事件驱动Q学习在呼叫接入控制中的应用

Application of event driven Q-leaning in call admission control

文章研究了计时报酬方式下最优呼叫接入控制问题,建立了系统的连续时间Markov决策过程(CT-MDP),根据系统特征引入后状态Q值更新方法,给出呼叫接入控制问题基于事件驱动Q学习优化算法,并给出一个数值仿真实例;仿真结果表明,该算法比Q学习具有收敛速度快、存储空间小的优势;根据实验结果分析了在最优策略下业务拒绝率与业务特征的关系。

Optimal call admission control（CAC） based on time compensation is concerned in this paper. The continuous-time Markov decision processes（CTMDP） for the system is established, and a method of afterstate Q-value updating is introduced according to the characteristics of the system. Then an optimal algorithm of event driven Q-learning is proposed to solve the call admission control problem. Finally, an example of numerical simulation is given. The simulation results show that the proposed al- gorithm needs less memory and has faster convergence than Q-learning. And on the basis of the experimental results, the relationship between the rejection rate of business and the characteristics of business is analyzed under optimal call admission policy.