Enhanced Radio Resource Management Algorithms for Efficient MBMS Service Provision in UTRAN

Christophoros Christophorou1, Andreas Pitsillides1, Tomas Lundborg2

1 University of Cyprus, Department of Computer Science, 75 Kallipoleos Street, P.O. Box 20537, CY-1678 Nicosia, Cyprus

christophoros@cs.ucy.ac.cy, andreas.pitsillides@cs.ucy.ac.cy 2 Ericsson AB

Torshamnsgatan 23, 164 80 Kista, Sweden tomas.lundborg@ericsson.com

Corresponding Author:

Name: Christophoros Christophorou Email: christophoros@cs.ucy.ac.cy Tel. Number: +357 22892687 Fax Number: +357 22892701

Address: 75 Kallipoleos Street, P.O. Box 20537, CY-1678 Nicosia, Cyprus Keywords: MBMS, UMTS, RRM, Dual transmission mode cell, Handover Control Abstract – As currently specified by 3GPP, Multimedia Broadcast Multicast Service (MBMS) bearer services can be provided within a cell either by Point-to-Point (P-t-P) or Point-to-Multipoint (P-t-M) transmission mode, but not both at the same time. If P-t-P transmission mode is selected for a cell, one Dedicated Channel (DCH) is established for each user within the cell that joined the MBMS service. Otherwise, if P-t-M transmission mode is selected, one Forward Access Channel (FACH) is established covering the whole cell’s area and commonly shared by all the UEs within. In this paper, we highlight the inefficiencies that can be caused with the aforementioned approach and introduce the “Dual Transmission mode cell” in which P-t-P and P-t-M transmissions (i.e. multiple DCHs and FACH) are allowed to coexist within the same cell. Hence, we propose a new radio resource allocation algorithm and solution to address them. Our proposed algorithm considers the instantaneous distribution and movement of the users within the cell and dynamically decides which users will use FACH and which DCH, in such a way that the requested Quality of Service (QoS) is supported with the least amount of transmission power (i.e. capacity) consumption. Moreover, with the “Dual Transmission mode cell”, new types of intra-cell handovers are introduced which we also analyse and propose a new handover algorithm to address them. The performance evaluation carried out showed that our proposed “Dual Transmission mode cell” approach, provides considerable gains, as well as outperforming all other related approaches, such as “UE Counting”, “Power Counting”, “Rate Splitting”, and “FACH with dynamic power setting”, in terms of capacity and link performance efficiency.

1

Table of Acronyms

3GPP 3rd Generation Partnership Project AH Activation Hysteresis AR Alteration Rate CCIT Cell Context Information Table CN Core Network CPICH Common Pilot Channel DCH Dedicated Channel FACH Forward Access Channel FACH_IT FACH Information Table HAA Handover Activation Area HAT Handover Activation Threshold HI_Table Handover Information Table HSDPA High Speed Downlink packet Access HTT Handover Trigger Threshold IMSI International Mobile Subscriber Identity MBMS Multimedia Broadcast Multicast Service MCCH MBMS Point-to-Multipoint Control Channel MD Mobile Device NBAP Node-B Application Protocol PP Pre-Trigger Predictor P-t-M Point-to-Multipoint P-t-P Point-to-Point QoS Quality of Service RNC Radio Network Controller RRC Radio Resource Control RRM Radio Resource Management SM Safety Margin TMGI Temporary Mobile Group Identity TTI Transmission Time Interval UE User Equipment UMTS Universal Mobile Telecommunications System UTRAN UMTS Terrestrial Radio Access Network Δt Handover delay time

2

1 Introduction Driven by the need for efficient data distribution when a large number of users want to receive the same data, the 3rd Generation Partnership Project (3GPP) consortium introduced the Multimedia Broadcast Multicast Service (MBMS) system [1][2] in Universal Mobile Telecommunications System (UMTS) Release 6 specifications [3]. MBMS system is considered as one of the most important steps of the UMTS network evolution since it provides it with a powerful tool to offer broadcast and multicast services efficiently. With MBMS system (see Figure 1), the Core Network (CN) sends only one stream of data to the Radio Network Controller (RNC), irrespective of the number of Node-Bs (also known as Base Stations) or User Equipments (UEs) that want to receive it. Then, the RNC replicates and distributes, as efficiently as possible, the MBMS content to the UEs within the cells.

Figure 1 Current 3GPP specified MBMS service provision approach

As currently specified by 3GPP in [2], MBMS bearer services can be provided within a cell either by Point-to-Point (P-t-P) or Point-to-Multipoint (P-t-M) transmission mode, but not both at the same time. If P-t-P transmission mode is selected for a cell, one Dedicated Channel (DCH) is established for each UE within the cell that joined the MBMS service. Otherwise, if P-t-M transmission mode is selected, one Forward Access Channel (FACH) is established covering the whole cell’s area and commonly shared by all the UEs within. The decision of which transmission mode is going to be adopted is made by the RNC based on a radio resource efficiency criterion [2].

Although the service and technical requirements are complete for MBMS [1]-[6], still many design issues [7] need to be considered for MBMS services to be provided in a more efficient way, especially in the UMTS Terrestrial Radio Access Network (UTRAN) where the radio resources are limited. In this paper we focus our research on this issue. First, we highlight the inefficiencies that can be caused with the current 3GPP approach [2]. To address them and achieve efficient MBMS service provisioning, we motivate the need of allowing P-t-P (multiple DCHs) and P-t-M (one FACH) transmissions to coexist within the same cell (which we refer to as the “Dual Transmission mode cell”). Then, we built on this idea, which first appeared in a brief 3GPP report [8], and propose a new radio resource allocation algorithm to efficiently manage the radio resources of this new type of cell.

The radio resource allocation algorithm we propose, allows part of the cell’s area (see Figure 4) to be supported using FACH (“FACH supported area”) while the rest of it is supported using DCHs (“DCH supported area”). Both at session initiation and also during the session, the size of these areas is dynamically adapted (shrinks or expands by adapting the transmission power devoted to FACH and by releasing or establishing

3

DCH connections) according to the instantaneous distribution and movement of the users within the cell, aiming to always provide the MBMS service in the most efficient way (in terms of capacity and Quality of Service (QoS)). Moreover, with this new approach, mobility issues arise, in terms of intra-cell handovers between the FACH and the DCH supported areas, that must be considered in order for this scheme to work efficiently. Thus, we also analyse these new types of handovers and propose a new handover algorithm to efficiently address them.

The paper is organized as follows: In section 2 the need for the “Dual Transmission mode cell” is motivated by emphasizing the gains that can be achieved compared to the current 3GPP specified MBMS service provision approach [2]. In section 3 we present related research work performed on the efficient provision of the MBMS service. The “Dual Transmission mode cell” concept, the challenges we address in our design and the problem formulation are described in section 4 together with our solutions in terms of new radio resource allocation and intra-cell handover algorithms. The performance evaluation of our proposed “Dual Transmission mode cell” approach is presented in section 5, together with a comparative evaluation with other competing approaches, such as “UE Counting”, “Power Counting”, “Rate Splitting”, and “FACH with dynamic power setting”. Finally, in section 6 we provide our conclusions and future work.

2 Motivation for “Dual Transmission mode cell” In order to motivate the need for the “Dual Transmission mode cell” approach, the following scenario was investigated highlighting gains that can be achieved.

The scenario we use (see Figure 2) considers the case where 30 stationary users, participating in a 3 minute duration 64 Kbits/sec MBMS streaming video session, are uniformly distributed at distances closer than 600 meters from the Node-B and two UEs (UE 1 and UE 2) located near the cell’s edge, are leaving and re-joining the MBMS service every 20 seconds. The simulation parameters used are provided in TABLE I. Two instances of this scenario were simulated, one using the current 3GPP [2] and the other the “Dual Transmission mode cell” approach. Results concerning the capacity requirements and processing effort introduced (in the RNC and Node-B) by each approach were obtained using OPNET Modeller and illustrated in Figure 3 and TABLE II. As can be observed, the gains that can be achieved when the “Dual Transmission mode cell” approach is utilized are significant. The scenario is further analysed below.

TABLE I SIMULATION PARAMETERS Simulation Parameter Value

Propagation environment Pedestrian Outdoor Cell radius 1000 meters

Node-B Antenna type Omnidirectional Shadow fading standard deviation log-normal of 10 dB Transmission Time Interval (TTI) 20 ms (as suggested in [9])

Channel coding Convolutional 1/3 coding rateDownlink Other-cell Interference factor 1.78 Thermal Noise power spectral density -174 dBm

Orthogonality factor 0.5

4

(a) Current 3GPP approach: Use either P-t-M or P-t-P transmission mode

(b) “Dual Transmission mode cell” approach: Adapt FACH coverage to support the UEs near the Node-B and use DCHs for UEs near the cell’s edge

Figure 2 Need for co-existence and dynamic adaptation of P-t-P and P-t-M transmissions

(a) Actual Transmission power used (watts)

(b) Average Transmission power used (watts)

Figure 3 Capacity requirements – 3GPP Vs “Dual Transmission mode cell” approach

At the initialization of the MBMS session, the current 3GPP approach (Figure 2.a) establishes one FACH allocating 2.0222 watts to cover the whole cell’s coverage area and support all the UEs within. The “Dual Transmission mode cell” approach (Figure 2.b), establishes one FACH allocating 0.2620 watts to support the 30 UEs located at distances closer that 600 meters from the Node-B and uses DCHs to support UE 1 and UE 2 located near the cell’s edge (allocating ~0.9 watts for both DCHs). By doing this, the total transmission power required is reduced from 2.0222 to 1.16 watts (38% less power; see Figure 3.a).

Then, when UE 1 and UE 2 leave the service, the current 3GPP approach establishes one DCH for each of the remaining 30 UEs (allocating a total of ~1.53 watts) and releases the FACH. The “Dual Transmission mode cell” approach just releases the DCHs of UE 1 and UE 2 and retains FACH’s transmission power to 0.2620 watts to support the 30 UEs near the Node-B. By doing this, the total transmission power used is reduced from 1.53 to 0.2620 watts (83% less power; see Figure 3.a).

When UE 1 and UE 2 join the service again, the current 3GPP approach releases all the DCHs (30 in total) and establishes one FACH covering the whole cell’s area (allocating 2.0222 watts). The “Dual Transmission mode cell” approach retains the FACH with transmission power 0.2620 watts to support the 30 UEs near the Node-B and just establishes two more DCHs to support UE 1 and UE 2.

5

TABLE II AGGREGATED PROCESSING EFFORT INTRODUCED IN THE RNC AND NODE-B DUE TO ESTABLISHING OR RELEASING OF CHANNELS

Current 3GPP specified approach Dual Transmission mode cell approach Action Times performed Action Times performed

Establish FACH 5 Establish FACH 1 Release FACH 4 Release FACH 0 Establish DCH 120 Establish DCH 10 Release DCH 120 Release DCH 8

As illustrated above, the “Dual Transmission mode cell” compared to the current 3GPP approach can achieve significant transmission power savings. As shown in Figure 3.b, the average transmission power used during the session is reduced from 1.8094 to 0.7583 watts (58% less). Also, the processing effort burdening the RNC and the Node-B during the session (see TABLE II), for the establishment or releasing of channels, is also considerably reduced (this is further analysed in [19]).

The potential gains that can be achieved with the “Dual Transmission mode cell” approach are highlighted above. As indicated, a vital aspect for the efficient MBMS service provisioning lies in the co-existence and dynamic adaptation of P-t-P and P-t-M transmissions, a feature not considered in the current 3GPP approach [2].

3 Related Work As mentioned in [10], with MBMS services more than 30% of the Node-B’s total power has to be allocated to a single 64 Kbits/sec MBMS service. This can make MBMS services too expensive since the overall UMTS network capacity is limited by the Node-B’s power source. Thus, in order to make MBMS services feasible and also achieve higher bit rates, some power saving schemes has to be introduced. Below we briefly discuss related Radio Resource Management (RRM) schemes.

Initially, as described by 3GPP and standardized in [2], the criteria used by the RNC for selecting either P-t-P or P-t-M transmission mode to provide the MBMS service within the cell, was based on a “UE Counting” threshold. This threshold is estimated based on the worst case scenario, i.e. all the UEs are assumed to be at the cell’s edge. Usually, this number cannot exceed 5 or 6 UEs. With this approach, the RNC periodically counts the UEs within the cell who joined the same MBMS service. If the number of the counted UEs is lower than the “UE Counting” threshold then the P-t-P transmission mode is adopted, otherwise, the P-t-M transmission mode is adopted.

The “UE Counting” is not the best possible approach because the current location of the users is not taken into account. Thus, later a more efficient approach [11][12], using the downlink transmission power required as the selection criterion, was adopted. With this approach, the RNC periodically notifies the UEs who joined the service to report to it the pathloss they experience in the cell. Then, the total downlink transmission power required by each transmission mode is estimated and the most efficient is adopted. Yet, with this approach dynamic adaptation of FACH’s transmission power is not allowed. This can result in redundant use of power in cases when all the users are located near the Node-B and P-t-M transmission mode is used.

With Rate Splitting [13][14], when P-t-M transmission mode is used, the MBMS data stream is split into several streams with different QoS. The lower layer is coded by itself to provide the basic video quality and broadcasted in the whole cell’s coverage area. The enhancement layer is coded to enhance the lower layer and broadcasted with less amount of power and only the users who have better channel conditions can

6

receive it. The enhancement layer when added back to the lower layer regenerates a higher quality reproduction of the input video. With this way, FACH’s transmission power requirement is reduced. However, the tradeoff is a worsening QoS for UEs receiving only the lower layer, which is neither desired nor fair.

FACH with dynamic power setting [15][16] is another approach that can be used for delivering MBMS services using FACH. With this approach, the transmission power devoted to FACH is determined based on the worst users’ pathloss within the cell. With this way, FACH’s transmission power requirement is reduced. However, the tradeoff is the need for very frequent reporting by the UEs in order for the RNC to keep track of all the users and adjust the FACH’s transmission power accordingly.

All the aforementioned approaches are based on the current 3GPP approach [2], which allows only one transmission mode to be utilized within a cell at any given time. The idea of the “mixed usage of multiple DCHs channels and FACH” first appeared in a brief 3GPP report [8] and later suggested in [16] and [17] as an appealing method for providing MBMS services. With this approach, some users lying within the inner part of the cell can be served using FACH while the rest of the users in the outer part can be served using DCHs. To the best of our knowledge, we were the first that proposed a radio resource allocation algorithm specifically implementing this concept (see [18] and [19]). In this paper we build on this concept and propose a new radio resource allocation algorithm that facilitates both at session initiation and also during the session an efficient channel combination decision. Moreover, this new concept introduces new intra-cell handovers that we also analyse and propose a new handover algorithm to efficiently address them. Later, a similar approach (building on our work [18]) focusing on the High Speed Downlink Packet Access (HSDPA), was proposed in [20].

4 Dual Transmission mode cell concept

With the “Dual Transmission mode cell” concept (see Figure 4) the group of users that join the same MBMS service is divided into two subgroups; the one subgroup is served using FACH and the other using DCHs. Based on these subgroups, the areas of the cell that will be supported using FACH (“FACH supported area”) and DCHs (“DCH supported area”) are decided and the appropriate channels are established and assigned to the users. Periodically, during the session, the subgroups (and thus the FACH and DCH supported areas) are dynamically adapted to always provide the MBMS service in the most efficient way.

Figure 4 “Dual transmission mode cell” concept

4.1 Main challenges of the proposed design Before describing our proposed approach implementing the “Dual Transmission mode cell” concept, we briefly discuss the characteristics of the FACH and DCH channels.

7

This will also assist in clarifying the main challenges of our design and the approach we adopt to address them in order to achieve the performance goal.

The FACH can be shared by several users without requiring additional radio resources (i.e. FACH offers capacity benefits). However, fast power control is not applied on FACH. Thus, a fixed amount of transmission power has to be allocated to this channel, typically at a rather high power level, such that the requested QoS is guaranteed right up to the coverage limit of the selected coverage area. Since FACH’s transmission power is fixed, its coverage range is bounded. Thus, if a user leaves its coverage area will experience QoS degradation.

The DCH is dedicated only to one user and allows the use of fast power control. Fast power control is a crucial aspect in UMTS because it improves link performance and enhances downlink capacity, by reducing the average required transmission power and interference to other users. The coverage of the DCH is not bounded but it can be increased at the expense of more power; the greater the distance of the user from the Node-B, the more the (exponential) increase in the transmission power required.

Based on the aforementioned, the main challenges in our design are:

• To form and modify the DCH and FACH subgroups in such a way that the benefits that FACH (capacity benefits) and DCH (fast power control) have, are exploited. The aim is to achieve minimum transmission power consumption within the required coverage area, during the MBMS service provisioning.

• During the MBMS service provisioning, to consider the bounded coverage range of FACH and the transmission power requirements of both FACH and DCH, in order to avoid any QoS degradation or any transmission power waste.

To address these challenges, we propose two algorithms that run in parallel in order for the performance goal to be achieved. The first algorithm runs in the RNC and is responsible for the efficient formation and dynamic modification of the DCH and FACH subgroups, as well as the management of the cell’s radio resources. However, this algorithm cannot run continuously as this will cause excessive terminal’s battery consumption and uplink noise, due to the reporting required by the UEs. Thus, this algorithm needs to run periodically during the session. However, during idle periods any mobility of the users will not be accommodated, which may result in:

• QoS degradation, if a user during this idle period moves from the FACH to the DCH supported area. This will occur since once outside of the “FACH supported area”, the signal strength of FACH is inadequate for the UE to decode the signal correctly.

• Transmission power waste, if a user during this idle period moves from the DCH to the FACH supported area. This will occur since the user will continue receiving the service using DCH in an area where the service is also reliably supported using the FACH already established in the cell.

Hence, the role of the second algorithm is to complement the first one during these idle periods of time. This algorithm runs in the UEs and is responsible for monitoring and efficiently executing an intra-cell handover between the FACH and DCH supported areas and therefore avoiding the aforesaid inefficiencies from occurring.

Below we formulate our problem and then propose a practical algorithm to solve it. The proposed radio resource allocation algorithm is described in detail in section 4.3,

8

while the intra-cell handover algorithm, which addresses the mobility issues, is analysed and described in section 4.4.

4.2 Problem formulation Assume a cell C. Within the cell there exist N users whose locations remain static for the time of interest. Area A is a subarea of the cell which will be supported by using FACH (see Figure 4). Since fast power control cannot be applied on FACH, the different channel impairments1 that degrade the signal quality during propagation cannot be efficiently compensated, resulting in non-homogeneous dispersion of the signal quality along the perimeter of the selected area A, with also as a result its coverage range bounded. Hence outside of area A the requested QoS cannot be reliably supported. Due to the non-homogeneous dispersion of FACH’s signal quality within the cell, the actual location of the user cannot be considered as a reliable criterion to indicate if the user lies within the area A. Thus, the criterion we adopt to indicate if a user lies within the area A, is the Common Pilot Channel (CPICH) Ec/No signal quality he receives. Hence, in order to ensure that we provide adequate QoS support to those users that will be served by FACH, we assume that a user ui,

, lies within the area A if ui_CPICH ≥ Q dB. In this way, the area A determines the m number of users served by FACH and satisfying their QoS, while the remaining N-m number of users are served by DCH, with their QoS also satisfied. So m is a function of the area A such that

}...3,2,1{ Ni∈

()f )(Afm = .

Our objective is to determine the area that will be supported using FACH, such that it minimises the total downlink power consumption. This optimal area will determine the m number of users that will utilize FACH. Note that the total downlink power comprises the fixed component of the power required by the FACH to support the m number of users within area A, plus the total power required for the N-m number of users using DCHs. Let denote the total downlink power required for a given number N of users, for a given distribution x of the users within the cell and a given FACH subarea of the cell A. The objective can thus be expressed mathematically as follows:

*A*A

),,( AxNPDL

),,(minarg* AxNPA DLA= (Problem 1)

The infinite space of the area A, makes the above problem extremely difficult to solve. To render the problem tractable, we discretize area A by introducing the idea of the zone areas. We divide the cell into a number J of zones Zj, }...3,2,1{ Jj∈ such that

. A user ui lies within the Zj if ui_CPICH ≥ Qj dB. For each zone Zj we

calculate the total downlink power consumption and we consider a relaxation of Problem 1 given by:

CZ jJj=∪

∈

)jZ,,(DL xNP

),,(minarg**, jDLJjZ

ZxNPAj ∈

=

In this way, the selected Zj determines the m number of users served by FACH, while the remaining N-m number of users, are served by DCH.

1 The channel impairment can be due to for example the distance dependent path loss, the location dependent shadowing, multi-path fading dependent on the speed and environment of the mobile, and the interference level dependent on cell position and neighbouring cell activity.

9

4.3 Proposed Radio Resource Allocation Algorithm Motivated by the problem formulation presented above, we develop a practical approach (algorithm) to solve the problem, also considering the users’ mobility. The proposed radio resource allocation algorithm aims to efficiently form (at session initiation) and modify (during the session) the subgroups of users that will be served using DCH and FACH, in such a way that the requested QoS is always supported with the least amount of transmission power consumption. However, due to the users’ mobility, this procedure is not easy, especially with QoS inefficiencies likely to occur for those users located near the “FACH supported area” coverage limit. Thus, before the subgroups are formed, it is vital for the algorithm to be informed not only about the way the users are distributed, but also it is important to know how the users are moving within the cell. Using this knowledge, the algorithm forms different possible combinations of group of users that should be served using FACH and DCH (i.e. transmission arrangements; see section 4.3.2.2), which can be used for the MBMS service provisioning. Then, the transmission power required for each is estimated and the most efficient among these combinations of groups is selected and adopted.

Based on the aforementioned, we divide our proposed algorithm into two phases:

• Initialization phase: The transmission arrangement that will be used in the cell at the initialization of the session is selected, the size of the FACH and the DCH supported areas is decided, the required radio resources are allocated and the appropriate channels (FACH and DCHs) are established.

• MBMS Session Ongoing phase: Periodically, during the session, the transmission arrangement used is dynamically modified to accommodate any changes on the users’ distribution and movement within the cell and provide the MBMS content as efficiently as possible.

Each phase described above will be executed in three steps:

• Information Collection step: All the necessary context information is collected, processed and stored locally in the RNC.

• Transmission arrangements estimation step: All possible transmission arrangements that can guarantee the requested QoS for all the users are formed and the total amount of transmission power required for each is estimated.

• Final Decision step: The most efficient transmission arrangement that will be used for the MBMS service provision within the cell is selected and adopted.

Below we provide the details of each phase of the algorithm. To simplify the description, we assume a one-cell scenario. In case of a multi-cell scenario, the actions described below will occur in each cell the MBMS service is made available.

4.3.1 Initialization phase The steps executed during the Initialization phase are described below.

4.3.1.1 Information Collection Step Just before the MBMS session start, the RNC creates a “Cell Context Information Table (CCIT)” (see TABLE III) in which the instantaneous context of the cell will be stored. Then, the RNC notifies the users within the cell who joined the MBMS service, to report to it their instantaneous context. This report must include:

10

• Instantaneous CPICH Ec/No signal quality received (dB) • Instantaneous CPICH Ec/No alteration rate experienced (dB/sec)

The CPICH Ec/No alteration rate is estimated by the UE, by considering the instantaneous and previous CPICH Ec/No measurements. This parameter indicates how fast the channel quality degrades or improves during the UE’s mobility. To avoid any inefficiency due to measurement errors, it is important to apply filtering on the CPICH Ec/No measurements to average out the effect of fast fading. Often a filtering period of 200 ms is chosen, which we adopt for the studies in this paper. Note that the aforesaid values will be also used by the UE for handover purposes (see section 4.4).

TABLE III CELL CONTEXT INFORMATION TABLE (CCIT) Cell ID (Uniquely identifying the Cell that this CCIT belongs to)

MBMS TMGI (Uniquely identifying the MBMS Service that this table is created for) MBMS Service QoS Requirements

QoS Class Bit Rate Delay Loss QoS ID (The MBMS QoS Requirements are mapped into a QoS ID)

Zone Area supported using FACH

UE_1 IMSI CPICH Ec/No Signal Quality Received

CPICH Ec/No Alteration Rate Experienced

Power required in case a DCH is used

Type of Connection

UE_2 IMSI CPICH Ec/No Signal Quality Received

CPICH Ec/No Alteration Rate Experienced

Power required in case a DCH is used

Type of Connection

One record for each user that

joined the MBMS Service and lies within

the cell ……………………………………….

For each user within the cell that has joined the MBMS service, the RNC creates one record in the CCIT in which the reported context information is stored. This makes the algorithm aware of the distribution and movement of the users within the cell. The CCIT is depicted in TABLE III, while the fields are described below:

• Cell ID: Uniquely identifies the cell for which the respective CCIT is created.

• MBMS Temporary Mobile Group Identity (TMGI): Uniquely identifies the MBMS service that this CCIT is created for.

• MBMS Service QoS Requirements: QoS class (i.e. conventional, streaming, interactive, background), bit rate, delay and loss requirements.

• QoS ID: The MBMS service’s QoS requirements are mapped into a QoS ID. This QoS ID is provided to the UE at the initiation of the MBMS service and assists it to identify the thresholds that will be used in case an intra-cell handover is going to occur (see section 4.4).

• Zone Area supported using FACH: The Zone Area (i.e. the part of the cell) that will be supported using FACH is indicated here. This info is broadcast in the cell and used by the UE for intra-cell handover purposes (see section 4.4).

• One record for every user that lies within the cell and has joined the MBMS service: Each record includes the following:

o UE’s International Mobile Subscriber Identity (IMSI): Uniquely identifying the user.

o CPICH Ec/No Signal Quality received: Depicts the MBMS users’ distribution within the cell (i.e. “channel quality” based distribution).

o CPICH Ec/No Alteration Rate experienced: Depicts the MBMS users’ movement within the cell.

11

o Power required in case a DCH is used: Estimated by the RNC based on the CPICH Ec/No signal quality received by the user.

o Type of Connection: “DCH” if the user will be supported using DCH; “FACH” if the user will be supported using FACH.

4.3.1.2 Transmission Arrangements Estimation Step The algorithm, by considering the cell’s CCIT, becomes aware of the instantaneous context of the cell. However, in order to form all possible transmission arrangements and estimate the transmission power required for each, it also needs to know the transmission power that should be devoted to FACH for guaranteeing a reliable reception of the MBMS service at different distances from the cell’s Node-B (referred to as different “Zone Areas”). This info will be provided to the RNC, in a form similar to that shown in TABLE IV (referred to as “FACH Information Table (FACH_IT)”).

TABLE IV FACH INFORMATION TABLE (FACH_IT) – 64 KBPS MBMS STREAMING VIDEO Cell ID (Identifying the Cell this FACH_IT belongs to)

QoS ID (Associating the FACH_IT with the MBMS service’s QoS requirements) Zone Area Zone Area coverage limit

CPICH Ec/No required to be received FACH transmission

power required Z1 (~100m coverage) ≥ 35dB 0.0002 watts Z2 (~200m coverage) ≥ 21.5dB 0.0032 watts Z3 (~300m coverage) ≥ 14dB 0.0163 watts Z4 (~400m coverage) ≥ 9dB 0.0517 watts Z5 (~500m coverage) ≥ 5dB 0.1263 watts Z6 (~600m coverage) ≥ 1.8dB 0.2620 watts Z7 (~700m coverage) ≥ - 0.9dB 0.4855 watts Z8 (~800m coverage) ≥ - 3.3dB 0.8283 watts Z9 (~900m coverage) ≥ - 5.4dB 1.3267 watts

Z10 (full cell coverage) ≥ - 7.2dB 2.0222 watts

Cell divided in ten Zone Areas (Z1–Z10)

In the FACH_IT shown above, created for a 64 Kbits/sec streaming video, we divided a cell with radius 1000 meters into ten Zones Areas (Z1 – Z10). The coverage limit of each Zone Area is defined based on a minimum CPICH Ec/No value required to be received by the UE for guaranteeing a reliable reception of the MBMS content using FACH. Using the example given in TABLE IV, if the user receives the CPICH Ec/No with greater than or equal to (≥) 35dB then it belongs to Zone Area 1 (Z1); if the user receives the CPICH Ec/No with greater than or equal to (≥) 21.5 dB then it belongs to Zone Area 2 (Z2), etc. Note that the users that belong to Z1 also belong to Z2.

It is worth mentioning that the number of the zone areas the cell is divided in is highly affected by the cell’s environment, the required capacity gains and the desired QoS. For example, dividing a high vehicular (i.e. speeds of 120 Km/h) cell in ten zone areas will not be efficient since frequent intra-cell handovers will occur, possibly degrading also the QoS experienced by the users. However, the more the zone areas the cell is divided in, the higher the capacity gains that can be achieved.

Using TABLE IV as an example, eleven possible Transmission Arrangements (TAs) can be formed for the provision of the MBMS service within the cell:

• TA 1: All the users will be supported using DCHs (no FACH used).

• TA 2: Support using FACH the users that belong to Z1 (i.e. those that receive a CPICH Ec/No ≥ 35 dB) and based on their mobility are predicted to stay within Z1 for the next C seconds. All other users will be supported using DCHs. The value of time C is fixed and pre-estimated and equals the time required by the

12

RNC to switch a user from the FACH to the DCH supported area without any QoS degradation, plus a small amount of time the UE must stay connected with FACH before triggering a handover (in order to avoid frequent intra-cell handovers). Thus, the users that will be supported using FACH are those whose:

CPICH Ec/No received ≥ 35dB + (CPICH Ec/No Alteration Rate x C)

We will refer to the CPICH Ec/No Alteration Rate x C as the “Mobility Predictor” parameter. Its role is to predict the mobility of the users during the transmission arrangements’ formation and eliminate the possibility for QoS degradation, or frequent intra-cell handovers. This is achieved by ‘guaranteeing’ that all the users that will be supported using FACH will stay within the selected “FACH supported area” for at least C more seconds.

• TA 3: Support using FACH the users that belong to Z2 (i.e. those that receive a CPICH Ec/No ≥ 21.5 dB) and based on their mobility are predicted to stay within Z2 for the next C seconds. All other users will be supported using DCHs.

• TA 4 to TA 10: Can be defined similarly to TA 2 and TA 3.

• TA 11: All the users will be supported using FACH (no DCHs used).

Once all possible transmission arrangements are formed and the total transmission power required for each is estimated, the Final Decision step is initiated.

4.3.1.3 Final Decision Step

In this step, the algorithm selects the transmission arrangement with the least amount of transmission power requirements. Once selected, the size of the FACH and the DCH supported areas is defined, the required radio resources are allocated, the users are informed about the channel that is selected for them to receive and the appropriate channels are established. Also the Zone Area of the cell that will be supported using FACH is broadcasted to the UEs to facilitate efficient intra-cell handover execution (see section 4.4). This info can be broadcasted through the MBMS Point-to-Multipoint Control Channel (MCCH) [2][21], which already includes information related to the MBMS service supported within the cell.

For easy reference, the algorithm described above is summarized below.

INITIALIZATION PHASE

First create the CCIT, collect the context information required and create the MBMS users’ records.

Execute Information Collection Step

Form all possible Transmission Arrangements (TA1 – TA11) than can be used and select the most efficient one.

For each MBMS user record in the CCIT, estimate the power required if a DCH is established; Total Power for TA1 = Total Downlink Power needed if all DCHs established; Minimum Power = Total Power for TA1; Selected Zone Area = Z0 (The case where all DCH connections are established);

For i = 1 to 9 do { Total Power for TAi+1 = FACH Power to cover Zi + Sum of DCHs Power for UEs outside Zi + Sum of DCHs Power for UEs within Zi but predicted to leave Zi in C sec;

If (Total Power for TAi+1 < Minimum Power) then { Minimum Power = Total Power for TAi+1; Selected Zone Area = Zi; } }

If (Total Power for TA11 < Minimum Power) Selected Zone Area = Z10;

Now establish the appropriate FACH and DCH channels according to the Selected Zone Area.

For every MBMS user lying in Selected Zone Area and NOT predicted to leave it in C sec Type of Connection = FACH

13

Otherwise Type of Connection = DCH

Establish FACH to cover the Selected Zone Area and support the users with Type of connection == FACH; For each MBMS user with Type of Connection == DCH establish one DCH; Set in the CCIT, the “Zone Area supported using FACH” field = Selected Zone area;

Inform the users within the cell about the Zone Area that will be supported using FACH.

Broadcast through the MCCH the Zone Area supported using FACH;

4.3.2 MBMS Session Ongoing Phase During the session, users might join or leave the service, enter or leave the cell or even move within the cell, thus possibly making another transmission arrangement more efficient than the one currently used. Hence, during the session, the algorithm must periodically be executed in order to accommodate these variations and dynamically adapt to the most efficient transmission arrangement.

4.3.2.1 Information Collection Step

Periodically (period T), the RNC broadcasts a notification message within the cell, asking the users to report to it their instantaneous context. Once a report is received, the related user’s record in the CCIT is updated with the new values. When all the reports are received, the algorithm becomes aware of the new cell’s context. Then, the Transmission Arrangement Estimation step is initiated. The Period T (frequency of reporting) is a parameter that highly influences the MBMS service provision efficiency. Due to its importance, it is further analysed in section 4.3.3.

Moreover, techniques for reducing the need for reporting will be beneficial both for the users (i.e. for battery consumption) and the network (i.e. for uplink noise). One technique to achieve this (which we use in this work), is to set a threshold value and have the users report only when their instantaneous CPICH Ec/No differs from the one previously reported by this threshold. Another technique, that can be used mainly in hot spot areas, is to set a minimum coverage for FACH and have the users report only when the CPICH Ec/No they receive is weaker than a threshold value (i.e. this threshold value will define the minimum FACH’s coverage limit).

4.3.2.2 Transmission Arrangement Estimation Step

Once the Information Collection step is finished, the updated CCIT with its related FACH_IT are passed as input to the algorithm and new possible transmission arrangements are formed (similarly to section 4.3.1.2). Then, the transmission power required for each is estimated and the Final Decision step is initiated.

4.3.2.3 Final Decision Step By considering the new possible transmission arrangements formed, the algorithm checks if a more efficient than the one currently used exists and adopts it. It is worth mentioning that the new transmission arrangement that will be used is determined based on the new Zone Area of the cell that will be supported using FACH.

It should also be pointed out that when the total downlink power used within a cell exceeds 10% of the upper bound value required to cover the whole cell using FACH, the algorithm automatically increases FACH’s transmission power in order to cover the whole cell and releases all DCHs.

For easy reference, the algorithm described above is summarized below.

14

MBMS SESSION ONGOING PHASE

Periodically (Period T) and until the End of the MBMS session do the following {

Collect the context information required and update the MBMS users’ records.

Execute Information Collection step;

Form the new possible Transmission Arrangements (TA1 – TA11), estimate the transmission power required for each and then select the most efficient one to adopt.

For each MBMS user record in the CCIT, estimate the power required if a DCH is established; Total Power for TA1 = Total Downlink Power needed if all DCHs established;

For i = 1 to 9 do { Total Power for TAi+1 = FACH Power to cover Zi + Sum of DCHs Power for users outside Zi + Sum of DCHs Power for users within Zi but predicted to leave Zi in C sec; }

Total Power for TA11 = FACH Power to cover Z10

Minimum Power = Total Power for TA1; Selected Zone Area = Z0; For i = 1 to 10 do { if (Total Power for TAi+1 < Minimum Power) then { Minimum Power = Total Power for TAi+1; Selected Zone Area = Zi; } }

The TA that will be used in the cell is decided (Indicated by the value selected for the new Selected Zone Area). Now perform the appropriate actions in order for the new TA to be adopted. There are three cases here:

If (new Selected Zone Area == Zone Area currently supported using FACH) then Do Nothing. The new TA will be the same as the one currently used. If a user joins/leaves the session, or performs an intra- or inter- cell handover, will be handled by other algorithms that run in parallel.

If (new Selected Zone Area > Zone Area currently supported using FACH) then { 1. Increase FACH power to cover the new Selected Zone Area; 2. Set in the CCIT the Zone Area supported using FACH = new Selected Zone Area; 3. Broadcast through the MCCH within the cell the new Zone Area supported using FACH; 4. For each user included in the new Selected Zone Area and Not predicted to leave it in C sec:

a. Inform him to start receiving the service using the FACH; b. Set his Type of Connection = FACH; c. Release the DCH previously established for him; }

If (new Selected Zone Area < Zone Area currently supported using FACH) then { 1. For each user removed from the new Selected Zone Area and also for each user that lies within the new

Selected Zone Area but predicted to leave it in C sec. a. Establish a DCH; b. Inform him to start receiving the service using the new DCH; c. Set his Type of Connection = DCH.

2. Based on FACH_IT decrease FACH power to cover the new Selected Zone Area 3. Set in the CCIT the Zone Area supported using FACH = new Selected Zone Area; 4. Broadcast through the MCCH within the cell the new Zone Area supported using FACH; } }

During the session, the maximum power that should be used for the distribution of the MBMS service within the cell is equal to the power required to support the whole cell’s coverage using FACH (i.e. 2.0222 watts). When exceeded, the algorithm automatically increases FACH’s transmission power to cover the whole cell.

Continuously during the MBMS session do the following { 1. If power required for FACH + power required for DCHs becomes 10% greater than the power required for

covering the whole cell’s area using FACH (i.e. 2.0222 watts) then: a. Increase FACH’s transmission power to cover the whole cell’s area; b. Notify all the users within the cell to start receiving the service using FACH; c. Set their Type of Connection = FACH; d. Release all the DCHs previously used by the MBMS users;

2. Set in the CCIT the Zone Area supported using FACH = Z10; 3. Broadcast through the MCCH within the cell the new Zone Area supported using FACH; }

4.3.3 Periodicity of Reporting (Period T) The period T is a parameter that highly influences the performance of the algorithm and should be chosen as a tradeoff between the capacity, the uplink noise rise and the terminal’s battery consumption efficiency. Figure 5 illustrates how the period T can affect the efficiency of the aforementioned factors. A three minute duration steaming video of 64 Kbits/sec was used. Sixty users were set-up to move randomly within the cell with a speed selected using a uniform distribution between 6 and 30 Km/hour.

15

Ten instances of the same scenario (using the simulation parameters provided in TABLE I) were simulated, each using a different period T (from 1 to 10 seconds).

Figure 5 Affect of period T on efficiency

As shown in Figure 5, the higher the value set for the period T, the higher the average downlink power required but the less the terminal’s battery consumption and uplink noise rise. Since not only the downlink capacity but also the uplink interference and the terminal’s battery life are also considered as critical factors in the MBMS service provision efficiency, the period T should be wisely chosen so as not to sacrifice the efficiency of the one factor over the other. Taking into consideration the results shown above, for the presented scenario, a period T of about 4 or 5 seconds is a good compromise choice. However, if the Network Operator desires maximum capacity gains, a shorter period T, of say 1 second, can be used. Also, if required, the period T can be dynamically adjusted based on the available capacity in the cell. For example, if the cell becomes very loaded and radio resources are getting limited, then the period T can be decreased for more frequent reporting and thus sacrifice battery consumption and uplink noise efficiency for increased capacity. Otherwise, if the cell’s radio resources is not an issue, the period T can be increased for less frequent reporting and thus sacrifice capacity efficiency for increased battery life and reduced uplink noise.

4.4 Mobility Considerations: Intra-cell Handovers The radio resource allocation algorithm we proposed above should not run continuously as this will cause increased terminal’s battery consumption and uplink noise, due to the measurement reporting required by the UEs. Hence, during the session there will be small periods of time which this algorithm will be idle and thus any mobility of the users (more specifically intra–cell handovers; see Figure 6) will not be accommodated, possibly resulting (as mentioned in section 4.1) either in QoS or capacity inefficiencies.

Figure 6 Types of intra-cell handovers that might occur in a “Dual Transmission mode cell”

To the best of our knowledge, these new types of handovers have not been explored yet, since this is a new issue arising with the “Dual Transmission mode cell” concept.

16

Thus, below we formulate a new intra-cell handover algorithm to efficiently address them. This algorithm will run in the UE and complement the radio resource allocation algorithm, running in the RNC, during its idle periods. Its role is to monitor and efficiently execute an intra-cell handover between the FACH and DCH supported areas and therefore avoid the aforesaid inefficiencies from occurring.

4.4.1 Proposed handover algorithm: Preliminaries and Main Concept As indicated in section 4, the FACH provides capacity benefits but the QoS can be guaranteed only up to the “FACH supported area” coverage limit. Thus, in order to guarantee the required QoS while at the same time take full advantage of the capacity benefits that FACH can offer, the handover must be executed (i.e. switch channels) on the “FACH supported area” coverage limit. Note that previous work of ours [22]-[24] used a similar approach but for executing an inter-cell handover.

Since some delay is caused from the time the handover is triggered until the UE switches channels, triggering the handover on the “FACH supported area” coverage limit, will result in the switching of channels at some distance away from it. Thus, in order to accommodate this delay, a “Pre-Trigger Predictor (PP)” parameter is used. This PP will be estimated by the algorithm during the UE’s mobility. Its value is based on three parameters (see equation (1)):

• Handover delay time (Δt): Estimated by the RNC and provided to the UE, using RRC (Radio Resource Control) signalling.

• CPICH Ec/No Alteration Rate (CPICH Ec/No AR) experienced by the UE: Estimated by the UE during its mobility; if the CPICH Ec/No AR value is negative (i.e. degrading) the UE is moving away from the Node-B and if is positive (improving) the UE is moving towards the Node-B.

• Safety Margin (SM) (measured in dB and ≥ 0): Although appropriate filtering is performed on the CPICH Ec/No measurements, some minor errors might still occur resulting in the handover’s execution slightly outside of the “FACH supported area”, which will result in some packet loss. Thus, the role of the SM is to shift the handover’s execution point from the “FACH supported area” coverage limit slightly inside its guaranteed coverage, aiming to lessen any possibility of QoS degradation during the handover. The value of the SM will be fixed and selected by the Network Operator, as a tradeoff between the QoS desired to be supported and the capacity gains desired to be achieved; the smaller the value of the SM, the higher the capacity gains but the higher the possibility for QoS degradation during handover. The SM value will be acquired by the UE from the “Handover Information Table (HI_Table)” (see TABLE V in section 4.4.2) stored locally in its memory.

PP = ( CPICH Ec/No AR x Δt) + SM (1)

The value of the PP along with the value indicating the “FACH supported area” coverage limit will form the “Handover Trigger Threshold (HTT)” (see equation (2)). Note that the “FACH supported area” coverage limit is indicated by the minimum CPICH Ec/No signal quality required to be received by the UE for guaranteeing a reliable reception of the MBMS content using FACH.

HTT = Minimum CPICH Ec/No Required - PP (2)

17

The role of the HTT is to indicate to the UE the exact time of triggering the handover so that to be executed on the “FACH supported area” coverage limit. That is the UE will trigger the handover when the measured CPICH Ec/No becomes equal to the HTT. Due to alterations occurring on the user’s movement, the value of the PP and as a consequence the value of the HTT can change through time. Thus, the HTT must be continually estimated by the UE during its mobility, in order to accommodate any changes on the user’s movement and achieve efficient handover execution.

Moreover, in order to prevent the UE from any unnecessary processing during its mobility, as in [22], we introduce the idea of the “Handover Activation Area (HAA)” (see Figure 8), inside which the handover is most likely to occur; i.e. the estimation of the HTT will occur only when the user lies within this area. The size of this area is determined by the value set for the “Activation Hysteresis (AH)” parameter which will be fixed and pre-estimated by the Network Operator. The value of this parameter is set to a value which will ensure the accurate and efficient execution of the handover. In order to assist the UE to indicate when this area is reached, we define the “Handover Activation Threshold (HAT)”; its role is to indicate to the UE when to start or stop the processing required for the estimation of the HTT. Based on the intra-cell handover type that is likely to be executed, the HAT is estimated differently:

• If a handover from FACH to the DCH supported area is likely to occur, then:

HAT = Minimum CPICH Ec/No Required + AH (3)

• If a handover from DCH to the FACH supported area is likely to occur, then:

HAT = Minimum CPICH Ec/No Required - AH (4)

The HAT will be estimated only once by the UE, since its value depends on two parameters that have fixed values. These values are both acquired from the HI_Table stored locally in the UE. Once the HAT is estimated, the HAA will be defined as the area (see Figure 8) between the “FACH supported area” coverage limit and the point where the measured CPICH Ec/No signal quality equals to the HAT.

The aforesaid outlines the main concept of our proposed handover algorithm and also the parameters and thresholds vital for the efficient handover execution are discussed. Next we provide details of the HI_Table and then present the details of our algorithm.

4.4.2 Handover Information Table (HI_Table) The HI_Table will be created by the Network Operator and provided to the UE during the installation of the MBMS application. Its role is to facilitate the UE for efficient intra-cell handovers execution. The following info will be included:

• Cell ID: Uniquely identifying the cell that this HI_Table is created for.

• QoS ID: Associating the HI_Table with the MBMS service QoS requirements.

• The “FACH supported area” coverage limit of each Zone Area: Indicated by a minimum CPICH Ec/No signal quality required to be received by the UEs.

• Activation Hysteresis (AH) and Safety Margin (SM) thresholds: The value set for these thresholds is highly affected by the CPICH Ec/No alteration rate experienced by the UE. However, as illustrated in Figure 7, the CPICH Ec/No alteration rate varies at different distances from the Node-B and for different speeds. Thus, the value of the AH and the SM, will not be the same for all Zone

18

Areas. Based on Figure 7, we propose in TABLE V values for the AH and SM thresholds for each Zone Area for a pedestrian operating environment. For a vehicular environment these values will be much higher.

TABLE V HANDOVER INFORMATION TABLE (HI_TABLE) Cell ID (Identifying the Cell this HI_Table is created for)

QoS ID (Associating the HI_Table with the MBMS service’s QoS requirements) Zone Area

FACH supported area coverage limit (dB) (minimum CPICH Ec/No signal quality required to be received)

Activation Hysteresis (dB)

Safety Margin (dB)

Z1 35 dB 0.925 dB 0.1 dB Z2 21.5 dB 0.535 dB 0.06 dB Z3 14 dB 0.325 dB 0.04 dB Z4 9 dB 0.245 dB 0.03 dB Z5 5 dB 0.2 dB 0.02 dB Z6 1.8 dB 0.165 dB 0.015 dB Z7 - 0.9 dB 0.145 dB 0.015 dB Z8 - 3.3 dB 0.125 dB 0.015 dB Z9 - 5.4 dB 0.11 dB 0.015 dB Z10 - 7.2 dB 0.095 dB 0.01 dB

(a) Pedestrian - 5 Km/h

(b) Vehicular - 60 Km/h

Figure 7 CPICH Ec/No Alteration rate experienced Vs Distance from the Node-B

Note that in case some of the information included in the HI_Table changes due to some reason, it will be up to the application installed in the UE, after a notification from the RNC, to acquire any updates made.

4.4.3 Proposed Intra-cell handover algorithm: Description As indicated in section 4.4.1, in order to minimise execution time overhead the main algorithm is activated only when the MBMS user is inside the HAA. Therefore, the proposed intra-cell handover algorithm is divided into two phases:

• Phase 1: Outside of the HAA

• Phase 2: Inside the HAA

Figure 8 Moving from the “DCH supported area” towards the “FACH supported area”

19

Phase 2 is further divided into two more steps, initiated based on two thresholds:

• Activation Step: Initiated by the HAT

• Trigger Step: Initiated by the HTT Below, we describe the algorithm by considering the case where a UE is likely to handover from the DCH to the FACH supported area (see Figure 8). Note that a similar logic can be applied with the vice versa case of intra-cell handover. For easy reference, the algorithm is also described using flow charts in Figure 9.

Phase 1: Outside the Handover Activation Area When in this phase (Position 1), the UE continually measures the CPICH Ec/No signal quality and acquires from the MCCH, the Zone Area of the cell currently supported using FACH. Assuming that FACH currently supports Zone Area 3 (Z3), the UE locates the related HI_Table in its memory (using the Cell’s ID and the QoS ID mapped to the MBMS service), acquires the “FACH’s supported area” coverage limit and the AH values associated with Z3 and estimates the HAT. Once this threshold is estimated the UE continuously compares it with the measured CPICH Ec/No. When the measured CPICH Ec/No becomes equal to or stronger than (≥) the HAT (Position 2), the algorithm transits to Phase 2, otherwise remains in Phase 1.

(a) From DCH towards the FACH supported area

(b) From FACH towards the DCH supported area

Figure 9 Intra-cell handover algorithm’s flow charts

Phase 2: Inside the Handover Activation Area Activation Step: This step is activated automatically upon transition to Phase 2 (Position 2) and indicates to the UE that it has entered the HAA. The UE upon entering this area requests from the RNC the handover delay time and also acquires from the HI_Table the SM threshold value associated with Zone Area 3 (Z3). Then,

20

the processing required for the estimation of the HTT is initiated. The estimation of this threshold must be performed continually in order to accommodate any changes on the user’s movement and achieve efficient handover execution. Thus, every time a new HTT value is estimated it is compared with the currently measured CPICH Ec/No. When the measured CPICH Ec/No becomes equal to or stronger than (≥) than the HTT (Position 3), the algorithm will transit to the Trigger Step.

From within the Activation Step, if the measured CPICH Ec/No becomes weaker than (<) the HAT, meaning that the user has left the HAA, then we go back to Phase 1.

In order to further reduce the processing effort required by the UE, the HTT will be estimated only when the CPICH Ec/No alteration rate experienced from the serving cell is increasing (i.e. the UE is moving towards the Node-B; see Figure 9.a). Otherwise, there is no need to estimate it since the user is either stable or moving away from the “FACH supported area” and an intra-cell handover from the DCH to the FACH supported area is not likely to occur. Similar logic is applied with the vice versa case of intra-cell handover, too (see Figure 9.b).

Trigger Step: The UE prepares and sends the handover request to the RNC (Position 3). The RNC processes the request and the handover is executed (Position 4).

4.4.4 Importance of Pre-Trigger Predictor (PP) Parameter As discussed in section 4.4.1, the aim of the PP parameter, is to accommodate the Handover delay time (Δt) caused and avoid any QoS degradation or capacity inefficiency during the intra-cell handover. In order to investigate the amount of Δt that is tolerated when an intra-cell handover is performing, a series of scenarios have been simulated. The following results were recorded; ~ 0.9 second Δt when the UE is handing over from the DCH to the FACH supported area, ~ 1.0 second Δt when the UE is handing over from the FACH to the DCH supported area (the additional 0.1 second is needed for the establishment of the DCH).

Taking into consideration the Δt values indicated above, we demonstrate the importance of the PP parameter in the handover triggering and also the QoS and capacity gains achieved by our algorithm. For the scenarios used the simulation parameters provided in TABLE I were used. The UEs are expected to receive an MBMS streaming video of 64 Kbits/sec and the Zone Area supported using FACH is Z7 (i.e. FACH covers ~700 meters of the cell’s area).

First, we demonstrate the impact of the PP parameter on the QoS during an intra-cell handover from the FACH to the DCH supported area. In this case the handover must be triggered when the measured CPICH Ec/No equals to “Minimum CPICH Ec/No Required minus (-) PP”. If the PP is not considered (i.e. PP = 0) the handover will be triggered on the “FACH supported area” coverage limit, and the channel switching from FACH to DCH will occur 1.0 second after the “FACH supported area” is left (during this 1.0 second the UE will continue receiving the service using FACH until the switching to DCH occurs). Since FACH guarantees the requested QoS only up to the “FACH supported area” coverage limit, it is obvious that this will have some impact on the traffic received. To illustrate this impact, six instances of the same scenario, each having the UE moving with a different speed, were simulated.

As illustrated in Figure 10, the packet loss experienced when the PP is considered is up to 2.34% when 120 Km/h speed is used. If the PP is not considered, we have losses from 1.36% up to 6.23% for low speeds (5 and 15 Km/h) and from 15 % up to 96.7 %

21

for vehicular speeds (30, 60, 90 and 120 Km/h). For many streaming applications, an acceptable QoS can still be provided, if the packet losses in the received service from the UE do not exceed 10%. This is met for all the cases when PP is considered.

(a) Throughput received (bits/sec)

(b) QoS degradation – Packet loss (%)

Figure 10 Impact of UE’s speed on QoS received during intra-cell handover

Similarly, during an intra-cell handover from the DCH to the FACH supported area, if the PP is not considered, the channel switching will be executed inside the “FACH supported area”, 0.9 seconds after passing the “FACH supported area” coverage limit (during this 0.9 seconds the UE will continue receiving the service using DCH until the switching to FACH occurs). The optimum case here would be to break the DCH connection and switch to FACH exactly on the “FACH supported area” coverage limit in order to release the DCH earlier and avoid any redundant use of power. In order to illustrate this impact, we use a scenario in which a group of UEs receiving the same MBMS service are travelling together with a speed of 30 Km/h and performing the intra-cell handover at the same time. Three instances of this scenario were simulated, having 1, 2 and 3 users, respectively.

Figure 11 Average Downlink Transmission Power requirements (watts)

As illustrated in Figure 11, significant transmission power savings can be achieved when the PP parameter is considered. A trivial amount (up to 2.6% more) of additional transmission power compared to the optimum case2 is used when the PP is considered. Note that this additional power is the tradeoff of using the SM parameter in the estimation of the PP, which shifts the handover execution point slightly inside the “FACH supported area” guaranteed coverage, aiming to lessen any possibility of QoS degradation during the handover. Then again, when the PP parameter is not considered, the additional transmission power used is considerable (from 53.8% up to 173.7 % more in this scenario). Thus, in this case the use of the PP is essential.

2 The optimum case reflects the situation in which the channel switching and thus the releasing of the DCHs occurs exactly on the “FACH supported area” coverage limit. In this case the average downlink power required should be 0.4855 watts, which is the amount of transmission power required to be devoted to FACH for guaranteeing a reliable reception of the MBMS service at 700 meters from the Node-B.

22

The results illustrated above demonstrate the importance of the PP parameter in the intra-cell handover triggering and also the QoS and capacity gains that can be achieved by our algorithm during the intra-cell handover process.

5 Performance Evaluation Our proposed “Dual Transmission mode cell” approach has been implemented and evaluated in the MBMS simulator [25]. This simulator was created during the B-BONE project using as a base the UMTS module provided by OPNET Modeller 11.0.A [26]. In order to illustrate the performance of our proposed approach compared to other approaches, we also implemented in the same simulator the “UE Counting” [2], the “Power Counting” [12], the “Rate Splitting” [13] and the “FACH with dynamic power setting” [15] approaches. The scenarios used for the evaluation are described in section 5.1 while the results obtained are analyzed in section 5.2.

5.1 Scenarios For the performance evaluation, two scenarios have been simulated. Scenario 1 is a simple one-cell scenario used for illustrating the gains achieved and the tradeoffs of each of the aforementioned approaches. Scenario 2 is a more complex scenario and used for illustrating the feasibility, the performance, and the usefulness of our proposed “Dual Transmission mode cell” approach in a multi-cell environment where the inter-cell interference is increased and also inter-cell handovers occur. It is worth mentioning that although inter-cell handovers are included in scenario 2, these are not in the scope of this paper thus will not be further discussed (for this issue, a discussion appears in [22]). For each scenario five instances were simulated (using the same simulation parameters provided in TABLE I), using the “UE counting”, the “Power Counting”, the “Rate Splitting”, the “FACH with dynamic power setting” and our proposed “Dual Transmission mode cell” approach, respectively. For “FACH with dynamic power setting” a Period T of 0.5 seconds was required in order to accommodate the mobility of the users and avoid any QoS degradation. For all other approaches the same Period T of 4 seconds was used (as proposed in section 4.3.3).

5.1.1 Scenario 1: One-cell scenario In this scenario all the users are placed randomly within the cell and receive an 8 minute duration 64 Kbits/sec MBMS streaming video clip. All users are set to move randomly within the cell, with a speed selected using a uniform distribution between 6 and 30 Km/hour. This scenario has been simulated using a high loaded (65 users) and a low loaded (15 users) cell. The results collected are analysed in section 5.2.1.

5.1.2 Scenario 2: Multi-cell scenario In this scenario we consider the case where a large number of users are moving from surrounding cells towards the central cell where a stadium is located and a football match is ready to start in 20 minutes. A lot of users have already reached the stadium while a lot of others are still approaching. During this time, in order to entertain the crowd already in the stadium, but also those that are still approaching, a 20 minute 64 Kbits/sec video clip is transmitted, showing the best scenes of past football matches between the two teams. In this scenario we assume that the major part of the crowd is already in the stadium (i.e. more than 5000 users) and wait for the football match to start. The results collected are analysed in section 5.2.2.

23

5.2 Analysis of Results

5.2.1 Scenario 1: One-cell Scenario The results collected from scenario 1, relate to the tradeoffs and the gains achieved by each approach. Due to paper’s space limitation, only results related to cells with pedestrian and low vehicular users are presented. However, similar results were obtained with higher bit rates video clips (256 Kbits/sec) and higher vehicular speeds.

(a) Instantaneous Downlink Power used (watts)

(b) Average Downlink Power used (watts)

Figure 12 High Loaded Instance (65 users) - Downlink power requirements (watts)

(a) Instantaneous Downlink Power used (watts)

(b) Average Downlink Power used (watts)

Figure 13 Low Loaded Instance (15 users) - Downlink power requirements (watts)

From the results illustrated above, we observe that our proposed “Dual Transmission mode cell” approach outperforms all others in terms of transmission power efficiency. More specifically, for the high loaded instance (Figure 12), the “Dual transmission mode cell” average power used is 1.0392 watts, whilst, for “Rate Splitting” is 1.1812 watts (13.7% more), for “FACH with dynamic power setting” is 1.3827 watts (33% more), for “Power Counting” is 1.8309 watts (76.2% more), and for “UE Counting” is 2.0222 watts (94.6% more). For the low loaded instance (Figure 13), the “Dual transmission mode cell” average power used is 0.6936 watts, whilst, for “Rate Splitting” is 0.8511 watts (22.7% more), for “FACH with dynamic power setting” is 1.0248 watts (47.7% more), for “Power Counting” is 1.1946 watts (72.2% more), and for “UE Counting” is 2.0222 watts (191.5% more).

For the “UE Counting” approach, a threshold of 6 UEs was used. Thus for both high and low loaded instances, a FACH covering the whole cell area was established allocating 2.0222 watts and retained throughout the session. With the “Rate Splitting” approach, when P-t-M transmission mode is used, two FACHs are established. The

24

first one allocating 0.9471 watts and transmitting the basic layer of the video clip (32Kbits/sec) in the whole cell area while the second one allocating 0.2357 watts and transmitting the enhancement layer (32Kbits/sec) covering 700 meters of the cell. Thus, the tradeoff with this approach is that UEs further than 700 meters from the Node-B receive a worse quality video. With the “FACH with dynamic power setting” approach only one FACH is used which is frequently adjusted based on the worst users’ pathloss within the cell. With this approach not only very frequent reporting is required but also it can sometimes result in redundant use of power in cases when only one UE is in a bad position. Our proposed “Dual Transmission mode cell” approach, manages to minimise the inefficiencies occurred by the other approaches and achieves significant transmission power savings and best QoS support.

Also, by comparing the results collected from the low and high loaded instances (see Figure 12.b and Figure 13.b), we can observe that the average downlink power used by the “FACH with dynamic power setting” approach in the low loaded instance (i.e. 1.0248 watts to support 15 users), is almost the same amount used by the “Dual Transmission mode cell” approach in the high loaded instance (i.e. 1.0392 watts to support 65 users). Thus, for the presented scenario, our proposed scheme manages to support 50 more users (433% increase) than the “FACH with dynamic power setting” approach. Similar benefits are obtained in comparison with the other competing schemes presented above.

(a) Instantaneous Channel Quality received (dB)

(b) Average Channel Quality received (dB)

Figure 14 Downlink Channel Quality (CPICH Ec/No) Received (dB)

By reducing the downlink transmission power requirements, the intra-cell interference caused in the cell is also reduced, thus improving the channel quality experienced by the users moving within. Figure 14 compares the instantaneous and average CPICH Ec/No experienced by a UE during its mobility (in the high loaded instance) when the different approaches referred above are applied. As illustrated in Figure 14.b, our approach outperforms all other approaches in terms of link performance improvement.

Next we consider the aggregated processing effort introduced by each approach in the high loaded instance (TABLE VI). The approach that outperforms all others in terms of processing effort efficiency is the “UE Counting”; only one FACH is established and maintained throughout the session. The “FACH with dynamic power setting” approach follows, which establishes only one FACH, but additional processing effort is required for its frequent coverage adaptation during the session. However, this processing effort is trivial since only one command is required to be sent by the RNC to the Node-B to adapt the power devoted to FACH. Then, with similar performance and medium amount of processing effort introduced come the “Rate Splitting” and the

25

“Dual Transmission mode cell” approaches, followed by the “Power Counting” which introduced almost three times more processing effort than the aforementioned.

TABLE VI AGGREGATED PROCESSING EFFORT INTRODUCED IN RNC AND NODE-B

UE

CountingPower

CountingFACH with dynamic

power setting Rate

Splitting Dual transmission

mode cell Action (Times Performed)

Establish FACH 1 3 1 4 1 Release FACH 0 2 0 2 0 Establish DCH 0 195 1 0 65 1 81 3

Release DCH 0 195 2 0 65 2 76 4

Increase FACH’s power 0 0 497 0 13 Reduce FACH’s power 0 0 416 0 11

1 Caused due to “From P-t-M to P-t-P” transmission mode switching 2 Caused due to “From P-t-P to P-t-M” transmission mode switching 3 Caused due to transmission arrangement adaptation (35) and intra-cell handovers (46) 4 Caused due to transmission arrangement adaptation (37) and intra-cell handovers (39)

From the results presented above, we observe that our “Dual Transmission mode cell” approach with medium processing effort introduced can support the requested QoS for all the users with the least amount of transmission power requirements and also provides the best link performance improvement than any of the other competing approaches. However, for the gains to be achieved, periodical measurement reporting is required by the UEs throughout the session. Below, results concerning the terminal’s battery consumption (Figure 15) and the uplink noise rise (Figure 16) by each approach, due to this reporting requirement, are presented and compared.

Figure 15 Terminal’s Battery Consumption due to reporting (watts)

(a) Instantaneous Uplink Noise Rise

(b) Aggregated Uplink Noise Rise

Figure 16 Uplink noise rise due to reporting (femtowatts 1 femtowatt = 1 x 10-15 watts)

26

As observed, in Figure 15 and Figure 16, the approach that outperforms all others in terms of terminal battery consumption and uplink noise rise efficiency is the “UE Counting”. Based on the aforementioned results, if we assume a 3.6 Volt Li-Ion 1100 mAh cell phone battery, a consumption of about 4 watts gets the battery fully discharged (Watts = Volt x Ampere). Thus, for an 8 minute duration video clip, the “UE Counting” consumes 2.85% of the battery life time, the “Power Counting” 4.45%, the “Rate Splitting” 6.45%, the Dual Transmission mode cell” 7%, while the “FACH with dynamic power setting” 28.42%. Also, the aggregated uplink noise introduced by each approach shows similar trends.

The results presented above (for the high loaded instance) are also summarized and compared in TABLE VII. Given the tradeoffs and the benefits achieved, we can conclude that our approach significantly outperforms the other competing approaches.

TABLE VII SUMMURY OF THE RESULTS PRESENTED

UE

Counting Power

Counting FACH with

dynamic power setting

Rate Splitting

Dual transmission

mode cell Average Transmission Power 2.0222 watts 1.8309 watts 1.3827 watts 1.1812 watts 1.0392 watts

Average Link Quality received 4.6750 dB 5.3897 dB 7.1325 dB 7.9195 dB 9.1550 dB QoS supported for all users Yes Yes Yes No Yes Processing effort introduced Minor Major Minor Medium Medium

Battery consumption 2.85% 4.45%, 28.42%. 6.45%, 7% Uplink Noise rise (femtowatts) 203.34 372.84 4009.49 516.03 553.61

5.2.2 Scenario 2: Multi-cell Scenario The results collected from scenario 2, relate to the transmission power savings and link performance gains that can be achieved by our proposed approach, in an environment that is closer to real cellular operating environments.

(a) Instantaneous Downlink Power used

(b) Average Downlink Power used

Figure 17 Multi-cell Scenario - Downlink power requirements (watts)

From the results illustrated in Figure 17 and Figure 18, we observe that our proposed “Dual Transmission mode cell” approach again outperforms all others in terms of transmission power and link performance efficiency. More specifically (see Figure 17), the “Dual transmission mode cell” average power used is 1.0786 watts, whilst, for “Rate Splitting” is 1.1828 watts (9.7% more), for “FACH with dynamic power setting” is 1.4531 watts (34.7% more) and for “Power Counting” and “UE Counting” is 2.0222 watts (87.5% more). By reducing the inter- and intra- cell interference

27

experienced by the users moving within the central cell, major improvement on the downlink channel quality received by the users was also achieved (see Figure 18).

(a) MBMS users near the cell’s edge

(b) MBMS users near the cell’s Node-B

Figure 18 Average Downlink Channel Quality (CPICH Ec/No) received (dB)

6 Conclusions and Future work In this paper we introduce the “Dual Transmission mode cell” concept in the MBMS service provisioning and propose a new radio resource allocation algorithm in order to efficiently manage the radio resources of this new type of cell. With this new concept, P-t-M and P-t-P transmissions are allowed to coexist within the same cell and dynamically adapt through time, in such a way that the requested QoS to be supported with the least amount of transmission power consumption. The main idea of introducing this new type of cell in the MBMS service provisioning is to take full advantage of the benefits that both transmission types can offer (i.e. the capacity benefits of FACH and the fast power control of DCH) and achieve increased radio network capacity and performance. A problem formulation is presented and then a solution approach is discussed, considering practical concerns. The “Dual Transmission mode cell” concept introduces new mobility issues, more specifically new intra-cell handovers, which are also addressed in this paper. The performance evaluation carried out showed that our proposed “Dual Transmission mode cell” approach, provides considerable gains and outperforms all other related proposed approaches in terms of capacity and link performance efficiency, thus highlighting its feasibility, importance and usefulness in the MBMS service provisioning. However, the aforesaid gains have been achieved having as tradeoff, small increases in terminal’s battery consumption and uplink noise.

Future work will include monitoring standardization progress in using HSDPA for the MBMS bearer services provision and extending our algorithm accordingly. Also techniques for reducing the need for context reporting by the users will be investigated and proposed. Moreover, we plan to investigate and propose an architecture and RRM algorithm that will allow the MBMS system to be integrated and be efficiently provided in systems beyond 3G (Heterogeneous Networks).

Acknowledgment This work is partially supported by the European Commission funded FP7 ICT-2007-216462 C-CAST project. The authors also thank OPNET Technologies Inc. for providing the software licences to carry out the simulations for this research.

28

References [1] 3GPP TS 23.246 V9.1.0, “3rd Generation Partnership Project; Technical Specification

Group Services and System Aspects; Multimedia Broadcast/Multicast Service (MBMS); Architecture and functional description (Release 9)”, June 2009.

[2] 3GPP TS 25.346 V8.3.0, “3rd Generation Partnership Project; Technical Specification Group Radio Access Network; Introduction of the Multimedia Broadcast Multicast Service (MBMS) in the Radio Access Network (RAN); Stage 2 (Release 8)”, March 2009.

[3] 3GPP TS 23.101 v9.0.0, “Technical Specification Group Services and System Aspects: General Universal Mobile Telecommunications System (UMTS) architecture (Release 9)”, December 2009.

[4] 3GPP TS 22.146 V9.0.0, “3rd Generation Partnership Project; Technical Specification Group Services and System Aspects; Multimedia Broadcast/Multicast Service; Stage 1 (Release 9)”, June 2008.

[5] 3GPP TS 22.246 V8.5.0, “3rd Generation Partnership Project; Technical Specification Group Services and System Aspects; Multimedia Broadcast/Multicast Service (MBMS) user services; Stage 1 (Release 8)”, March 2008.

[6] 3GPP TR 23.846 6.1.0, “3rd Generation Partnership Project; Technical Specification Group Services and System Aspects; Multimedia Broadcast/Multicast Service (MBMS); Architecture and functional description (Release 6)”, December 2002.

[7] Ogunbekun J., Mendjeli A., “MBMS service provision and its challenges”, 4th International Conference on 3G Mobile Communication Technologies 2003 (3G 2003), 25-27 June 2003, Pages: 128 – 133.

[8] 3GPP TSG RAN WG1#28bis, “Power Usage for Mixed FACH and DCH for MBMS,” Lucent Technologies, R1-02-1240, October, 2002.

[9] C. Chaikalis, “Efficient TTI for 3G Multimedia applications”, Advances in Multimedia, Hindawi Publishing Corporation, New York, United States, Year of publication: 2007, Volume: 2007, Issue: 1, ISSN 1687-5680.

[10] 3GPP TSG RAN WG2 #29 “Considerations on power allocation for MBMS”, Samsung, R2-021669, June 2002.

[11] B-BONE, Deliverable 3.2, “Report on Broadcast/Multicast RAN Resource Management Algorithms”, August 2005. http://seacorn.cs.ucy.ac.cy/bscw/bscw.cgi/d108910/WP3_AUEB_D32.doc

[12] 3GPP TR 25.922 V7.1.0, “3rd Generation Partnership Project; Technical Specification Group Radio Access Network; Radio resource management strategies (Release 7)”, March 2007.

[13] 3GPP TSG RAN WG2 #31, “MBMS Power Usage”, Lucent Technologies, R2-022110, August, 2002.

[14] S. Armando, S. Nuno, C. S. Joao, E. Patricia, C. Americo, “Effective Radio Resource Management for MBMS in UMTS Networks”, An International Journal on Wireless Personal Communications, Volume 42 , Issue 2 (July 2007), Pages: 185 – 211, 2007.

[15] 3GPP TSG-RAN WG1#29, “Power Control for FACH,” Lucent Technologies, R1-021324, November 2002.

[16] M. Chuah, T. Hu, W. Luo, “UMTS Release 99/4 airlink enhancement for supporting MBMS services”, IEEE International Conference on Communications 2004 (ICC 2004), Paris, France, June 2004, Volume: 6, Pages: 3231- 3235.

[17] C. Bouras, A. Alexiou, V. Kokkinos, “Power saving techniques in MBMS multicast mode”, 3rd ACM Workshop on Wireless Multimedia Networking and Performance Modeling 2007 (WMuNeP’07), Chania, Crete Island, Greece, October 22 - 22, 2007, Page(s): 62 – 69.

[18] C. Christophorou, A. Pitsillides, “A new Approach for Efficient MBMS Service Provision in UTRAN”, IEEE Symposium on Computers and Communications 2008 (ISCC'08), 6-9 July 2008, Marrakech, Morocco, Pages: 356 – 361.

29

30

[19] C. Christophorou, A. Pitsillides, V. Vassiliou, “An Enhanced Radio Resource Allocation Approach for Efficient MBMS Service Provision in UTRAN”, IEEE International Symposium on Personal, Indoor and Mobile Radio Communications 2008 (PIMRC 2008), 15-18 September 2008, Cannes, French Riviera, France, ISBN: 978-1-4244-2643-0, Pages: 1 - 5.

[20] A. Alexiou, C. Bouras, V. Kokkinos, E. Rekkas, “A Novel Mechanism for Radio Capacity Maximization during MBMS transmission in B3G Networks”, The 11th ACM/IEEE International Conference on Modeling, Analysis and Simulation of Wireless and Mobile Systems (MSWiM 2008), Vancouver, British Columbia, Canada, 27 - 31 October 2008, Pages: 185 – 192.

[21] 3GPP TS 25.331 V8.4.0, “3rd Generation Partnership Project; Technical Specification Group Radio Access Network; Radio Resource Control (RRC); Protocol Specification (Release 8)”, September 2008.

[22] A. Pitsillides, C. Christophorou, “MBMS Handover Control: A New Approach for Efficient Handover in MBMS Enabled 3G Cellular Networks”, Elsevier Computer Networks Journal (COMNET), December 2007, Volume 51, Issue: 18, Pages: 4897-4918.

[23] C. Christophorou, A. Pitsillides, “An enhanced approach for efficient MBMS handovers in 3G networks”, IEEE Symposium on Computers and Communications, 2008 (ISCC 2008), 6-9 July 2008, Marrakech, Morocco, Pages: 362 – 367.

[24] C. Christophorou, A. Pitsillides, “A NEHO based MBMS Handover Control Approach”, IEEE International Symposium on Personal, Indoor and Mobile Radio Communications 2008 (PIMRC 2008), 15-18 September 2008, Cannes, French Riviera, France, ISBN: 978-1-4244-2643-0, Pages: 1 -5.

[25] B-BONE MBMS System Level simulator: http://b-bone.ptinovacao.pt/docs/BBONE-Reference-Model.rar, October 2006.

[26] OPNET University Program: http://www.opnet.com /services/ university/