Scalable and Safe Multi-Agent Motion Planning with Nonlinear Dynamics and Bounded Disturbances

Authors: Jingkai Chen, Jiaoyang Li, Chuchu Fan, Brian C. Williams11237-11245

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

Reproducibility Variable Result LLM Response
Research Type Experimental We demonstrate our method by benchmarking in 2D and 3D scenarios with ground vehicles and quadrotors, respectively, and show improvements over the solving time and the solution quality compared to two state-of-the-art multi-agent motion planners.
Researcher Affiliation Academia 1 Massachusetts Institute of Technology 2 University of Southern California
Pseudocode No The paper describes the methods and algorithms in prose, including mathematical formulations for MILP and an explanation of the Priority-based Search, but does not include any structured pseudocode or algorithm blocks.
Open Source Code No The paper does not provide any specific links to open-source code for the methodology described. It mentions using 'Dry VR (Fan et al. 2017)' and 'Gurobi 9.0.1 (Gurobi Optimization 2020)' which are external tools.
Open Datasets Yes We compare S2M2 against ECBS-CT on two benchmark maps from the Grid-Based Path Planning Competition (GPPC)2, namely Arena (49 × 49) and Den502d (251 × 211). We use map Wall (13 × 13 × 5) from (Hönig et al. 2018). (Footnote 2: GPPC: https://movingai.com/GPPC)
Dataset Splits No The paper uses benchmark maps for evaluation and discusses random generation of initial and goal locations but does not explicitly describe train/validation/test dataset splits (e.g., percentages or sample counts) for its experiments.
Hardware Specification Yes All experiments were run on a 3.40GHZ Intel Core i7-6700 CPU with 36GB RAM with a runtime limit of 100s.
Software Dependencies Yes We used Dry VR (Fan et al. 2017) to generate reachability envelopes and Gurobi 9.0.1 (Gurobi Optimization 2020) as the MILP solver.
Experiment Setup Yes We set the maximum velocity in both planners to be 1. The cost multiplier for the motion primitive model is set to 1, which means no preferred action is specified. We set the focal weight ω (i.e., suboptimality ratio) for ECBS-CT to 1.2 and 1.5. In post-processing, we set the total iterations to 7 and continuity degree to 4.