The main goal of modelling a Multi-Agent System (MAS) is finding its optimal joint policy, which maximize the collective payoffs of all agents. This can be a complex problem, especially for real-world systems, like heterogenes traffic aiming to ensure safe movements or optimize autonomous driving. Our updated work here: https://lnkd.in/eu4iYbu3 contains a multi-agent simulator with minimum requirements and elegant design, in addition to baselines and a novel multi-step reinforcement learning models, thoroughly evaluated on many relevant criteria, as shown in the video below. Paper published here: https://lnkd.in/ena2vDaE
An interesting idea.
Thanks for using our inD and uniD dataset 🙌