Symmetry Breaking for k-Robust Multi-Agent Path Finding

Authors: Zhe Chen, Daniel D. Harabor, Jiaoyang Li, Peter J. Stuckey12267-12274

AAAI 2021 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Experimental results show very large gains in success rate for k-CBS; not only on classic MAPF benchmarks but also in two application specific domains: in warehouse logistics, where k-robust plans are desirable and in railway scheduling where k-robust plans are mandatory. and Experiments The implementation is based on CBS with rectangle, corridor and target reasoning of Li et al. (2020) and support for k-robust planning is added on top of it. It is programmed in C++ and experiments were performed on a server with AMD Opteron 63xx class CPU and 32 GB RAM.
Researcher Affiliation Academia Zhe Chen,1 Daniel Harabor,1 Jiaoyang Li,2 Peter J. Stuckey1 1Monash University, Australia 2University of Southern California, USA
Pseudocode No The paper describes algorithms and methods but does not provide structured pseudocode or algorithm blocks.
Open Source Code No The paper does not provide an explicit statement or link for the open-sourcing of their code for the described methodology.
Open Datasets Yes We use 25 even scenarios from movingai.com to evaluate our algorithms., We use a 31 79 Warehouse map (Li et al. 2020) with randomly generated problems to evaluate the performance of robots in warehouse system., The Flatland challenge is a railway scheduling challenge 1 provides a simplified railway simulator using a directed grid map, where trains cannot move backwards. Railway systems have headway control, one train cannot start to enter a railway block if another train currently occupies the block. Hence the railway domain requires k = 1 robust plans. Our experiments use flatland-rl v2.0.0 to generate experiment problems. Note, no target conflicts can occur in the Flatland challenge scenarios since trains disappear when reaching their destination. We have two settings for evaluation: (1) a 100 100 fixed dense map contains fixed 140 start and goal locations, and each pair of start and goal are connected by 5 railways; and (2) a 100 100 incremental sparse map has one start and goal location per agent, and each pair of start and goal are only connected by 1 railway. The experiment on railway maps, Figure 5f shows that KCBSH and K-CBSH-RM performs substantially better than KCBS. k-robust corridor reasoning helps further improve performance on sparse maps. and 1Swiss Federal Railways, 2019, Flatland Challenge. https://www.aicrowd.com/challenges/flatland-challenge
Dataset Splits No For each map, we keep increasing the number of agents, and for each number of agent we solve 25 different instances.
Hardware Specification Yes experiments were performed on a server with AMD Opteron 63xx class CPU and 32 GB RAM.
Software Dependencies Yes It is programmed in C++ and experiments were performed on a server with AMD Opteron 63xx class CPU and 32 GB RAM. and Our experiments use flatland-rl v2.0.0 to generate experiment problems.
Experiment Setup No The time limit is set to 90s for each instance.