Research On The Handoff Mechanism Of WLAN Based On Hybrid Markov Prediction Model

Because of its good bandwidth with 54Mbps or more,IEEE 802.11 WLANhas been extensively used currently. At the same time, most of the applicationsconcentrate on data services, and also the performance of multimedia services areunsatisfying which are normal applications on wired LAN such as VoIP,audiofrequency,video frequency etc. Especially, when the users are moving andmaking handoff between different AP, the QoS of transmitting may be impactedsignificantly. Thus, more researches should be done on the technology of how tosatisfy the real time communication. This paper's aim at the things about WLANas follows: how to reduce the decrease of performance of terminal when makinghandoff, such as reduce the handoff delay and packets dropping, and how toreduce using of channel resources and storage space.This paper analyzed and compared all the modes of handoff, and concludedas follows. Hard handoff mode has small delay while drops more packets, andsmoothness handoff mode just the other way round. Forehand transmitting basedhandoff mode both has small delay and drops less packets, but it has to use morechannel resources and storage space. All of them use the channel search mode tofind the target cell of the handoff. According to the three steps of handoff, thereare three kinds of delays, delay of searching,delay of authentication and delay ofre-association. We can see that all the three handoff modes have the same delay ofsearching, which made us to think about the using of technology of mobile pathprediction. If the new AP can be predicted, the search process would be simplified,and the authentication and packet forwarding would be done previously, thusdelay is reduced.Based on the thoughts above, the first thing to do is how to predict the newAP. Many scientific institutions and scholars have researched on the mobileprediction and proposed some solutions currently, but we have to select apredictor with good performance in real application. Scholars in DartmouthCollege in US compare and analyze these solutions based on users' mobilitytraces collected from their own campus's wireless network, and found that simplerk-order Markov predictors performed as well or better than other predictors. Theyalso indicate that 2-order Markov predictor is the best one in k-order Markovpredictors. Unfortunately, 2-order Markov predictor has a problem that its statespace expands too fast as the increasing of AP's number, which challenges theMarkov predictors' practicability in this field.Aimed at k-order Markov predictors' state space expanding problem, thispaper proposed k-step Markov model, and validated its feasibility. And based onthis , proposed a hybrid multi-step Markov model, which uses EM arithmetic toestimate the hybrid-parameter of several single Markov models. This new model'scomplexity of state space is O ( N ) while k-order Markov model's is O ( N 2), andnew model's complexity of storage space is O ( N 2) while k-order Markovmodel's is O ( N 3),where N is the number of the WLAN. By calculating theconditional entropy and analyzing the practice on real data set, it is finally provedthat the new model has the similar performance with k-order Markov predictormodel while uses smaller costs, It can preferably predict the new AP(success ratiocan reach 75～80%).Based on the prediction of new AP using this new Markov predictor model,and associated with forehand transmitting based handoff mode, this paperproposed a new handoff mode which named as prediction based handoff mode,and also, it analyzed every step of the whole handoff process.When the MH can work normally after entering a cell, we can use the hybridmulti-step Markov model to predict the new AP, and then inform the current AP toprocess pre-authentication to new AP. This paper is based on the service setidentifier--SSID closed system authentication mode, and describes thepre-authentication process in detail. AP. Authentication process of AP to MH iscompleted before the handoff, and not be impacted from MH mobile speed, thus,decreased handoff's time. And authentication process is not between thepredictive AP and MH, but between the current AP and the predictive AP, thus,reduce occuping the wireless channel of MH.Though the AP can be predicted by predictor, the success ratio is between75% and 80%, not 100%. Therefore, we should consider the situations with wrongprediction and correct prediction, then, first of all need to be resolved is thequestion of how to judge the prediction is correct or not, this is done in phasesjudgement.The first phase, when the RSSI of MH receiving from the current APbelows a certain threshold, MH sends active-scan-request to prediction-Ap. If MHreceives scan-response, it indicates that MH has entered prediction-Ap area;IfMH does not receive scan-response, it indicates that MH has not enteredprediction-Ap area, it shows a failure prediction.The second phase, although MHreceived scan-response from prediction-Ap, but it is difficult to forecast whetherMH can move to prediction-Ap's cell with uniform velocity or not, and therefore,it needs stakeout for them, and process further judgement. We scout the changesof RSSI receiving from current-AP and prediction-Ap, then, have the timelyjudgement: MH moves to prediction-Ap, MH moves to current-AP or dosen'tmove, and MH moves to a third AP. It shows wrong prediction if MH moves to athird AP.If it is correct of the prediction, then, MH sends message to current-AP andinform transmiting packets to prediction-AP. If it is wrong of the prediction, then,MH processes the normal mode of searching AP, and transact the handoff processof based on AP pre-forward. Clearly, when it is successful of the prediction, MHonly sends scan-request to prediction-AP, it simplifies the scan process and savesthe handoff'time. Furthermore, MH only transmits packets to prediction-AP, itreduces using channel resources and storage space.This paper uses the concept of the two-tier coverage region, and analysisesthe factors of overlap region size, node movement speed and so on. Theconclusion is that it can reduce network costs when MH processes previousauthentication and transmits packets in the second layer, and in overlap region ofthe current AP and adjacent AP. It also use timer to control transmitted packets,and reduces channel resources and storage space wasted. This paper complete thetransmitted data packets using the encapsulation of the data frame ,and in thesame time the new AP set up a FIFO buffer for the MH to storage of data packets.Based on the analysis of the prediction of over cells handoff, this papercompletes the realization of the handoff mode in the AP protocol architecture ,andoperation flow of three stages -MH sends associate request to the AP,previous-authentication request to the new AP, and re-associate request to thenew AP.This paper also analyzed the performance of this proposed handoff mode,and combined the predictive probability of correct and wrong, then accounted theaverage of handoff delay and occupying channel and buffer. And, compared itwith other three modes. It can be concluded that this new mode has evidentlyreduced the handoff delay, and uses less channel resources and storage space.Therefore, it's performance is better than the other three modes.