Learning Verified Safe Neural Network Controllers for Multi-Agent Path Finding
Authors: Mingyue Zhang, Nianyu Li, Yi Chen, Jialong Li, Xiao-Yi Zhang, Hengjun Zhao, Jiamou Liu, Wu Chen
AAAI 2025 | Venue PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We validate our approach through shape formation experiments and UAV simulations, demonstrating significant improvements in safety and effectiveness in complex multi-agent environments. Experimental results are indeed promising. |
| Researcher Affiliation | Academia | 1College of Computer and Information Science, Southwest University, Chongqing, China 2 Zhongguancun Laboratory (ZGC Lab), Beijing, China 3Department of Computer Science and Engineering, Waseda University, Tokyo, Japan 4School of Computer & Communication Engineering, University of Science and Technology Beijing, Beijing, China 5School of Computer Science, University of Auckland, Auckland, New Zealand |
| Pseudocode | No | The paper describes the methodology and framework in prose and uses figures (e.g., Figure 1: Our framework, Figure 2: The loss function and neural network architecture) but does not contain explicit pseudocode blocks or algorithm sections. |
| Open Source Code | No | The paper does not provide any explicit statements about releasing source code or links to code repositories for the described methodology. |
| Open Datasets | Yes | As shown in Fig.4, we have chosen three types of scenarios for simulation evaluation: (1) a random map, characterized by cuboidal obstacles that are positioned arbitrarily; (2) a cross map, wherein the spatial arrangement of obstacles emulates a crossroads; (3) a street map, abstracted from the open building dataset pertaining to Portland, USA (Burian et al. 2002). |
| Dataset Splits | No | The paper mentions training and testing phases and variations in the number of drones and maps used for testing (e.g., 'use a random map during training. In the testing phase, the number of drones is expanded from 4 to 1024, and the maps in-'), but it does not specify exact percentages, sample counts, or refer to standard splits for the dataset used. |
| Hardware Specification | Yes | All the simulation experiments are conducted on a desktop running Ubuntu 16, powered by Intel(R) Core(TM) i7-7700 CPU@3.6GHz and an Nvidia Quadro P600 GPU. |
| Software Dependencies | No | The paper mentions 'Ubuntu 16' as the operating system and 'Marabou' as a tool ('We verify hθ i stu,i using the Marabou (Katz et al. 2019)'), but it does not specify version numbers for Marabou or any other critical software libraries or frameworks used in the experiments. |
| Experiment Setup | Yes | c1, c2, c3, c4, c5 are set to 1, 1, 1, 0.1, and 0.05, respectively. γ1 = γ2 = γ3 = 0.01. |