An Agent-Based Self-Adaptive Mechanism with Reinforcement Learning

In order to solve the problem in choosing action for a system in a dynamic environment, a self-adaptive mechanism combining the technology of agent and reinforcement learning is presented in this paper. With such a mechanism, the system determines all possible initial states of the agent's execution strategy, and adopts Q-learning algorithm on all the initial states. And then, the best result of all learning results is chosen as the current execution strategy. Meanwhile, agents can share learning results to improve the efficiency of the system. At the end of this paper, a case study is illustrated to validate the effectiveness of the proposed mechanism.