Learning Optimization Approach To Call Admission Control

Resource management is significant in the network control. Especially with the emergence of new business networks, it has become more prominent in the network control. As the network resources are limited, it's each network operator's goal that how to improve customer satisfaction, get the most appropriate allocation of resources and enhance the revenue. Call Admission Control (Call Admission control, CAC) is an important part of the management of network resources. which can achieve the rational allocation of resources, reach the goal of improving the network for revenue and customer satisfaction through actualizing different admission policies for different types of operational implementation. The admission control is studied by reinforcement learning approach in the integrated service networks and cellular wireless communication networks in the paper.CAC under the fixed reward and cumulative reward is studied in the integrated services networks. The system is model as continuous-time Markov decision process (CTMDP). Combined event-driven optimization of thought and the characteristics of CAC, the afterstates event-driven Q learning is proposed, which slove the application of algorithm to CAC. The experiment results show that the algorithm proposed can solve admision control problem effectively and need less data storage.The admission control problem of handoff call priority is studied in the cellular wireless communications networks. According to the problem description, we model it as CTMDP. The priority of handoff call is improved by giving it a greater reward, and we use event-driven Q learning algorithm to solve the problem. Finally, simulation examples illustrate that, compared to always accept policy, the event-driven Q Learning can heighten profit and lower handoff call dropping probability.