Individual and Social Behaviour in the IPA Market with RL

摘要

Market-based mechanisms offer a promising approach for distributed resource allocation. Machine Learning has been proposed to influence and optimize market-based resource allocation. In particular, Reinforcement Learning (RL) has been used to improve the allocation in terms of utility received by resource requesting agents in the Iterative Price Adjustment (IPA) mechanism. This paper analyses the individual and social behaviour of agents in the IPA market-based resource allocation with RL. In particular, it presents results of experimental investigation on the influences of the amount of learning in the agents' behaviour aiming at determining how much learning is sufficient and the theoretical-experimental explanation of the agents' behaviours using game theory.