The mobile nodes (MNs) roam around the WCDMA/WLAN interworking networks at any place, any time by two different radio frequency (RF) modules for two radio interfaces. Thus, the dual-mode mobile device with two radio interfaces is able to connect with WCDMA and WLAN networks simultaneously. Moreover, each MN has an MAC layer to combine multiple services into a single data stream so that they can transmit over a single radio interface.
When an MN requests a service, the service can choose its target network according to the result by the proposed method as long as the radio modules are available. During the transmitting, the service can be configured to attach to another network for transmission quality consideration. On the contrary, the handoff decision is a totally different situation, which depends on the availability of radio modules. The re-configuration is not allowed because selecting the occupied radio module may degrade the transmission quality of the existed services.
2.2.1 Mobility model
Here, the mobility pattern is characterized by the velocity and moving direction of MN.
Assume the current position of MN is −−−−−−−→
(x(ta), y(ta)) in the observation point ta, and after a period of observation time T , the MN will be in −−−−−−−→
(x(tb), y(tb)) where tb = ta+ T is the next
observation point. It is formulated as,
−−−−−−−→
(x(tb), y(tb)) =−−−−−−−→
(x(ta), y(ta)) + νT · d, (2.1)
where d is the unit vector of the MN’s moving direction, and ν is the velocity of the MN. The vector d of the MN in the urban area is different to in the suburban area because the streets in the urban area are often in a planning chart. Thus, d in the urban area is (1, 0), (0, 1), (0, −1) with 13 probability on each while in the suburban area is (1, 1), (1, −1), (−1, 1), (−1, −1) with
1
4 probability on each as shown in Fig. 2.2. The reason of the given value is that the MNs move either in going straight or in turning 90 degrees around the corner following the street deploying in the urban area; and on the contrary, the MNs in the suburban area moves randomly. Three velocities are adopted: 3 kilometers maximum per hour for pedestrians, and 40 kilometers and 80 kilometers maximum per hour for vehicles. When the MN hits the boundary of system coverage, the MN’s position will be reset to a randomly given position where is not alike the original home cell.
Figure 2.2: (a) Mobility scenario in the urban area, and (b) Mobility scenario in the suburban area.
2.2.2 Service classes
3GPP [18] defines four traffic classes, which involves conversational class, streaming class, interactive class and background class. The conversational class represents real-time multi-media applications such as telephony and video. The steaming class is for streaming type of applications, like video on demand (VOD). The interactive class includes applications for Web-browsing, chat room, etc. The background class involves others services that uses best effort transmission. In the thesis, four traffic classes are clustered into two groups, real-time group and delay-tolerant group. As implied by the name, the real-time group includes most delay sensitive applications which involves conversational and streaming class. Compared to the real-time group, the delay-tolerant group cares data error rate more, which includes interactive and background class.
According to the service requirement in real-time group, the QoS constraints contain low delay, low bandwidth variation, handoff frequency and low jitter. Also the real-time group is able to tolerant to a certain level of packet loss. On the other hand, the QoS constraints of delay-tolerant group include data integrity, high bandwidth and packet loss to have a reliable transmission. The transmission rate variations are allowed in delay-tolerant group. Here, we consider voice service and Web-browsing as representative services of real-time group and delay-tolerant group, respectively. We adopt user mobility and QoS requirements, involving delay, bit error rate and data rate, for the utility function.
2.3 Source Model
As mentioned before, voice and Web-browsing services are considered in the thesis. The source models of these services are shown below.
Figure 2.3: A pattern of on-off voice model.
2.3.1 Voice service
Packet voice streams are based on the classic on-off model as shown in Fig. 2.3. This assumption is based on the observation that the human speech is decomposable in talkspurt (ON) and silence (OFF) period, where α and β are the mean of talkspurt and silence periods, respectively. The two periods are assumed to be exponentially distributed.
2.3.2 Web-browsing service
A Web-browsing session model is shown in Fig. 2.4 [19]. The session is composed by
Figure 2.4: A pattern of Web-browsing session model.
number of packet calls and the reading time for simulating the behavior of Internet surfing.
When the MN clicks a link on Internet, a packet call including a number of datagrams is transmitted to the MN via the wireless medium. After the MN downloads the information, a period of reading time is needed to digest the information. The length of packet call Npc is generated by geometrical distribution with mean κNpc. The packet calls are separated by
reading times Dpc which simulates the user’s studying the information and are also exponen-tially distributed in mean time κDpc. During a packet call, the number of datagrams Nd is geometrical distributed, whose mean is κNd. The size Sdof a datagram is Pareto distributed, and is generated with mean κSd. The datagrams are parted by inter-arrival times Dd which is also in geometrical distribution with mean κDd. In the simulations, we assume that the dwell time of a WWW session is exponentially distributed and the Web session arrivals is based on Poisson distribution.
Chapter 3
Utility Function-based Access Selection (UFAS) Method
3.1 Introduction
The utility function-based access selection (UFAS) method contains three parts, cells classification, utility function (UF) computation, and target cell determination. As shown in Fig. 3.1, for a request from the context manager module within RNC, the cells classification is to filter out unfeasible cells for the MN. Every cell in the candidate cell group comes through the UF computation to get its utility value. Based on the utility values, the target cell determination selects a best-fit cell for the service on an MN. In the following subsections, each part is described in details.