Assessment of QoE and User Perception Models

Quality has been defined as the result of a perception and judgment process, during which the user compares the perceived characteristics of the speech sound (the so-called "auditory event") to the expected or desired characteristic. Because of the necessity of perception and judgment processes, subjective tests are still the only valid and reliable means for the purpose of quantifying the impact of different types of degradations on perceived overall quality (QoE). However, instrumental models such as PESQ and the E-model have shown to provide valid estimations of the results of perceptual tests, within the limits of applications they have been de signed for. As a consequence, these models are widely used instead of auditory tests, e.g. for network planning and monitoring. Still, their range of validity needs to be respected, and this is why they cannotsimply be applied without further validation to NGMN scenarios.

In the case of transmitted speech, which is the focus of the present chapter, the perceived quality of the given conditions is collected in auditory tests. In a listening-only situation, for instance, a selection of test participants is asked to judge the quality of a number of processed speech samples. The text material, read by different speakers, is chosen according to the aim of the experiment, and the recorded clean speech files are processed by the system of interest and are finally presented to the listeners. Their task is to judge upon the perceived quality of the processed speech sample, providing a quantitative measure of the QoE.

The Telecommunication Standardization Sector of the International Telecommunication Union (ITU-T) provides information about how such tests have to be performed in detail. There, it is specified how to choose "balanced" speech material, "normal" speakers as well as "normal-hearing" test participants. In order to increase the reliability of the experiment, the recording and play-back situation is specified, as well as further test parameters which might have a significant impact on the measurement results.

The judgments of the listeners are usually limited to the identification or scaling of pre-defined properties of the percept. Therefore, a set of predefined scales are available. In the area of speech transmission, a 5-point category scale is usually employed (Absolute Category Rating, ACR). For the collection of overall quality ratings, this 5-point scale is labeled with the attributes "Excellent", "Good", "Fair", "Poor", and "Bad". Subsequently, the ratings are averaged overall participants, leading to an arithmetic mean (Mean Opinion Score, MOS) for each processing condition.

In contrast to the listening-only situation, a bi-directional communication system is employed in real-world conversation situations. Thus, the ecologically most valid method for quality assessment is the conduction of conversation experiments with two interacting participants. Here, the interlocutors are asked to have a conversation on an arbitrary or pre-defined topic. By means of, e.g., the abovementioned ACR scale, the opinion about the just finished connection is supposed to be judged. 

Another standardized paradigm, which can be regarded as a trade-off between the "artificial" listening-only experiments and the quite complex conversation tests and it is also called the "CallQuality" method. In this process, the perceived quality of a "simulated" telephone conversation is assessed. The participants are asked to listen to short extracts of a normal conversation, and verbally answer questions regarding the content of the stimulus they just heard. After five of these stimuli, they rate the quality of the entire simulated conversation on the ACR scale. The answering part was introduced to come close to a real conversation with its turn-takings, and to distract subjects from concentrating on the quality until rating it (the test participants are instructed to try to put themselves into the position of an interlocutor).

Once the perceptual effects have been quantified, instrumental quality prediction models can be developed that are capable of estimating the subjective ratings. One type of recommended models by ITU-T, the so-called "Perceptual Evaluation of Speech Quality", is based on the application-layer speech signals. Here, the quality of transmitted speech in a listening-only situation is estimated by comparing the clean and degraded signals on a perceptual level, i.e. by taking advantage of psycho-acoustic knowledge, such as the Bark-scale transform, loudness functions, time/frequency masking, asymmetries of "positive" and "negative" error components, as well as insensitivities of certain variations in delay, the spectrum, and the amplitude. PESQ has been extended towards wideband speech by applying a flat input filter and a different mapping function.

Currently, the requirements for a successor of PESQ are discussed in Study Group 12 of ITU-T which is supposed to improve known drawbacks of PESQ and is valid for an even wider range of distortions (e.g., audio bandwidth, time-warping). NGMN-specific degradations, however, will not explicitly be covered by this model.

For conversational speech quality, the ITU-T recommends the so-called E-Model, which is a parametric model usually employed for offline quality estimation, e.g. for network planning. The MOS values are estimated on the basis of transmission channel parameters commonly known in classical telephonometry (e.g., loudness ratings, noise levels), but also in the context of packet-based networks (e.g., packet loss rates). These parameters are subsumed by so-called impairment factors, which are assumed to be additive on a psychological scale, the so-called transmission rating scale (R-scale). The eventual quality estimates are then obtained by a non-linear transformation of the R-values, i.e. the summations of the perceived impairments.

Until now, none of these models have been validated to correctly predict the effects of NGMN handovers and/or codec changeovers on user perception. However, such instrumental models are indispensable to rapidly design network handover and codec changeover strategies which provide an optimum quality to the user. The auditory investigations presented in this chapter can therefore be considered as a basis for the development of NGMN-capable quality models.

Mobility in 4G Networks

The Internet Protocol (IP) was developed in the early 1980s with the aim of supporting connectivity within research networks. However, in the last decade IP has become the leading networking protocol. It is the basic tool for a plethora of client/server and peer-to-peer networks; it is predominant in both wired and wireless worlds, and now the current scale of deployment is straining many aspects of its more than 30 year old design. To overcome the limitations inherent in the original IP design, IPv6 has been proposed as the new protocol that will provide a firmer base for the continued growth of today's complex networks.

More people will access the Internet via wireless rather than via wired connections, and each user will have a set of wireless devices interconnected that will be accessing a great variety of IP-based services. There are over 3 billion mobile subscriptions in the world, and this number continues to grow, one for every two inhabitants. In the light of this growth, IPv4 faces many problems related to its limited address space and mobility capabilities applied to the mobile world.

Every mobile device is potentially capable of accessing IP services, Wi-Fi networks are becoming ubiquitous; the growth in the hotspot market is being accompanied by similar growth in other wireless technologies (e.g., Bluetooth, Ultra Wideband, and satellite), posing the urgent need for a new identification scheme and an adequate support for mobility.

Whereas the main thrust of IPv6 is to solve the addressing problem, it also provides important functions to enable mobility (e.g., scaling and ease-of-configuration). IPv4 has difficulties managing mobile terminals for several reasons such as address configuration and location management. However, in order to drive the evolution in the current mobile world and avoid the humongous effort to migrate all computers and equipments from IPv4 to IPv6, Charles Perkins, from Nokia Labs, originally proposed Mobile IPv4 (MIPv4). This protocol extension was projected to enable IPv4 devices to support micro and macro mobility. MIPv4 has disadvantages in comparison with its successor, Mobile IPv6 (built as a natural part of IPv6 protocol, and less of an extension, as in the case of IPv4), but neither Mobile IPv4 nor Mobile IPv6 were intended to support seamless roaming in heterogeneous environments such as 4G Networks.

The rapidly growing demand for "anywhere, anytime" high-speed access to IP-based services is becoming one of the major challenges for operators. As the demand for mobility increases, mobile terminals need to roam freely across heterogeneous systems forming the present wireless landscape. Currently, Mobile IP stands as the de facto solution for mobility management in 4G networks.

This work targets the user-driven evaluation of mobility in 4G networks, using MIPv4 as the underlying support protocol (Figure 1). In particular, it is of interest to evaluate the impact of vertical handovers (i.e. network handovers) and terminal mobility on user perception. A network handover occurs when a mobile device changes its point of attachment to the Internet and the former point belongs to a different wireless technology; for example, a mobile phone handing off from the cellular network to a public hotspot. A brief introduction to the basic concepts of Mobile IPv4 is included next to introduce readers to its basic concepts.

 

Figure 1: Mobile IPv4 protocol

Mobile IP Version 4

Mobile IPv4 defines mechanisms that add support to a terminal (i.e. the Mobile Node) to change its point of attachment to the Internet whilst remaining reachable through a permanent address (the Home Address, HoA) and preserving all the active connections it had before travelling to its new location. While a Mobile Node (MN) is connected to its Home Network (i.e. the network its Home Address belongs to), no special mode of operation is needed, and packets are forwarded (using normal IP routing) between the Mobile Node and any other node it is communicating with (the Correspondent Node, CN). When a MN is connected to a network other than its Home Network (i.e. it is visiting a Foreign Network), the MN acquires an IPv4 address belonging to the address space of the Foreign Network it is visiting (supported by the Foreign Agent, FA), called the Care-of Address (CoA).

The MN announces its CoA (by sending a Binding Update message, BU) to a special entity, called the Home Agent (HA) that is located in the MN's Home Network. The HA intercepts the packets addressed to the MN's Home Address while the MN is away from its Home Network and establishes a bidirectional tunnel with the MN's CoA, in order to redirect these packets to the MN's current point of attachment to the Internet. The MN also uses this tunnel to send its traffic to the Correspondent Node, avoiding in this way any ingress filtering.

Latency on MIPv4-enabled links can be very high, especially for interactive applications that require real-time response. Therefore, the research community is working on mechanisms that decrease this latency as much as possible, at least to levels that support real-time applications. Two of the most significant proposals are Fast Handovers for Mobile IPv6 (FMIPv6) and Hierarchical Mobile IPv6 (HMIPv6).

FMIPv6 aims to decrease the total latency to almost only the Layer 2 handover time. This approach has been shown to perform well in intra-technology (i.e. horizontal) handovers. The HMIPv6 approach is designed to reduce the degree of signaling required and to improve handover speed for mobile connections by managing local mobility in a more efficient way. Previous work has analyzed which of these approaches (i.e. FMIPv6 and HMIPv6) performs better, the conclusion being that a combined approach would be optimal. However, given the implementation complexity that this would require, the FMIPv6 optimization by itself is good enough. 

Future 4G networks, where heterogeneity will be more the rule rather than the exception, have challenging characteristics when performing vertical handovers (also known as inter-technology). The present work focuses on the impact of this new type of mobility on the user perception. The enhancement (better performance, lower latencies, etc) of the underlying protocols (i.e. MIPv4) is not in the scope of this research. 

HWN* SIMULATIONS HWN* Framework

Accurate modelling of communication networks often results in models that are intractable to mathematical analysis and computer simulation becomes the option for modelling and analysis of complex systems. This section provides a description of the modelling concepts, simulation techniques and tools that have been used in the modelling and evaluation of the HWN* system. Algorithms proposed in conjunction with the HWN* infrastructure have been evaluated using OMNET++, which is a discrete event simulation environment with GUI support. OMNET++ is an object-oriented simulation tool, which consists of hierarchically nested modules written in C++. Nested modules are used that implement comprehensive and accurate modelling, which can reflect a good estimation of the actual system performance. In order to introduce events such as discontinuous transmission, session arrival & departure and etc, it is necessary to introduce the notion of time into the simulations. Events are triggered at some time instance in the simulation. For example, the event of session arrival at one node usually triggers the event of the arrival of that packet at a downstream node. Indeed the time delays are often random variables but events allow session arrival and departure processes to be modelled.

The HWN* physical layer needs to calculate the Bit Error Rate (BER) based on the SIR variation influenced by path loss, shadowing and fading models. The simulation of HWN* mainly considers packet level and session level modelling, but not bit level. Fast fading such as multipath fading is averaged out or captured in the BER and Packet Error Rate (PER). The propagation characteristics change from place to place when a MT moves. Thus, the transmission path between the transmitter and the receiver can be modelled from simple direct line of sight (LOS) to one that is severely obstructed by buildings, foliage and other terrain. Three mutually independent propagation phenomena considered are path loss, shadowing and multipath fading.

As complicated simulation models are evaluated, each simulation run may consume random numbers from several streams, which should be from several independent Random Number Generator (RNG) instances. Therefore different random streams or say simulation benches have been use for different tasks. For example, a random stream for generating packets and another random stream for simulating packet delivery rates in the transmission are different and not overlap. For all algorithm and system evaluation, we implement the Mersenne Twister RNG with 623-dimensional equal-distribution. The independent simulation processes run independently of one another and continuously send their observations to the central analyser and control process. This process combines the independent data streams and calculates from these observations an overall estimate of the mean value of each parameter. The Akaroa method provided by OMNET++ is also used which halts a simulation. It is decided by the 95% confidence intervals for all simulations to precise whether the result has enough observations.

Probabilistic models are applied to generate voice traffic, multimedia traffic and web traffic in this simulation environment for QoS oriented results evaluation. Voice traffic is modelled down to a packet activity with a different degree of granularity. Packet based voice source traffic models incorporate a voice activity aspect, which allows transmitting speech data when voice frames are detected, and facilitates discontinues data transmission. Hence, the voice source model has to include a higher degree of granularity and is represented as a sequence of consecutive talk spurts and silence periods within each call. The duration of the active and inactive periods reveals a negative exponential distribution for the duration of both active and inactive periods. The values applied for the investigations of this work are, a mean duration for the active periods of 1.35 seconds and a mean duration for the inactive periods of 1.70 seconds. The activity factor of the source is defined as the proportion of the time that the source is active. Several factors impact the nature of multimedia traffic and its transport requirements such as target quality, compression technique, coding time. The use of video traces is to facilitate algorithm evaluation with respect to QoS. As modeling the video sources always requires the original trace to be fully evaluated first, a simpler approach adopted in this research is to incorporate traces directly into OMNET++. Another benefit from incorporating the video traces directly into network simulators is the vast amount of video source models. Direct utilisation of video traces in OMNET++ facilitates the fastest method to incorporate video sources into existing network models. An interface has been provided for video trace is implemented, which is capable of detecting the different video trace file formats and to feed the data into the simulator accordingly. For particular environments and service types only specific traffic has to be instantiated. Whereas the evaluation of routing and inter-system traffic balancing of HWN* has instantiated all voice, multimedia and web services. Web traffic modelling is represented by a generic model with three levels session, activity and packet level. Session level consists of pages visited in a web session where a client starts an application, uses it for a time and then disconnects from the system. The moments when sessions arrive can be described by a Poisson model. The duration of sessions in real time applications depends on applications and in non real time is controlled by Transmission Control Protocol (TCP). Activity Level consists of a set of object applications such as images, sound and applet. The density of information in one application depends on the application itself. To describe this property, the activity level represents the application as a detailed succession of activity and inactivity periods in an ON/OFF model. ON represents page downloading time followed by an OFF period for the reading time. Packet Level decides the transmission of IP packets. If the User Datagram Protocol (UDP) is used, the packet inter arrivals can follow any distribution with packets following a truncated version of distributions to respect transmission limitation. In case of TCP, inter-arrivals of packets are determined by a TCP Pareto distribution.

To also include cellular network node movement characteristics, routing and handover algorithm evaluation of the HWN* system implements an Attraction (Attractor) Points oriented Mobility Model (APMM) based on the random waypoint model. The algorithm first selects N attractors that are distributed at points where MTs will originate from or progress towards. Prior to heading for attraction points, nodes are grouped together using Cell Type Transition Probability, each subscriber selects a destination area with probabilities. In a typical MT movement, at the beginning, all MTs are scatted around in a metropolitan environment. After 100, 150 and 200 simulated minutes, the trend of MTs have been moved to several pre-defined attractor points which locate in the middle of northwest, northeast, southwest, southeast and the city centre. The northeast and southeast regions may have higher attractor probability than the rest ofhot spots and therefore more MTs gather at the right hand side. It is flexible to change the geographical location of hot spots by revising the attractor points. Meanwhile, with a speed control mechanism, the current MT speed is configured correlated to the previous speed value and a smaller sampling time makes the speed change more smoothly.

HWN* configuration deals with message exchange, core network layer structure and individual algorithm implementation issues. The multiple access systems of the cellular component in HWN* deal with the inter-cell and intra-cell interference caused by common access to a shared band of frequencies. The mutual interference happens between BS and MT, RN cellular interface and BS, and RN cellular interface and MT. A TDMA based standard interface has been modelled. Each session is assigned a time slot within a frame which it keeps until it is handed off. No other sessions within the same cell are assigned the same slot, and thus users within a cell do not interfere. The MANET component of HWN* employs the contention based Carrier Sense Multiple Access/ Collision Avoidance (CSMA/CA) for the multiple access, between MT and MT, and MT and RN MANET interface, as it is the most adopted access protocol for MANETs and 802. 1x networks. When regarding the RN MANET interface as a MANET node, if MANET node cluster head selection is performed and along with RN provide the time beacon for synchronisation, a syncronised CSMA/CA can support fast data transfer, adopts ACK for successful transmissions and implements the handshaking mechanism between RN and MT to reduce collisions.

HWN* RN Site Planning

Plans have already been proposed to select RN sites for in other hybrid wireless networks. The solutions depend on the varied network performance objectives. Among these the packing based RN placement plan is simple, effective and straight-forward but is only suitable for ideal HWN* deployment scenarios where the BS sites are also packed with similar discipline plus the geographical site availability issues are not considered for both BS and RN. It is well known from planar geometry that to cover a two dimensional district with equal sized circles, the best possible packing solution can be obtained by surrounding each circle by six circles. But to have connections between the RNs to form a virtual RN backbone, an overlap is needed between relay cells. The framework therefore considers a situation where the locations of the RNs are centered with maximum coverage.

As the research addresses radio resource management, routing and node mobility problems, the RN site location planning should not be designed too idealistic such as in the packing based RN placement or not only include one or two parameters. We therefore propose a novel heuristic RN placement algorithm considers both physical distance through multi hop distance and Signal to Interference Ratio (SIR) through channel availability. The algorithm is devised in three steps described as:

1. Identify ideal RN locations based on radio resource management and MT mobility behaviour, then generate a set of RN position candidates.
2. Further formulate the RN site positioning as a constrained optimisation problem, of which the goal is to maximise the overall network throughput, the potential gain of MANET RN based services and minimise the hop distance and delay, so that more MTs can be served with guaranteed QoS.
3. Test RNs positioning sites combination recursively and update each RN's position based on performance result.

These three procedures are executed recursively until the algorithm converges. By imposing such apian, in practice, one can expect the system performance at disadvantaged locations should be improved with enlarged network dimensioning. In order to provide fixed RN assistance for both cellular and MANET interfaces radio resource management and routing, initial RN placement site candidates and possible topologies are considered in the following scenarios as presented in Figure 1. The scenario I, II and III cover the mobility management problems, scenario IV discusses the routing problem and scenario V is concerned with the relay structure.

 

Figure 1: The RN node test scenarios I, II and III for the cellular network, the test scenario IV for the MANET and virtual RN backbone scenario V for RN positioning evaluation

When RNs participate in cellular resource sharing and traffic handover along with the BSs and MTs (BSONRN mode), the situation is a little more complex than the BSON mode as more hops are involved during resource transfer, although RN does not have the traffic admission functionality. The heuristic algorithm first proposes to place RNs in positions within the coverage of several BSs, such as the shaded area presented in the Scenario I where the RN is located within the coverage of both BS1 and BS2. The RN can assist advanced communication mechanisms such as cell breathing traffic balancing and TDMA based soft handover. For example, suppose that in Scenario I the RN is associated with both BS1 and BS2. If BS1 reduces signal coverage radius to improve interference and capacity in considered area, the RN will lose its association to BS1, and it can transmit data to BS2. The BS2 may at this stage reduce or increase its coverage radius based on the information and traffic condition. The software handover support is similar to load balancing since the RN also acts as data relayer. Suppose a MT is moving from BS1 to BS2 and it currently receives data from the RN. If the received signal from BS2 becomes larger than it from BS1, the MT performs an intra cellular network handover from one cell to another cell without changing the serving RN. Scenario II places together two RNs at the coverage edge of BS1 and BS2 to facilitate the communication between two cells (It is assumed each cell has only one BS). From the SIR values combining with Shannon's formula, if the bandwidth allocation λB ratio to BS transmissions and the bandwidth allocation ratio to RN transmissions λR together is 1, Scenario II can have a slightly better system capacity performance compared to Scenario I because the average received signal power strength in Scenario I is lower than in Scenario II. However, the sites deployment of Scenario II can introduce much larger latencies, service interruption time and equipment cost, while most cellular coverage overlaps. To conclude, the shaded area in Scenario I is considered as a better site candidate plan other than Scenario I. Scenario III is also considered as an ideal location candidate. The RN at this position extends the cellular coverage to places without fixed infrastructure support.

As the attractor point mobility model is used to model MT movement, the MTs at some stage converge to the attraction points, dwell for a certain period of time and move within a short radius. In order to mitigate resource contention in the "hot spot", reduce service interruption time and latencies for ad hoc communications, a RN should be positioned where more MTs can associate to the RN within one or two hop distance as illustrated in Scenario IV. It is also important to place RN in places based on MT traffic density prediction. Scenario V presents the RN candidate sites combination topologies, it is preferred that RNs can compose a mesh like service layer with virtual backbone connections using ad hoc frequencies thus Scenario V topology right is recommended other than Scenario V topology left. The next step of the heuristic algorithm is to decide the number of RNs needed to assist traffic relaying with guaranteed QoS. The strategy first updates system traffic load information and initialises the HWN* system. MTs then continue moving using attraction point mobility model. As BSs and attraction points sites are pre-defined by the system so only MTs trajectories are needed to be used for Scenario IV analysis. A system performance result for a RN site combination is recorded and compared with theoretical analysis. The RN's location will be changed at disadvantaged locations if the result is not satisfactory. The algorithm runs recursively until an optimal solution is found. Meanwhile, in the final sites selection, a hard distant limit δ is introduced to solve BS RN overlapping problem to regulate the distance between one RN candidate site and any BS.

HWN* Routing Management

Efforts must be invested on HWN* to avoid congestion, compute the next route, discover medium resources and gather data. This section proposes a novel cascaded Adaptive Distributed Cross-layer Routing (ADCR) scheme for the HWN* framework, using a minimal number of hops and considers dynamic routing models aimed at reducing latency, preserving communications and delivering good overall and per node throughput. A cross-layer network design that seeks to enhance the system performance by jointly designing MAC and network layers has been adopted.

The cascaded ADCR includes three sub packet transmission modes labeled as One-Hop Ad-Hoc Transmission (OHAHT) for point-to-point direct communication, Multi-Hop Combined Transmission (MHCT) for radio resource relaying using fixed RNs or MTs, and Cellular Transmission (CT) for a traditional cellular service. The RNs of the core network compose a mesh structure with fixed routing tables using ad hoc frequencies, while BSs are connected to each other via wires and the link between the RN and BS are established using cellular channels and directional antennae. In areas without infrastructure support, two MTs may communicate directly, or through intermediate MTs. When a MT transmits packets to a BS through RNs, the RNs extend the signalling coverage of the BSON thus enhanced resource sharing performance is expected. The QoS flows can consume all the bandwidth on certain links, thus creating congestion for, or even starvation of best effort sessions. Statically partitioning the link resources can result in low network throughput if the traffic mix changes over time. Thus, a mechanism that dynamically distributes link resources across traffic classes based on the current load conditions in each traffic class is critical for performance. By proposing the cascaded ADCR for HWN*, the framework discourages applications from using any route that is heavily loaded with low priority traffic. Traditional routing strategies that use global state information are not considered. Problems associated with maintaining global state information and the staleness of such information is avoided by having individual MTs infer the network state based on route discovery statistics collected locally, and perform traffic routing using this localised view of the network QoS state. Each application, categorised by the service class with the choice of three possible transmission modes, maintains a set of candidate paths to each possible destination and route flows along these paths. The selection of the candidate paths is a key issue in localised routing and has a considerable impact on how the ADCR performs. The high priority traffic is given high priority in accessing comparatively expensive cellular resources, while low priority traffic tries to access lower cost ad hoc resources. Per MT bandwidth is used as the metric for local route statistics collection since it is one of the most important metrics in QoS routing. Furthermore, important metrics such as end-to-end delay can be improved by an increased bandwidth as long as the traffic load is not largely increased. As in QoS based investigation for inter system handover management. Traffic sessions are divided by HPUs, NPUs and LPUs. In case of network congestion, CT mode may temporarily become unavailable to NPUs when HPUs are not fully accommodated, while LPUs sessions may be only granted MHCT and OHAHT mode access to mitigate network congestion, reduce transmission delay and improve per MT throughput.

To avoid having higher traffic classes being influenced by lower traffic classes in terms of queueing delay, a waiting time limitation is placed on each traffic class using a forced starvation packet switch model. A traffic flow maintains two queues: a slot queue and a packet queue. The slot queue is decoupled for traffic class identification from the packet queue for transmission. The RN is specially designed so that it reserves QoS guaranteed free channels for signalling information exchange in the slot queue. Each MT and RN maintains a routing table. A RN's routing table contains the information of other fixed RNs thus the routing delay and multi-hop signalling overhead in MANET RN mode are largely reduced. On the other hand, a RN can also analyse the current system traffic load condition through feed-back from other fixed RNs. The purpose of bandwidth reservation is to let RNs that receive the relaying discovery command in the slot queue check if they can provide the bandwidth required for the connection. Start and finish tags are associated with slots but not with packets. When a packet arrives for a flow, it gets added to the packet queue, and a new slot is added to the slot queue. Corresponding start and finish tags are assigned to the new slot. The way to raise priority in slot queue is that the packets related to a high profile have shorter back-offtime to increase the probability of early medium access. As for the status table maintenance, information flooding is restricted to a limited scope. Once a positive acknowledgment message is confirmed by a requesting RN, the relay paths will not be changed unless resource contention happens. Given the fact that maintaining global RNs channel status in each RN slows down RN response time, each RN only updates neighbouring RNs' information, periodically.

In OHAHT, the requesting MT A first broadcasts SEARCH messages to every node in its transmission range including its associated RN and BS. If the destination MT B is within its transmission range and there is no ad hoc based media contention between MT A and MT B, MT B can respond to MT A with an ACK message. Once MT A confirms the acknowledgment, it starts a connection SETUP session immediately. The OHAHT transmission model can be extended to multi-hop ad hoc communication and it is only activated on demand.

The MHCT involves RNs acting as intermediate nodes for message relaying. BS provides assistance on cellular link establishment using cellular location registers. In the connection setup process for communication between MT A and MT B via the RN infrastructure, MT A first broadcasts SEARCH messages to every node to find MT B. After the SEARCH session, MT A may find that the cellular resources can be used through RNs by receiving three ACK messages from the serving BS of MT B, RNs and the MT B. It may also find that the ad hoc frequency based RN mesh is adequate for communications by receiving two ACK messages from the serving RN of MT B, and the MT B, respectively. These positive acknowledgments require MT B to send an ACK to its serving BS and serving RN, then the serving BS and the serving RN feedback the ACKs to MT A. Once the positive ACK is confirmed, MT A can either start a connection SETUP from MT A → RN, then RN → BS, and finally BS → MT B, or from MT A → RN then RN → M T B. The DATA transmission process follows the same packet delivery route, and further route discovery is prohibited to reduce the signalling overhead.

The label routing concept originated in Asynchronous Transfer Mode (ATM) network is introduced to MHCT mode since the position of RN is fixed and label based RN switching provides faster packet forwarding than routing because its operation is relatively simple.

The label is a fixed-length identifier. Multiple labels can identify a path or connection from source MT to destination MT. The structure of label has four parts. The first part CAST of the label is 3 bits and relates to cast options. Only uni-cast is considered in the current research. As practically the RN mesh can be very large, the second part of the Label is 20 bits which includes a node identifier and is unique in the network ensuring conflicts do not occur. The third part is 3 bits QoS, which means class of service. The last part of the label is TTL. All label information has a time-related restriction. After time out for a label, all corresponding entries will be deleted from the label routing table. Label routing uses a label to directly index into a connection table entry to determine the next hop, lending itself to a simpler lookup implementation than the complex IP routing and hop-by-hop IP address lookup. All intermediate nodes in the virtual connection can forward packets more efficiently. The path from a source MT to a destination MT is identified by multiple labels. The protocol distributes labels and set up new route after the path is computed by the routing protocol. The path finding process dynamically initialised by the label request packet carrying a unique label and flow information, where low path setup delay is guaranteed. The path between MT and destination MT is composed of multiple segments. The path is separated by segments and all data packets are relayed by these segments. Each segment is a real connection between two nodes and labeled by the sending-side node of label relay message in this segment. Figure 1 presents an example routing presentation and routing tables for three nodes MT A, RN B and MT C. The relay nodes only need to find the available entry indexed by a label in the packet, swap it with the respective Label out of this entry, and then send it out to the next relay node. Furthermore, the fixed RN placement reduces the frequency of label entry changes with reliable service.

 

Figure 1: Presentation of routing presentation and routing tables for three nodes MT A, RN B and MT C

In the CT mode, the source node MT A first broadcasts SEARCH messages to every node to find destination MT B. After the SEARCH session, the MT A finds that it is able to communicate with MT B directly via BSs, while the connection can be setup through a virtual wireless backbone. The positive acknowledgment of a connection requires MT B to send an ACK to its serving BS, then the serving BS informs the serving BS of MTA or the BS feedbacks the ACK to MT B when both MT A and MT B share the same serving BS. Once the positive ACK is confirmed, MT A starts connection SETUP from MT A → BS, then BS → BS, and finally BS → M T B. The DATA transmission process follows the same packet switched delivery route. The cellular transmission also includes cellular transmission with RN support. For cellular based RN connection, the packet tries to establish a link with a BS first then the traffic will be connected to BS RN mode. Dynamic channel allocation can be realised in a distributed manner given that the channel usage does not break the two channel interference constraint which are cosite constraint where there are minimum channel separations within a cell and non-cosite constraint where minimum channel separation between two adjacent BSs is kept.

HWN* MANAGEMENT FRAMEWORK AND ALGORITHMS

HWN* Handover Management

In the HWN* context, each communication mode exhibits different capacity characteristics therefore selecting the most appropriate mode for a particular service request is critical to guarantee QoS to the end user. A balanced traffic load distribution between the available modes is also important to avoid one mode becoming excessively loaded, which leads to an unstable state. The HWN* requires an algorithm that includes the service transfer function between the cellular communication modes and MANET communication modes. Another novelty of the HWN* handover mobility management is that user segregation is provided to guarantee QoS to particular user classes by the prioritisation of user requests when the HWN* system approaches a congested state. As the RN provides stable infrastructure support for a MANET, it will be beneficial to switch more service from the cellular service to the MANET service. The HWN* concentrates on the mode selection considering differentiated user negotiation, multiple handover triggers and congestion control. Here the handover mobility management procedure  refers to a process of transferring a MT from its serving BS to another BS or from one medium access interface to another interface. The RN using cellular frequencies relays the handover traffic but doe s not have traffic admission functionality. While for the MANET mode and MANET with RN mode, the procedures refer to a new path finding process similar to rerouting, which is described as part of the cascaded ADCR routing algorithm. To mutually mitigate the traffic burden between communication modes, the HWN* dynamically allocates the radio resources without a fixed plan.

Three service classes are devised to describe subscriber behaviour: High Profile Users (HPUs), Normal Profile Users (NPUs) and Low Profile Users (LPUs). Principally, HPUs get the guaranteed QoS service and the class is assured with any amount of bandwidth, firm end-to-end delay bounds and limited queue loss for data flow. Next are NPUs with less QoS guaranteed medium access opportunities compared to HPUs, but the users will get a close service quality as the one received by HPUs. LPUs are a best-effort class with absence of QoS specification using currently available medium resources. HPUs have the highest access priority in any of the communication modes of the HWN*, and traffic admission of NPUs and LPUs has to consider the impact of ongoing HPUs sessions. NPUs are also configured to have a higher probability than LPUs in terms of resource acquisition and this probability is decided by an Association Level (AL) set, which will be described later. Inter-system and intra-system mobility management are addressed separately to reduce system complexity.

MANET RN and MANET modes can potentially become the primary communications methods if distributed management, link reliability and security problems of the MANET mode can be solved. Specifically for the MANET mode, the issue is that data relaying via a third party MT is not safe as discussed in MCN review. Data relaying via provider owned fixed RN provides a permanent solution. The RN integration largely moderates the problem by providing reliable relays between communication parities. Upon communication request, HPUs are configured to search for cellular based service, NPUs sfor a MANET RN service while LPUs for MANET service by default. The signalling, path discovery and route establishment for MANET and MANET RN modes are completed before the BSON and BSON RN modes, if all four modes are available. Inter-system and intra-system traffic handovers are only triggered when essential to avoid unnecessary network management overhead.

The control entity for inter system mobility cooperation is called the HWN* Mobility Controller, which is responsible for managing the modification of a route in an attempt to maintain or enhance the QoS level. The unit is located in each BS and a BS periodically monitors receives, specifies, filters and analyses frequencies in use by nearby MTs, RNs and BSs, geographically locates MTs so that the terrain blocks interfering signals, and may use directional antennas to reduce unwanted signals. The BS also gathers the information on the MANET with RN mode channel availability from its associated RN. A MT, either requiring differentiated service or not, makes a distributed decision on inter-network and intra-network handovers based on information gathered from its associated BS mobility controller as well as information gathered after probing one hop point to point direct communication. To obtain an effective handover process, while reducing the unnecessary handover rate, it is proposed to setup a dedicated Status Check Point embedded in the HWN* Mobility Controller where the necessary measurements are taken and then fed back to the handover algorithms of the nodes involved. The check result indicates the likelihood of a handover, which depends on interference level and physical layer information such as Bit Error Rates (BER), velocity, buffer size, etc. Since the HPUs applications have higher priorities over NPUs and LPUs, subscribers from this profile are more likely to get an accept ticket, which is issued by the Negotiation Unit in continuing with traffic handover. If an accept ticket is not issued, the MT will not use the status check data to request continuing with the handover process unless the status check point data necessitates handover. Then the mobility controller will decide to accept, queue or reject the MT handover request. The negotiation process between HPUs, NPUs and LPUs is based on Association Level (AL), which makes decisions and feeds back to the HWN* Mobility Controller. The AL is a set of parameters monitoring channel availabilities, an AL that scores higher than the threshold means that the channels are already occupied by ongoing sessions. The AL set is further sub-classified as AL in the BSON, BSON RN, MANET and MANET.

The handover algorithms in the HWN* should allow subscribers to seamlessly move without dropping their communication session and considers differentiated QoS issues, for example, it must guarantee QoS for HPUs that agree to pay more than NPUs and LPUs. Two handover types are therefore used as described previously, Intra Network Handover occurs when a MT enters into another entity that belongs to the same network with a cellular TDMA MAC interface. An Inter Network Handover happens when a MT leaves the serving network and communicates with another entity that belongs to a different network. For all service classes, the intra network handover is selected before considering inter network handover as less traffic management overhead may be produced. In intra network handover, after obtaining pass tickets from the Status Check Pointand the Negotiation Unit, the Network Selector entity that embedded in the mobility controller informs the MT if the RN should participate in handovers or not, then the MT makes a local decision. The inter network handover is seen as a switching process between ad hoc and cellular services which has different properties and procedures compared to the intra network handover. The Status Check Point is activated first to avoid extra expense and the Negotiation Unit keeps monitoring the channel availability status and updating the AL in a short interval to grant or reject handovers. The Network Selector always tries to divert the traffic back to intra network handover before it is informed that the intra network handover is not possible. If the MT is currently communicating in ad hoc modes, the selector will search for available B Ss in neighbouring cells. If the MT is using cellular modes, it will look for either direct point-to-point communication or search for a fixed RNs involved multi-hop communications. The Network Selector also uses several short network search expire time s for both intra-network search and inter-network search to make sure a MT is not long isolated during the network selection process