Fault-Tolerant Offline Multi-Agent Path Planning

Authors: Keisuke Okumura, Sébastien Tixeuil

AAAI 2023 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We evaluate DCRF with named failure detectors in grid environments and observe that DCRF can address more problem instances compared to computing a set of vertex disjoint paths... This section evaluates DCRF in both SYN and SEQ with NFD. We present a variety of aspects including merits to consider MAPPCF and bottlenecks of the planning. Figure 7 shows the results.
Researcher Affiliation Academia 1 Tokyo Institute of Technology, Japan 2 Sorbonne University, CNRS, LIP6, Institut Universitaire de France, France
Pseudocode Yes Algorithm 1 presents DCRF.
Open Source Code Yes The supplementary material is available on https://kei18.github. io/mappcf.
Open Datasets Yes Each scenario was tested on two four-connected grids (size: 32x32 and 64x64) obtained from (Stern et al. 2019).
Dataset Splits No The paper mentions preparing '25 well-formed instances' but does not specify any training, validation, or testing splits for these instances, nor does it refer to a validation set in the context of model training.
Hardware Specification Yes The experiments were run on a desktop PC with Intel Core i9-7960X 2.8 GHz CPU and 64 GB RAM.
Software Dependencies No The paper states 'The code was written in Julia', but it does not specify a version number for Julia or any other software libraries or dependencies used in the experiments.
Experiment Setup Yes The initial paths were obtained by prioritized planning... Single-agent pathfinding was implemented by A , adding a heuristic that penalizes the use of common vertices... We applied the refinement over initial paths... The attempt is a failure if it returns a solution before the timeout of 30 s.