Study Of Vertical Handoff Algorithm In Heterogeneous Networks

With the growing of wireless access technology, the next generation of wirelesscommunication network (4G) would be a heterogeneous network architecture composed of differentwireless access technologies. As one of key underlying technologies in the convergence ofheterogeneous networks, it is imperative to investigate how to smoothly switch amongheterogeneous access networks, i.e., strategies of vertical handover. This thesis theoreticallyanalyzes and experimentally studies the vertical handover algorithms in heterogeneous networks.Firstly, some basic theories are introduced, such as the characteristics of UMTS and WLAN,the concept of vertical handoff, integrated structures of UMTS and WLAN. Then, existingstrategies of vertical handover are analyzed and summarized.Secondly, this thesis presents a new type of analytical model which can theoretically evaluatethe performance indicators of various vertical handover algorithms, including the number of totalhandovers, the resource utilization of network and some QoS factors in real-time applications. Then,for two commonly used vertical handover algorithms, the theoretical performance values obtainedby the proposed analytical model are compared with those obtained through simulations. Theconsistency between theoretical and simulated values illustrates that our proposed analytical modelcan evaluate well the performance indicators of various vertical handover algorithms. Meanwhile,we also briefly discuss the intrinsic tradeoff among vertical handover algorithms.Finally, a new vertical handover scheme, the neural network based difference threshold fuzzylogic algorithm NNDTFH is proposed in this thesis. First by pre-judgment, NNDTFH reduces thesampling points which need subsequently processing, and therefore lowers system load; then, signalstrength in next moment is predicted by BP neural network, thus network conditions can beestimated in advance; then, the overall performance of the current access network is obtained byfuzzy logic processing; finally, the handover decision is made based on difference threshold. TheNNDTFH is simulated in Matlab, and compared with two commonly used algorithms: HVH andDTVH. Results show that the proposed algorithm NNDTFH not only mitigates the ping-pong effect,but also prolongs the time when UE stays in WLAN effectively, as a result, improve the systemperformance effectively.