Decentralised Multi-Agent Reinforcement Learning for Dynamic and Uncertain Environments

DMCA

Decentralised Multi-Agent Reinforcement Learning for Dynamic and Uncertain Environments

Cached

Download Links

[arxiv.org]

by Andrei Marinescu , Ivana Dusparic , Adam Taylor , Vinny Cahill

BibTeX

@MISC{Marinescu_decentralisedmulti-agent,
    author = {Andrei Marinescu and Ivana Dusparic and Adam Taylor and Vinny Cahill},
    title = {Decentralised Multi-Agent Reinforcement Learning for Dynamic and Uncertain Environments},
    year = {}
}

OpenURL

Abstract

Multi-Agent Reinforcement Learning (MARL) is a widely used technique for optimization in decentralised control prob-lems. However, most applications of MARL are in static en-vironments, and are not suitable when agent behaviour and environment conditions are dynamic and uncertain. Address-ing uncertainty in such environments remains a challenging problem for MARL-based systems. The dynamic nature of the environment causes previous knowledge of how agents interact to become outdated. Advanced knowledge of poten-tial changes through prediction significantly supports agents converging to near-optimal control solutions. In this paper we propose P-MARL, a decentralised MARL algorithm en-hanced by a prediction mechanism that provides accurate in-formation regarding up-coming changes in the environment. This prediction is achieved by employing an Artificial Neu-ral Network combined with a Self-Organising Map that de-tects and matches changes in the environment. The proposed algorithm is validated in a realistic smart-grid scenario, and provides a 92 % Pareto efficient solution to an electric vehicle charging problem.

DMCA