'C++' as the implementation language but does not provide specific software dependencies with version numbers (e.g., specific compilers, libraries, or frameworks with their versions).
Experiment Setup Yes MCPP uses a sample size of m = 100, except in Figure 1. Each run uses a set of 100 instances. We set a runtime limit for all mechanisms at 3600 seconds for the random-32-32-20 and dens312d maps and 5400 seconds for the dens520d and Paris-1-256 maps.