Effective and Stable Role-Based Multi-Agent Collaboration by Structural Information Principles
نویسندگان
چکیده
Role-based learning is a promising approach to improving the performance of Multi-Agent Reinforcement Learning (MARL). Nevertheless, without manual assistance, current role-based methods cannot guarantee stably discovering set roles effectively decompose complex task, as they assume either predefined role structure or practical experience for selecting hyperparameters. In this article, we propose mathematical Structural Information principles-based Role Discovery method, namely SIRD, and then present SIRD optimizing MARL framework, SR-MARL, multi-agent collaboration. The transforms discovery into hierarchical action space clustering. Specifically, consists structuralization, sparsification, optimization modules, where an optimal encoding tree generated perform abstracting discover roles. agnostic specific algorithms flexibly integrated with various value function factorization approaches. Empirical evaluations on StarCraft II micromanagement benchmark demonstrate that, compared state-of-the-art algorithms, SR-MARL framework improves average test win rate by 0.17%, 6.08%, 3.24%, reduces deviation 16.67%, 30.80%, 66.30%, under easy, hard, super hard scenarios.
منابع مشابه
Multi-Agent User-Centric Specialization and Collaboration for Information Retrieval
The amount of information on the World Wide Web (WWW) is rapidly growing in pace and topic diversity. This has made it increasingly difficult, and often frustrating, for information seekers to retrieve the content they are looking for as information retrieval systems (e.g., search engines) are unable to decipher the relevance of the retrieved information as it pertains to the information they a...
متن کاملStable Multi-agent Systems
We present an abstract declarative semantics for multi-agent systems based on the idea of stable set, and argue that it can be suitably employed to describe, and to some extent verify, the dynamics of complex systems of autonomous and heterogeneous interacting agents. We view agents as black-boxes, whose semantics is abstractly understood as an input-output transformation from the agents’ obser...
متن کاملA multi agent method for cell formation with uncertain situation, based on information theory
This paper assumes the cell formation problem as a distributed decision network. It proposes an approach based on application and extension of information theory concepts, in order to analyze informational complexity in an agent- based system, due to interdependence between agents. Based on this approach, new quantitative concepts and definitions are proposed in order to measure the amount of t...
متن کاملMulti-agent systems: effective approach for cancer care information management.
Physicians, in order to study the causes of cancer, detect cancer earlier, prevent or determine the effectiveness of treatment, and specify the reasons for the treatment ineffectiveness, need to access accurate, comprehensive, and timely cancer data. The cancer care environment has become more complex because of the need for coordination and communication among health care professionals with di...
متن کاملذخیره در منابع من
با ذخیره ی این منبع در منابع من، دسترسی به آن را برای استفاده های بعدی آسان تر کنید
ژورنال
عنوان ژورنال: Proceedings of the ... AAAI Conference on Artificial Intelligence
سال: 2023
ISSN: ['2159-5399', '2374-3468']
DOI: https://doi.org/10.1609/aaai.v37i10.26390