Architecture and Schemes for Intelligent Mobility Management in Future Mobile Telecommunication Systems Kuo-Hsing Chiang, Nirmala Shenoy Department of Computer Systems Engineering, RMIT University GPO Box 2476v, Melbourne, Victoria, 3000, Australia Email: kuo_hsing@hotmail.com, nirmala.shenoy@rmit.edu.au Tel: +61 3 9925 5360, Fax: +61 3 9925 5340 Abstract - In this paper, we present a novel terrestrial mobile network architecture that aims at the simplification of mobility management for the terrestrial mobile telecommunication networks. The work is in line with the UMTS approach. The proposed architecture uses the hierarchical cell concept and facilitates strategic placement of database control. A macro-cell based approach for location updating and a combined macro/micro cell based paging control plays a key role in our proposal. Another significant feature of the proposed scheme is that, each micro-cell has the same signalling load, regardless of the location updating traffic. A Location Agent (LAg) that is co-located with Signalling Transfer Point (STP) and hierarchically structured is also proposed to reduce mobility management signalling traffic to and from the HLR. A swap technic is applied at the location databases, which further reduces the signalling traffic. The simplifications achieved through the use of this architecture benefit both location updating and paging procedures by optimising the traffic in both cases. I. INTRODUCTION The future mobile systems aim at enabling subscribers to communicate with anyone, anytime and anywhere. Mobility management is always a key issue for the mobile networks to achieve more efficient location management and handover control. The existing techniques for location management are location updating (registration) and paging. The first generation mobile networks, such as Advanced Mobile Phone Service (AMPS), page the Mobile Station (MS) in every cell within the network. The second generation mobile networks, such as Global System for Mobile communication (GSM) [1] use Home Location Register (HLR), Visitor Location Register (VLR) and Location Area (LA) concepts to achieve more bandwidth efficiency in air interface. In this paper, we propose an architecture for an integrated wireless/fixed network in line with the UMTS approach [2,3] but is very much superior from the point of view of optimal signalling traffic and shared signalling loads with faster access during paging. The architecture has been modeled to achieve efficient mobility management by using hierarchical distributed database and intelligent paging techniques as suggested in the review paper [9]. The basic architecture described in section II, is the significant contribution based on which optimal mobility management schemes are proposed. A hierarchical Location Agent concept used for location management is explained in section III A. In section III B, the swap technic is applied to location database management to further reduce the signalling traffic. In section III C, a location updating using macro-cell and the information flows are discussed. A macro/micro cell based intelligent paging scheme as applied in a location area is proposed in section III D. Section III E describes the location management in picocell environment and the handover control is briefly discussed in section III F. Section IV provides discussion and conclusions. II. BASIC ARCHITECTURE Figure 1 illustrates the proposed network architecture. The proposed network is built on concepts from UMTS and GSM. The changes proposed by us have been highlighted using Italics. The basic network consists of three sub-networks: radio access network, core network, and a signalling and intelligent network. • Radio access network provides mainly radio related functions, the basic local cell site switching (CSS) and transmission functions, and interfaces with core and signalling networks. A group of pico-cells are connected to one Pico-cell CSS (PiCSS) and a group of micro-cells and one macro-cell are connected to a BSC/CSS. The BSC/CSS coverage area equals to one macro-cell coverage area and also equals to a Location Area. The PiCSS and BSC/CSS are at the same level of hierarchy and connected to LE/MSC. This type of umbrella architecture has been proposed earlier, but here we have slightly rearranged it and utilised this architecture to help strategic placement of databases, thereby achieving optimised mobility management. • The core network consists of enhanced local and transit switches of B-ISDN and also includes a transmission system to provide switching, call control and connection control. • Intelligent and signalling network will provide specific services and location management. Distributed databases will be required in Intelligent Network (IN) to store user related data (user/service profiles) to support access to the network and mobility services. A LAg co-located with STP is proposed. The function of the LAg briefly, is to provide hierarchical pointers to the VLRs a mobile station is traversing, where the hierarchy is controlled by the mobility of the user. A detailed explanation of this concept follows. STP Intelligent and Signaling Network HLR MS initial entry into remote or visited network: Signaling Network MS STP STP/LAg LE/MSC/VLR Core Network LE/MSC/VLR ( Remote or Visited Network ) VLRnew STP/LAg HLR Location updating (or registration) Location updating (or registration) Location updating (or registration) LE/MSC/VLR Location updating (or registration) ACK Location updating (or registration) ACK PiCSS BSC/CSS PiBCSS MS moving from VLRold to VLRnew : Radio Access Network MS Pico-cells Pico-cells Micro-cell Micro-cell Figure 1. Proposed Network Architecture III. MOBILITY MANAGEMENT A. Location Management Using Hierarchical LAgs at STPs In GSM to locate a mobile user, the HLR is always accessed first and the locality of connection is not exploited. In the future network, the Local Exchange (LE) may be enhanced to include Mobile Basic Call State Model. The LE shall be able to handle both fixed and mobile calls [4]. Based on this approach, a local database at the LE can control both fixed and mobile calls that are in the same LE coverage area. To reduce the signalling traffic when an MS roams to other networks or away from its home LE, a Location Agent (LAg) concept is used. In our configuration the LAg is distributed at the SS7 STPs. This means that the SS7 STP will be enhanced to include extra information. If the MS is roaming to other VLRs under the same LAg coverage area, the LAg will record the VLR identifiers. On the other hand, if the MS is in the remote part of the network or in the other networks, the LAg may acts like HLR Agent and replicates all the necessary information for the call set-up and mobility management. This will be a distributed HLR in STP level in the network. Figure 2 illustrates the information flows of the location updating by LAg in remote or visited network. A STP/LAg connects from one to several LE/MSCs. We propose locality of connection to be exploited as follows: in Figure 3 the LE is an integrated MSC and SSP therefore the call control can be localised at LE/MSC level. For example, when a fixed user initiates a call to a mobile user and both users are under the same LE control, the network will first check the local database which is co-located with LE therefore the call can be set up faster and the signalling traffic will be reduced. The other approach shown in Figure 4 is to connect the LEs and MSCs to a common STP that will be the signalling anchor point and therefore saves the signalling traffic to HLR. This approach can be used in the current mobile network when the LE and MSC are not integrated. STP/LAg HLR Location updating Macro-cell BSC: Base Station Controller HLR: Home Location Register LAg: Location Agent LE: Local Exchange MSC: Mobile Switching Centre PiCSS: Pico-cell CSS STP: Signaling Transfer Point VLR: Visitor Location Register CSS: Cell Site Switch ( Remote or Visited Network ) VLRold VLRnew Location updating Location updating ACK Location updating ACK Figure 2. Information Flows of the Simplified Location Updating by Lag in Remote or Visited network In our configuration, the LAgs are distributed at SS7 STPs. Based on the users’ mobility patterns any one of the LAgs will be the location agent for that mobile user. For example in figure 5, LAg111 is at the lowest level for low mobile users, LAg11 at the next level for medium mobile users and LAg 1 at the top level for highly mobile users. MS1 is a subscriber who spends most of his time under the local STP/LAg111 and his VLR pointer is recorded only at Lag111 while MS3 has higher mobility and often goes across different STP coverage areas and hence his VLR pointer will be at LAg1. An inherent advantage of this scheme is that the information to be stored in the LAgs is distributed fairly across the different hierarchical layers of STP/LAgs. If the LAg acts as a HLR Agent then for highly mobile users like MS3, having his profiles recorded at LAg1 will save some traffic when he moves from coverage under STP/LAg112 to STP/LAg121. If LAg1 didn’t have his profiles then the user’s information has to be transmitted via LAg112, LAg11, LAg1, LAg12, and LAg121 to the new VLR. Our approach has a shorter signalling path to transmit user’s information comparing with location forwarding strategy proposed in [10]. The proposed scheme will avoid heavy database processing at a particular STP and also reduce the signaling traffic to a remote HLR. The signaling load will be localized and distributed more evenly in all STP regions. Analytical studies of comparing different location management schemes are being conducted. STP/LAg LE/MSC LE/MSC STP/LAg LE/MSC Figure 3. LE and MSC are Integrated LE MSC LE MSC LE Figure 4. LE and MSC are Separated STP/LAg1 STP/LAg12 STP/LAg11 STP/LAg111 STP/LAg112 VLRold STP/LAg121 STP/LAg122 VLRnew MS1(LAg111) MS2(LAg11) MS3(LAg1) Figure 5. Distributed LAg’s for Different Class of User B. Swap technic for Location Management We have further divided the conventional location database (e.g. VLR) into two parts: current and history databases. The current database records the user/service profiles of the MSs that are currently residing in the VLR coverage area and is in the primary storage. The history database is stored in the secondary storage. It records the profiles of the MSs that visited the VLR previously in a certain period. The swap technic is used to exchange information between the primary and secondary databases. In the conventional mobile network (e.g. GSM, IS-41), the location updating algorithm will update the profiles at VLRnew and the profiles at VLRold will be deleted. In our approach, the profiles at VLRold will be swapped from the primary database to secondary database. The profiles are deleted only based on the mobility patterns and the timers. By using the swap technic at VLR, the signalling traffic for profiles download will be reduced significantly. The comparison between conventional and our approach is illustrated in Figure 6. There is a profiles download procedure every time when the MS roaming from a VLR to another VLR in the conventional approach. In our approach, the profile can be swapped and it is not necessary to be downloaded to the VLRs. In conjunction with the hierarchical algorithm proposed in the last section, the swap algorithm applied at location databases can further reduce the signalling traffic for location management. C. Location Updating Using Macro-cell In the conventional location updating approach (e.g. GSM), the location updating occurs only in perimeter cells, regardless of periodical updating. The radio resources consumed in the perimeter cells are much more than the inner cells. This causes load unbalance for each cell and the different design for perimeter and inner cells. Other researchers [11,12] use different dynamic LA management and updating schemes which need to record the different users’ mobility parameters. However, it is generally not easy to have dynamic LA for different MS’s as the MS’s must be able to identify the boundaries of LA’s which are continuously changing [9]. In our proposal, the existing macro-cells are used as LAs and a BSC/CSS control area equals to a macro-cell coverage area and also equals to a LA. This is the novelty of our approach as this leads to simplified approaches to location updating and paging. Figure 7 shows the network planning for the BSC or CSS. In this approach, the macro-cell will broadcast the LA identifier which, could be the same as macro-cell identifier to save signalling overhead. The main advantage of this architecture is to simplify the location updating control through macro-cell. Another important effect is that each micro-cell will have the same signalling load, regardless of the location updating traffic, as it does not have to broadcast the LA identifier. The macro-cell can also act as backup resources during inter micro-cell handover failures. Because the macro-cell and micro-cells are under the same BSC/CSS control it makes BSC/CSS control of radio resources easier during handover and location management. This concept can be extended to the pico-cell and micro-cell environments. The architectural and functional models proposed in this work can be easily scaled to fit the environment. Figure 8 illustrates the simplified location updating procedures and information flows. Conventional Location Updating Procedures: VLRnew STP VLRold HLR Location update req. Location update resp. (with Auth. parameters or profiles) Location update req. Location update resp. (with or without profiles) Location cancellation req. via STP11 Location cancellation resp. via STP11 Location Updating Procedures Using Swap technic: A. MS moves from VLR1 to VLR2: STP11 VLR2 VLR1 HLR Location update req. Location update resp. (with profiles) SWAP at VLR1 Update location pointer at STP11 Location update resp. (with profiles) B. MS returns to VLR1 from VLR2: VLR1 STP11 VLR2 HLR Location update req. Location update resp. (without profiles) SWAP at VLR2 Update location pointer at STP11 Location update resp. (without profiles) SWAP VLR1 Figure 6. Information Flows of Conventional and Swap Location database Management BSC/CSS xi 2 R yi R P( x, y ) {( xi 2 R, yi R), ( xi 2 R, yi R), ( xi 2 R, yi R), ( xi 2 R, yi R)}, xi 2 R yi R 0 x 16,0 y 8 • For example at P2 (16,4), R=2, the first paging area composes cells {(14,16), (13,15), (15,15), (12,4), (14,4), (16,14), (13,3), (15,3), (14,2)}. If there is no response from PA1, the rest of cells in the LA will be paged. Due to the service reliability reasons, if the MS is still not responding to the first 2 pages, we need to page the whole LA either using macro-cell (if micro-cell paging failed) or micro-cells again. LA Macro-cell • Micro-cells Figure 7. Overlaid Cell Planning Model MS BSC/CSSold BSC/CSSnew VLR Y Monitoring the LA or macro-cell identifiers Location updating request Location updating request LA update Location updating ACK Location updating ACK Figure 8. Information Flows for Location updating Control Using Macro-cell D. Macro/micro Cell Paging In the conventional network, the whole service area is divided into LAs which are a logical group of cells. In our approach, the LAs are geographically distributed according to BSC/CSS (macro-cell). We are proposing a combined macrocell and micro-cell paging which is depending on the user’s class (slow/fast). From figure 7 and 9, each macro-cell coverage area is one LA. Each user has his own Paging Area (PA), as we propose a per-user position based PA algorithm with more information and intelligent control [6,7] based on user’s speed, last contact cell, etc. [8]. The multiple-step paging scheme has already been proposed in [13], we plan to use it in our architecture and have proposed a technique through which we can determine the cells to be paged easily. Hence the BSC/CSS covering one LA will have a number of records of paging details for the different users. If a user belongs to the fast moving class then the macro-cell based paging scheme may be applied. In general the paging scheme is explained below and shown in figure 9. • Each micro-cell’s position is identified with references to X-Y axis. • R = paging radius and all cells within this radius will be paged. R is dependent on the user’s class (e.g. pedestrian, car, public transport, etc.). For example in Figure 8, P1 is a user in office, P2 is a pedestrian, and P3 is a user driving a car in the city. • When a user is called, the first paging area = PA1, and is given by 8 7 6 5 4 3 2 1 .. .. ... . 0 Y First PA 8 7 (4,6) 6 (8,4) 5 4 3 (16,4) (5,3) . ....... 2 4 6 8 10 12 14 16 R P3 R R P1 P2 2 1 X 0 2 4 6 8 10 12 14 16 X * The shadow area is covered a macro-cell and is equal to one BSC/CSS coverage area ** P1, P2, P3 are the last contact cells of the users Figure 9. Planning for Location and Paging Area A description of the macro/micro-cell based paging algorithm is given in the flow chart of figure 10. The network first checks the user’s class whether the user is fast or slow moving. In the case of fast moving user, the macro-cell paging will be used. If the network realises that user belongs to slow moving class then the intelligent paging algorithm is launched. The information flows of an incoming call for fast moving user are shown in figure 11. GSM uses a time-triggered periodic location updating to trace the MS’s position and status (ON/OFF). In our approach, when the MS is turning off, the MS will send a signalling message to inform the network (BSC, VLR, and HLR) its status so there is no necessity to have periodic location updating to make sure of the MS’s status. This reduces the signalling traffic considerably and can be easily adapted to the proposed architecture. This signalling message can be sent by the macro-cell base station. In the case where MS moves outside the network coverage area or moves into a blind spot, there is no signal between MS and network. This is a critical situation in our approach. During the noncontactable period, if there is an incoming call to the MS, paging will fail and the network will treat the MS as switched off and the following incoming calls will not be paged. Once the MS moves into the coverage area and regains the signals, the MS will send a location updating message to inform network its status/position immediately. between PiBSC and BSC and the PiBSC communicates directly to the LE. Paging Request User's Class Selection Fast Moving User? Yes Page the LA using Macro-cell No Paging the surrounding cells of the last contact cell within the LA according to •R " MS Response? Yes Stop No Paging the remain cells in the LA MS Response? REFERENCES Yes Stop No Paging the whole LA using macro-cell Figure 10. Flow Chart of Paging Control MS Micro-cell BS Macro-cell BS BSC/CSS LE/MSC Incoming call to MS Paging MS using macro-cell (for fast moving class user) Paging MS { Paging response Connection setup OR { IV. CONCLUSION In future mobile network systems, signaling traffic for mobility is predicted to be much higher and complicated. In this paper we have approached this problem by suggesting some enhancements to the proposed UMTS architecture, resulting in reasonable signaling traffic for mobility management. The model has been studied with respect to location updating and paging signaling traffic and the corresponding information flows and control points proposed. However we further need to study and analyse this model to provide estimation on the actual signaling traffic and perform some comparative studies with existing architecture. Paging response Connection setup Figure 11. Information Flows of an Incoming Call for Fast Moving User E. Location Management in Pico-cell Environment In figure 1, a functional entity called Pico-cell BSC (PiBSC) which connects to LE is proposed. The functions of the PiBSC are similar to BSC. The mobility management in the pico-cell environment is similar to the micro-cell environment. F. Handover When the MS moves from a pico-cell to a micro-cell, the forward handover control should be used instead of backward handover control which been used in most of micro-cell and macro-cell cases. This is because there is no connection [1] M. Mouly, M.-B. Pautet, “The GSM System for Mobile Communications”, published by the authors, 1992. [2] T.. Norp, A. J. M. Roovers, “UMTS Integrated with BISDN”, IEEE Comm. Mag., Vol. 32, No. 11, Nov. 1994, pp 60-65. [3] E. Buitenwerf, G. Colombo, H. Mitts, P. Wright, “UMTS: Fixed Network Issues and Design Options”, IEEE Personal Comm., Vol. 2, No. 1, Feb 1995, pp 30-37. [4] K.-H. Chiang, N. Shenoy, J. Asenstorfer, “Intelligent Handover and Location Updating for a Third Generation Mobile Network”, IEEE GLOBECOM’98, Vol.4, Nov. 1998, pp 1963-1968. [5] 3GPP, “Technical Report on the Gateway Location Register”, V1.1.0, 1999-04. [6] G. L. Lyberopoulos, J. G. Markoulidakis, D. V. Polymeros, D. F. Tsirkas, and E. D. Sykas, “Intelligent Paging Strategies for Third Generation Mobile Telecommunication Systems”, IEEE Trans. on Vehicular Technology, Vol. 44, No. 3, Aug. 1995. pp 543-554. [7] N. E. Kruijt, D. Sparreboom, F. C. Schoute, and Prasad, “Location Management Strategies for Cellular Mobile Networks”, Journal of Electronics and Communication Engineering, Vol. 10, No. 2, 1998, pp 64-72. [8] S. Mishra, O. K. Tonguz, “Most Recent Interaction Area and Speed-based Intelligent Paging in PCS”, IEEE VTC’97, Vol .2, May 1997, pp 505-509. [9] I. F. Akyildiz, J. Mcnair, J. S. M. Ho, H. Uzunalioglu, and W. Wang, “Mobility Management