Multicopy Reinforcement Learning Agents

Alicia P. Wolfe, Oliver Diamond, Brigitte Goeler-Slough, Remi Feuerman, Magdalena Kisielinska, Victoria Manfredi

Keywords: Cooperative Agents, Multiagent RL

TL;DR: The paper examines agents which attempt to overcome uncertainty by making copies of themselves.

Abstract: Inspired by the problem of optimal packet duplication in Wireless Networks, this paper examines a specific type of cooperative multi-agent problem in which an agent makes multiple identical copies of itself in order to achieve a noisy and difficult single agent task more reliably or efficiently. The agent must balance the cost of sending more copies with the improvement in speed or success rate generated by the extra copies. Due to the space of possible joint reward functions, this specific case has some differences from the more general cooperative multi-agent problem. We propose a learning algorithm for this multicopy problem which takes advantage of the structure of the value function to efficiently learn how to balance the costs and benefits of adding additional copies.

Type Of Paper: Full paper (max page 8)

Anonymous Submission: Anonymized submission.

Submission Number: 40