Multi-Agent Credit Assignment in Stochastic Resource Management Games

TY - JOUR

T1 - Multi-Agent Credit Assignment in Stochastic Resource Management Games

AU - Mannion, Patrick

AU - Devlin, Sam

AU - Duggan, Jim

AU - Howley, Enda

N1 - © Cambridge University Press, 2017. This is an author-produced version of the published paper. Uploaded in accordance with the publisher’s self-archiving policy. Further copying may not be permitted; contact the publisher for details.

PY - 2017/8/24

Y1 - 2017/8/24

N2 - Multi-Agent Systems (MAS) are a form of distributed intelligence, where multiple autonomous agents act in a common environment. Numerous complex, real world systems have been successfully optimised using Multi-Agent Reinforcement Learning (MARL) in conjunction with the MAS framework. In MARL agents learn by maximising a scalar reward signal from the environment, and thus the design of the reward function directly affects the policies learned. Inthis work, we address the issue of appropriate multi-agent credit assignment in stochastic resource management games. We propose two new Stochastic Games to serve as testbeds for MARL research into resource management problems: the Tragic Commons Domain and the Shepherd Problem Domain. Our empirical work evaluates the performance of two commonly used reward shaping techniques: Potential-Based Reward Shaping and difference rewards. Experimental results demonstrate that systems using appropriate reward shaping techniques for multi-agent credit assignment can achieve near optimal performance in stochastic resource management games, outperforming systems learning using unshaped local or global evaluations. We also present the first empirical investigations into the effect of expressing the same heuristic knowledge in state- or action-based formats, therefore developing insights into the design of multi-agent potential functions that will inform future work.

AB - Multi-Agent Systems (MAS) are a form of distributed intelligence, where multiple autonomous agents act in a common environment. Numerous complex, real world systems have been successfully optimised using Multi-Agent Reinforcement Learning (MARL) in conjunction with the MAS framework. In MARL agents learn by maximising a scalar reward signal from the environment, and thus the design of the reward function directly affects the policies learned. Inthis work, we address the issue of appropriate multi-agent credit assignment in stochastic resource management games. We propose two new Stochastic Games to serve as testbeds for MARL research into resource management problems: the Tragic Commons Domain and the Shepherd Problem Domain. Our empirical work evaluates the performance of two commonly used reward shaping techniques: Potential-Based Reward Shaping and difference rewards. Experimental results demonstrate that systems using appropriate reward shaping techniques for multi-agent credit assignment can achieve near optimal performance in stochastic resource management games, outperforming systems learning using unshaped local or global evaluations. We also present the first empirical investigations into the effect of expressing the same heuristic knowledge in state- or action-based formats, therefore developing insights into the design of multi-agent potential functions that will inform future work.

U2 - 10.1017/S026988891700011X

DO - 10.1017/S026988891700011X

M3 - Article

SN - 0269-8889

JO - The Knowledge Engineering Review

JF - The Knowledge Engineering Review

ER -