A Deep Reinforcement Learning Approach for Large-Scale Service Composition

Abstract

As service-oriented environments become widespread, there exists a pressing need for service compositions to cope with the high scalability, complexity, heterogeneity and dynamicity features inherent in these environments. In this context, reinforcement learning has emerged as a powerful tool that empowers adaptive service composition in open and dynamic environments. However, most of the existing implementations of reinforcement learning algorithms for service compositions are inefficient and fail to handle large-scale service environments. Towards this end, this paper proposes a novel approach for adaptive service composition in dynamic and large-scale environments. The proposed approach employs deep reinforcement learning in order to address large-scale service environments with large number of service providers. Experimental results show the ability and efficiency of the proposed approach to provide successful service compositions in dynamic and large-scale service environments.