Expressive Multi-Agent Communication via Identity-Aware Learning
Authors: Wei Du, Shifei Ding, Lili Guo, Jian Zhang, Ling Ding
AAAI 2024 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Experimental results on various benchmarks demonstrate IDEAL can be flexibly integrated into various multi-agent communication methods and enhances the corresponding performance. We conduct various experiments on four MARL benchmarks: Predator Prey, Traffic Junction, Battle, and SMAC. |
| Researcher Affiliation | Academia | 1 School of Computer Science and Technology, China University of Mining and Technology, Xuzhou 221116, China 2 Mine Digitization Engineering Research Center of Ministry of Education of the People s Republic of China, Xuzhou 221116, China 3College of Intelligence and Computing, Tianjin University, Tianjin, 300350, China |
| Pseudocode | Yes | Algorithm 1 shows the instantiation of IDEAL. |
| Open Source Code | No | The paper does not provide any explicit statement or link indicating that the source code for the described methodology is publicly available. |
| Open Datasets | Yes | To demonstrate the effectiveness of IDEAL, we conduct various experiments on four MARL benchmarks: Predator Prey, Traffic Junction, Battle, and SMAC. We employ the Predator-Prey benchmark introduced in (Singh, Jain, and Sukhbaatar 2019). We utilize the Traffic Junction benchmark as introduced in (Sukhbaatar, Fergus, and et al. 2016). We select Battle scenario from MAgent (Zheng et al. 2018). The Star Craft Multi-Agent Challenge (SMAC) (Vinyals et al. 2019) is a benchmark designed within the popular strategy game Star Craft II. |
| Dataset Splits | No | The paper mentions that training occurs (e.g., "agents are trained not in a supervised learning way but in the typical MARL way (using reward signal)"), and evaluates performance on benchmarks, but it does not specify explicit training, validation, or test dataset splits (e.g., percentages or sample counts) for the used datasets. |
| Hardware Specification | No | The paper does not provide any specific details regarding the hardware used for running the experiments, such as GPU or CPU models, or memory specifications. |
| Software Dependencies | No | The paper mentions various multi-agent reinforcement learning methods and graph neural networks, but it does not specify any particular software dependencies with version numbers (e.g., programming language versions, library versions) required to reproduce the experimental environment. |
| Experiment Setup | Yes | The detailed hyper-parameters are given in the Appendix. All implementations have been extended to support multiple rounds of message-passing... The two difficulty levels are defined as follows: 10 × 10 grid with 5 predators, and 20 × 20 grid with 10 predators. ... Nc= 10, p = 0.2... with Y = 40 and Z = 24. ... reduces the scope of vision for ally agents from 9 to 2. |