Heterogeneous Wireless Network Access Decision Algorithm Based On Deep Q Network

Mobile cellular network and wireless local area network,as the two mainstream wireless access methods at the present stage,have a large number of overlapping coverage areas in many scenes.However,these two networks are incompatible with each other due to the different technical design,which brings great challenges to the integration of heterogeneous wireless networks.Therefore,the technology of heterogeneous wireless network convergence is a subject worthy of study.To solve the technical problems of heterogenous wireless network convergence,this paper designs a communication protocol for unified control in heterogeneous wireless network.A heterogeneous wireless network convergence architecture and related communication procedure are proposed based on the software defined network technique.By using the method of traffic-labeling,it achieves heterogeneous convergence and seamless handover with about 3ms handoff delay.To solve the optimizing problems of heterogenous wireless network access decision,this paper designs a self-learning strategy based on deep reinforcement learning technique,which can model the raw wireless signal automatically.Based on the self-learning strategy,we propose a new access decision alogorithm.By taking the training method from deep q network,it achieves real-time heterogeneous access strategy with high throughput,high stability and low ping-pong effect.