GoBigger: A Scalable Platform for Cooperative-Competitive Multi-Agent Interactive Simulation

Authors: Ming Zhang, Shenghan Zhang, Zhenjie Yang, Lekai Chen, Jinliang Zheng, Chao Yang, Chuming Li, Hang Zhou, Yazhe Niu, Yu Liu

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

Reproducibility Variable Result LLM Response
Research Type Experimental We evaluate several state-of-the-art algorithms on Go Bigger and demonstrate the potential of the environment.
Researcher Affiliation Collaboration Sense Time Research, Shanghai AI Laboratory, Shanghai, China
Pseudocode No The paper describes algorithms for collision detection and neural network architectures, but these are presented as textual descriptions and figures, not formal pseudocode blocks or algorithms.
Open Source Code Yes Both Go Bigger and its related benchmark are open-sourced. More information could be found at https://github.com/opendilab/Go Bigger.
Open Datasets No The paper utilizes its own simulation environment, Go Bigger, for experiments rather than a pre-existing, publicly available dataset in the typical sense.
Dataset Splits No The paper does not provide specific training/validation/test dataset splits (e.g., percentages or counts) for reproduction. It evaluates agents within a continuous simulation environment.
Hardware Specification No The paper does not specify the hardware (e.g., CPU, GPU models) used to run the experiments.
Software Dependencies No The paper mentions using established reinforcement learning algorithms and environment design principles (e.g., gym), but it does not provide specific version numbers for any software libraries or dependencies.
Experiment Setup Yes Appendix B.3 provides detailed experimental setups for each algorithm, including hyperparameters: "DQN The batch_size is set as 512. learning_rate is 1e-4, replay_buffer_size is 2e4." and similar for PPO, QMIX, MAPPO, VMIX, COMA.