Multi-agent Reinforcement Learning for Service Composition

This paper investigates the multi-agent cooperation problems in Web services domain. For Pareto-optimal Nash equilibrium, reinforcement learning algorithms are used to solve the coordination problem in cooperative environments. Most previous works study the deterministic gain of a state. However, in practical service environments, the gain may be nondeterministic due to unstable Quality of Service (QoS). To avoid local optimal solution, we use an improved update function. The experimental results show that proposed reinforcement learning algorithm outperforms other learning methods.