WO2016156547A1

WO2016156547A1 - Tracking area configuration

Info

Publication number: WO2016156547A1
Application number: PCT/EP2016/057166
Authority: WO
Inventors: Fabrice Saffre; Mark Shackleton
Original assignee: British Telecommunications Public Limited Company
Priority date: 2015-03-31
Filing date: 2016-03-31
Publication date: 2016-10-06
Also published as: GB201714724D0; GB2550826B; GB2550826A

Abstract

A method of configuring tracking areas in a network comprising a first cell associated with a first tracking area, a second cell associated with a second tracking area, and a boundary between the first tracking area and the second tracking area, the method comprising: calculating a first value based on the number of handovers between the first cell and the second cell; calculating a second value based on the number of handovers between the second cell and another cell in the second tracking area; comparing the first value and the second value; and using the result of the comparison in a decision to change the boundary between the first tracking area and the second tracking area.

Description

TRACKING AREA CONFIGURATION

This invention relates to apparatus, systems and methods in the area of wireless or mobile data transmissions, specifically in respect of the configuration of radio cell groupings like tracking areas, which are of particular but not exclusive application in Long Term Evolution (LTE) networks.

In the LTE standard, a tracking area (TA) is a site comprising at least one geographically defined cell which are serviced by base stations or access points (AP). More usefully, they comprise a plurality of cells which is grouped into a TA zone or area which defines a single destination with a TA identity for traffic sent to mobile terminals or other "User Entities" (UEs, sometimes also referred to as "User Equipment" devices) such as mobile telephones or laptops. APs may broadcast the TA code (TAC) which is a TA's identity as one of the overhead parameters, so that it is not essential to keep track of the precise location of a particular UE as long as it remains within the TA it is associated with. This helps to simplify the routing process of transmissions to a destination UE located within a designated TA. In an ideal situation, the shape, size and location of TAs reflect the presence and dominant movement patterns of UEs within it so that the number of handovers crossing TA boundaries can be reduced or minimised. This can help to reduce the number of update messages from UEs associated with a change of TA as well as reducing the amount of paging that is needed to locate UEs. For instance, cells covering a major transportation axis (e.g. a main road or a train line) are preferably configured to form an "elongated" TA along the length the road which UEs travel along. On the other hand, cells in a TA providing coverage in a busy hub (e.g. shopping centre or a train station) could be formed into a more "spherical" clustered topology. The organisation of cells into tracking areas having an ideal configuration is a typical optimisation problem, the complexity of which increases with scale.

It should be understood that TAs can be organised into hierarchical or aggregate groupings that could themselves effectively be treated as if they were a single TA, an example of which are TA Lists in LTE networks. Hence a grouping such as a TA List (under the LTE standard) can comprise one or more atomic TAs. In the description, the term "Tracking Area" (TA) covers either a single atomic TA or such a grouping of atomic TAs as appropriate. The groupings described below may be generalised and considered as groupings of transceiver units; in which one or more cell transceiver units serviced by an AP are grouped into TAs, and one or more TA transceiver units are grouped into TA Lists. It should also be noted that for some wireless technologies or deployment scenarios it may be more feasible to dynamically reconfigure groupings of atomic TAs in real-time than it is to change the configuration of individual atomic TAs.

Configuration methods as generally known are described in e.g. S. Modarres Razavi (201 1 ) "Tracking Area Planning in Cellular Networks (Optimization and Performance Evaluation)", Linkoping Studies in Science and Technology Licentiate Thesis No. 1473. This describes an offline optimisation method which may be used when movement patterns are known and the number of cells involved is relatively small, and generates a TA configuration comprising a fixed allocation of cells to static tracking areas. This method is not easily scalable and performs less well with situations where the movement patterns of UEs or the precise location of APs is unknown, resulting in proposals for sub-optimal TA configurations.

Furthermore, this method cannot re-shape TA configurations to take into account changes in UE movement and activity in real-time. The process of resetting and re-drawing of some or all of the TA borders may consume such significant time and resources, that the cost penalty may discourage TA reconfiguration to such an extent that the sub-optimal performance at certain times of day is simply accepted, or else tolerance levels for sub-par performance set at a very high level for e.g. situations where UE traffic patterns slowly change over longer periods of time. Hence, a fixed allocation of cells to static TAs (even if they were optimal on initialisation ("Day One") or at an earlier stage) may over time be or become permanently, periodically or occasionally sub-optimal.

A method which can be operated at run time to dynamically reconfigure TAs is set out in US8804566. This disclosure describes a central network entity such as a Mobility

Management Entity (MME) making a decision to reconfigure a TA area by enlarging it, in which the decision is based on a determination that the load (in the form of tracking area update messages) at a border region or sector of the TA reaches a given threshold. There is no indication of what this threshold is based on. The MME can then reconfigure the affected TA with the aim of reducing the load ~ this is done so that the affected sector is no longer a border sector, by adding sectors from another TA so enlarging the TA area. This approach could successfully reduce the load to a particular TA by bringing sectors into it; however this method of "poaching" sectors from neighbouring TA could result in an even greater number of TA update messages being generated, e.g. where there is a significant number of handovers between the sectors which have now been moved into the new TA, and sectors which remain in the previous TA from which sectors have been moved. A further possible consequence is that this method could generate a significant amount of sector movement between TAs, as each TA seeks to bring sectors into their boundaries: this has the opposite effect to the intention to reduce the number of messages being generated. Moreover, it may be expected that at an extreme, the eventual outcome would be a single or universal "TA" containing all sectors.

It would desirable to re-configure TAs in real-time in a dynamic manner to respond to changes in UE movements and patterns over time, in an empirically robust way which optimises the overall performance of the network.

According to a first aspect of the invention, there is provided a method of configuring tracking areas in a network comprising a first cell associated with a first tracking area, and a second cell associated with a second tracking area, a boundary between the first tracking area and the second tracking area, the method comprising

calculating a first value based on the number of handovers between the first cell and the second cell,

calculating a second value based on the number of handovers between the second cell and another cell in the second tracking area,

comparing the first value and the second value, and

using the result of the comparison in a decision to change the boundary between the first tracking area and the second tracking area. Preferred embodiments enable the generation of a clear basis on which to make a decision about changing a boundary between two tracking areas which comprise cells which are handing over connections to each other. A working premise of approaches of such embodiments is that if the calculated values assessing the strength of association that a cell has with its current TA is weaker than the strength of association with another TA, then the cell should join or be associated with that other TA, by being re-designated to it. In certain circumstances, the strength of association of the cell with its own TA is high enough so that instead of being recruited to the other TA, the neighbouring cell associated with the other TA is recruited to join the first cell's TA. The values which are derived for comparison of the relative strengths of remaining associated with a current TA and of moving to a new TA, are based on the number or level of handovers. Strength of association with a current TA is based on the level of intra-TA handovers, i.e. within the same TA. Strength of association with another TA with which there is a handover is based on the level of inter-TA handovers, i.e. out of the current TA.

In preferred embodiments, the result of the comparison of values representing the strength of association with the current TA versus another TA can result in decisions to change the TA boundary comprising recruitment out (where a cell is "lost" to the other TA), recruitment in (where a cell of the other TA is re-designated to the cell's TA), or where no action is taken so that the status quo is maintained. In certain applications, a cell can be re-designated to a new TA which may generated in a manner which will be detailed below.

In a preferred application, a cell stress value indicative of the level of cross-TA boundary handovers can be generated, which can be used e.g. to identify those cells in most need of re-designation. An aim is to reduce the level of cross-TA border handovers as far as possible, which would have the effect of reducing traffic and management overhead due to a drop in e.g. the number of tracking area update messages, as well as potentially improve quality levels for UEs and customers. According to a second aspect of the invention, there is provided a method of operating a network in which tracking areas are being configured according to the first aspect.

Methods according to preferred embodiments can be used to operate a network which can help enable reconfiguration of tracking areas on a regular basis, in response to changes in user entity movement patterns and routes, or even on a pre-emptive basis in anticipation of such changes.

According to a further aspect of the invention, there is provided a network comprising a first cell associated with a first tracking area, a second cell associated with a second tracking area, a boundary between the first tracking area and the second tracking area, and a controller arranged to configure the first tracking area and the second tracking area by

comparing the first value and the second value, and

using the result of the comparison in a decision to change the boundary between the first tracking area and the second tracking area. Systems, methods and apparatus embodying the present invention will now be described by way of example only, with reference to the following drawings, wherein: Figure 1 is a topographical view of a network comprising TAs and the UE traffic therein; Figure 2 is a topographical view of the network with a different UE traffic and TA patterns; Figure 3 is a flowchart of the steps for generating a "cell stress" value for a cell;

Figure 4 is a flowchart of the steps for changing the boundary of a TA; and

Figure 5 is a flowchart of the steps for the cell recruitment process.

Figure 1 is a view of network 2 (which might be operated under the LTE standard) comprising a plurality (here, 13) of cells (one of which is shown with reference numeral 4) each served by an AP. The topographical layout shown in this example comprises a road (R) which separates two buildings (B1 and B2). Each cell overlaps with at least one other cell within the network area. Users carrying UEs (one of which is shown with reference numeral 6) travel within the geographic area in the direction depicted by an arrow, and their connections to the network are maintained by way of handovers between cells in the known manner. In this exemplary network, the geographic network area is divided into three TA regions "A", "B" and "C" as generally denoted by the boundary lines (one of which is shown with reference numeral 8) between each TA, as well as by the identification of each individual cell as shown, so there are five cells that belong to TA "A", four cells within TA "B", and four cells within TA "C". Cells would typically advertise their associated TA by broadcasting the TA identity value that may be numeric, but in the exemplary network this is denoted "A", "B" or "C" for ease of reference. The movement and trajectories of the UEs in this example are such that most of them stay within the bounds of a particular TA. Such a substantially circular pattern could suggest that the UEs are moving around in a shopping centre or office area within buildings (B1 and B2), for example. In such a situation, the TAs are operating at or near optimum and the network is, at least along this metric, in a stable state if there are no cross-TA boundary handovers. Where there are some UEs straying out of a TA and crossing a TA boundary, the connection is passed from a "B" cell to a cell in TA "C", resulting in a cross-TA boundary handover. At this point, the UE might transmit a TA update message upon detection that it has moved out of the previous TA and into a new one. As is known, handovers are processed between source and destination APs but when a TA update message is generated it is notified to an MME which keeps track of which TA the notifying UE resides within, to enable subsequent paging of the UE to be carried out when necessary. Depending on policy adopted by the network operator, the number and frequency of such cross-boundary handovers, it may be decided that the TA boundaries should be redrawn e.g. where a particularly high number of TA update handovers generate operational overhead exceeding a pre-determined threshold level or where quality of service or experience (QoS or QoE) requirements are not met. Figure 2 depicts the same network topography but with very different UE movement and trajectories and patterns. In particular, there are a significant number of UEs (one of which is shown with reference numeral 6') which are travelling along the road (R) in a linear trajectory. This pattern is more suggestive of movement of UEs which are travelling to, from or through the network area, for example during peak or rush hours. An ideal TA configuration for this kind of UE traffic looks significantly different from that proposed for the traffic patterns of Figure 1 . Here, there are four TA areas, "A", "B", "C" and "D", in which TA areas, "A", "B" and "C" continue to serve UEs whose traffic patterns are non-linear. TA "D" serves almost exclusively UEs on a linear trajectory on the road. If the UE pattern changes over time (e.g. time of day on a working day) from that shown in Figure 1 to that of Figure 2, a static TA configuration would be sub-optimal at least for part of that time. So for example, UE travelling along the road could potentially pass through each of the TAs "A", "B" and "C" in turn if the TA configuration remained as shown in Figure 1 . A TA "D" of Figure 2, on the other hand, is more disruptive than useful for UE traffic of the type shown in Figure 1 , as UEs could potentially stray into TA "D" and quickly leave as it heads back into its TA.

In a TA configuration for a network which has been decided on "Day One" using e.g.

conventional "offline" methods resulting in the TAs depicted in Figures 1 or 2 (as the case may be), there is a clear benefit in being able to "re-draw the map" to accommodate changes in UE traffic patterns. Preferred embodiments address the need for the generation of an optimal TA configuration of a network in real-time by using a "cell recruitment" process incorporating a determination whether the advantage of "re-zoning" a cell to join a

neighbouring TA to which it has been handing over UE connections, outweighs the downside of doing so. The advantage of re-designating a cell to another TA lies in a reduction of (or prevention of an increase in) the number of cross-border handovers, with its attendant penalties on operation and performance.

In the present embodiment, a routine may be implemented in a network Operation,

Administration and Management (OAM) element located at (or in connection with) the MME, which is configured to initiate a TA configuration assessment process. This process can be triggered by the detection (during a particular monitoring period) of a high number (e.g.

exceeding a threshold level) of cross-TA border handovers or TA update messages, or it can be set to run at a certain time (e.g. at 4.00 pm on working days, or where there is a scheduled sporting event on a certain day), or else it can be set to run periodically (e.g. every 2 minutes). As with other methods (e.g. "offline" optimisation), the present method may be based on maintaining a record of the number of handovers between cells, with a record of those handovers which have been with cells in the same TA, and those that cross the border between two TAs. With this initial information, preferred embodiments may derive a measure of "cell stress" which is obtained for a particular cell (e.g. cell "i"), which is indicative of the number (or fraction) of handovers in which the cell was involved, that cross the TA boundary. If there are none, then the stress value is zero (minimum), if all the handovers (over a certain monitoring period) are to/from a cell belonging to another TA, then the stress value is one (maximum). All the cells in the particular TA may be assessed for their stress level: this can be carried out by arranging all the cells within the TA from most-stressed (first) to least- stressed (last). Here, each cell is assessed and a recruitment process carried out for each cell in turn, until a "stop-condition" is met. A possible default stop-condition may be finding a cell whose stress level is zero. When the recruitment process is triggered, a cell (cell "i") collates information about its neighbouring cells (e.g. cell "j") with which it has exchanged UE connections. These neighbours may or may not be within the same tracking area as the cell collating the information. In one application, the neighbours (cells "j", "ji", "j₂", etc.) may be arranged into an ordered list starting with the cell with which it has experienced the most handovers and ending with the one with which it has experienced the fewest. When it reaches the first of its neighbours that belongs to a different tracking area, an assessment of the relative advantage of designating a cell to another TA is carried out.

This process involves two cells: cell "i" associated with a TA, and a cell "j" associated with a different TA. In this part of the process, the effect of re-designating the TA of a particular cell is compared to the effect of maintaining the status quo where the two cells remain in their respective current TAs. One way of considering this is to use the terminology of comparing the "attack" and "defence" values of cell "i" and cell "j" respectively. If cell "i" has a high attack value or factor relative to the defence value of its handover partner cell "j" associated with another TA, this means that cell "i" has an association with its TA which is stronger than the association that cell "j" has with its own TA. In such a situation cell "i" could not join the TA of cell "j"; conversely cell "j" could be recruited to join to TA of cell "i". If on the other hand, the attack value of cell "i" is lower relative to the defence value of cell "j", its association with its current TA is weaker, and it is worth considering that cell "i" should be recruited to join cell "j" in its TA so that future handovers to each other will no longer cross a TA boundary.

Applications of preferred embodiments can further and optionally involve a comparison also being made in the "opposite direction" in the form of a "counter-attack" which compares the attack value of cell "j" and the defence value of cell "i". This provides a form of confirmation of the decision as to whether to re-designate cell "i" or not. The decision to re-draw the TA boundary can also be influenced by other policies implemented by the network operator in the form of e.g. an inertia factor in which very small differences in the respective attack and defence values are ignored owing to the cost of changing the TA boundary, resulting in the maintenance of the status quo.

Even if the TA boundary is unchanged after the recruitment action (when seemingly nothing has happened), useful information about the relative (or lack of) advantage to be gained from the recruitment of cells into a new TA is generated. This allows for a decision to be made on the basis of actual hard data in real-time. Preferred embodiments can be used simply to generate such data to help justify or explain TA re-designations which have been made perhaps on the basis of other methods. In particular, this approach may be contrasted with known methods in which cells are re-designated without empirical knowledge of the usefulness of doing so at a particular (or any) time, in spite of the cost penalty in terms of added overhead and the like in effecting the TA re-designation.

A variant of the process would be to calculate many of the key variables such as cell stress and handover strengths (based on frequency of handovers), but deliberately to effect no changes to TA composition. The resulting variables could then be used directly or to formulate metrics to indicate the efficacy of TA configurations, possibly generated by other means. Examples of metrics include total stress across a region comprising several TAs, or a global ratio across such a region of inter-TA versus intra-TA handovers.

It will also be appreciated that variations whereby cells at a busy "border crossing" between TAs attempt to reduce their stress level through a reorganisation of local TA membership (i.e. by recruiting or being recruited) can be realised. For instance, making the resolution of a recruitment attempt a nonlinear function of the attack and defence strengths, or including a saturation element making a TA "weaker" when it approaches limit size would quantitatively change system dynamics, which could be beneficial or detrimental to overall performance depending on network characteristics and mobile devices movement patterns.

It may be observed in general that recruitment of a candidate cell into a TA tends to benefit that TA experiencing cross-TA boundary handovers (as it has to contend with fewer cross- boundary handovers upon recruitment of that partner cell). This is even though the recruitment might significantly penalise the TA losing the cell, as its cross-border handover level might increase with the loss of the cell. Conventional real-time TA re-designation methods provide that TA borders are redrawn without any consideration taken of the effect on the network as a whole. The present approach on the other hand determines the relative merits of re-designating a cell in a way which benefits the network as a whole as opposed to just benefitting one TA within the network. In particular, the optional "counter-attack" part of the process refines the final outcome of the initial comparison of the cells' relative strengths of association, as this can help balance out the initial outcome of the first comparison step by providing an even fuller picture of the advantage (or otherwise) of re-designating the TA of one of either cell. To this end, the "counter-attack" step could be included in the process even if cell "i" initially "won" the contest in having an attack factor higher than the defence factor of cell "j". As will be appreciated, various other permutations of the steps of the method can be applied according to need and usefulness in the particular case.

In summary, the above describes the process of TA reconfiguration in respect of a single cell pair "i" and "j". The outcome is either that one of the cells joins the other's TA, or else the status quo is maintained following an objective determination in real-time that TA re- designation for either cell will not result in a reduction (or prevention of an increase) in the number of cross-TA boundary handovers. As alluded to above, such a method is different from the "trial-and-error" process of other optimisation heuristics such as genetic algorithms, which may be based on guesswork about possible outcomes to set thresholds and to determine when a reconfiguration exercise should be run.

Preferred embodiments are described in terms of the respective cell's "attack" and "defence" values, but it will be appreciated that this is just a way to articulate the concept of obtaining an empirically verifiable value on the decision to maintain the status quo, or to designate a cell to another TA (and if so, in which direction) by comparing the relative benefits of different courses of action. In more general terms, the process starts with the premise that cross-TA handovers are to be avoided or minimised (so as to avoid or minimise cell stress), and the process then moves to generate values for each possibility. While preferred applications can be thought of as involving contests of strength between two cells in different TAs which have participated in a handover with each other, simpler applications may consider the question of whether a cell should stay put with its current TA, or whether it should move to a

neighbouring TA with which it has participated in a handover. In such a scenario, values are generated for comparing the advantage of staying versus the advantage of being designated, or comparing the advantage of staying versus the disadvantage of staying. For example, a value based on the number (or rate) of handovers within a cell's own TA (during a particular monitoring period) can be compared with a value based on the number (or rate) of handovers to cells outside the designated TA. In other words, different implementations could cover an external contest of strength between two cells, or obtaining a ratio of cell i's own "outward" handovers (which are directed out of its TA) versus "inward" handovers (which are directed within its TA). As all applications include an element of comparison, however, this approach enables the generation of an objective, potentially-quantifiable value of a particular decision, or at the least a notion of the relative benefit of taking one course of action over the other(s). Advantageously, this is a metric which takes into account the benefit to all or part of the network, and not to just one or both TAs in question. The above process, and a method to reconfigure TA boundaries, will be described in greater detail with reference to the flowcharts of Figures 3 to 5.

Figure 3 depicts the determination of the "stress" level of cell "i" (or equivalently in this context, its AP) resulting from its participation in a cross-TA border handover. In this implementation of the process, it is determined if a handover Hy with cell "j" took place within the same TA. Starting at step S1 , an AP or a cell "i" detects a UE handover in the usual way to or from cell "j". In step S2, it is determined or checked whether cell "j" is on the neighbour's list of cell "i". If it's not already on, the list is updated by adding cell "j" in step S3, and the handover count Hy, between cells "i" and "j" and the number of handovers which cell "i" has participated in is incremented in step S4. The next part of the process serves to determine the stress suffered by cell "i" in the form of numbers of handovers made with cells outside its TA. In step S5, it is determined if the handover Hy crossed the TA border of the TA of cell "i". Cell "i" holds a record of the cell with which it has made the most number of handovers Hi,_max, and at step S6, a comparison is made to see if this number exceeds the number of exchanges made with cell "i". If the number of handovers to cell "j" does not exceed this, then the record of internal handovers H,_n within the TA is updated in step S8. If however the number of handovers between cells "i" and "j" are the highest within the TA, the records are updated in step S7 and S8. The output of this sub-routine is a Hm value indicating the level of handovers within the TA of cell "i". To calculate the stress level of cell "i", a value is need for the level of handovers to any cells outside of the TA of cell "i". At step S9, the record of the number of handovers Hout outside the TA of cell "i" is updated. A stress level S, of cell "i" is generated in step S10 using the formula

wherein optional parameter a allows a network operator to tune the output, e.g. to emphasise particularly high Hout levels. A parameter value larger than 1 amplifies the result. Upon establishment of the stress level for cell "i", the routine ends (step S1 1 ).

The stress value of each cell "i" in the same TA is used in the next part of the process set out in Figure 4 which describes the cell recruitment process at a TA-wide level. This part of the process decides if a particular cell "i" should be re-designated to another TA (resulting in a change of the TA border). The process commences at step S20, which noted above can be automatically triggered by an event (e.g. detection of a high number of handovers or TA update events) or on a periodic basis. In one implementation, all the member cells of a particular tracking area are sorted into a list according to their stress levels, e.g. from the highest to the lowest stress ^"levels, per step S21 . It will be appreciated that other ways of organising the list are possible, and that ordering the members of the list is not essential to their processing.

The process then loops through steps S22 to S26 until all the cells in the list have been considered. In this routine, every cell "i" of the TA which has participated in a cross-border handover (so that it has a positive level of stress associated with it) is identified (step S23). A recruitment action by cell "i" commences at step S24, which is a sub-routine described in greater detail in the flowchart of Figure 5. There can be three outcomes of the recruitment action: the particular cell "i" is lost to the neighbouring TA to which cell "j" belongs, cell "j" is recruited to join the TA of cell "i", or a "stalemate" results in neither cell being re-designated to a new TA. In the first two cases, the TA boundary has changed (step S25) owing to the re- designation of cell "i" or "j". If cell "j" is recruited to join the TA of cell "i", the total cell numbers of the respective TAs are incremented/decremented by one (step S26, wherein a "mod" is a change in TA membership). As an optional step in S27, the process can be arranged to determine if the cell count in the TA has reached a predetermined threshold value, such value being e.g. a desirable maximum number of cells in the TA to ensure that it does not become "too big", as will be explained below. Furthermore, it is known that paging traffic load tends to increase as TAs become bigger: a limit is ideally placed on the number of changes or the number of cells by which the TA is allowed to grow to ensure that any efficiencies gained by the cell joining the TA are not lost in the overall measure. If the threshold in step S27 is reached, the recruitment process can terminate (step S29) immediately as the TA has reached a maximum size and cannot take on any further cell. If not, the process can return to step S28 and be repeated if other cells on the list remain to be tested. As will be appreciated, variations are possible, e.g. configuring the process to continue on condition that a member cell "i" must leave the TA before another cell "j" can be recruited to join.

If on the other hand it is found at S25 that the cell "i" remains associated with its TA for reasons that will be described below in connection with Figure 5, then the process loops to assess the next cell "i" in the TA if there remain any which have yet to be processed (loops of steps S22 to S28) until there are no more cells in the TA which have participated in a cross- boundary handover, at which point the recruitment process ends (step S29). As mentioned above, the flowchart of Figure 5 describes the recruitment action subroutine which starts (step S30) when triggered in step S24 in the TA. Essentially, this process looks at the strength of the links with the TA it is associated with in the first instance, in terms of the level of handovers made with cells within the TA. If what dominates are the handovers with cells outside of its own TA, the system may decide that this cell "i" is better associated with another TA. By using an approach based on the relative "attack" and "defence" strengths of a pair of handover partner cells to recruit or to be recruited to and from each cell's TAs, methods according to preferred embodiments may be based on robustly objective criteria to make the decision of when and whether to redraw TA boundaries, which can be carried out in real-time and in response to cell, TA, and customer demand.

In this example, the neighbouring handover partner cells to a particular cell "i" who are not associated with the same TA as cell "i" are identified and sorted in order of the number of handovers made with cell "i" in step S31 . The list can be in the order of the most frequent handover partner to the least frequent (i.e. by decreasing Hy value), based on a snapshot view of the network condition at the time of the recruitment action. Starting with the first partner cell "j" on the list (here, the one with the most number of handovers with cell "i", step S32), an attack value is calculated at step S33 for cell "i" vis-a-vis the defence value for cell "j", representing the "reasons" for cell "i" retaining its association with its current TA.

There are a number of ways of obtaining and expressing the relative strength values of the cells, of which the following is an example as set out in step S33:

Ai = ¾ X

Dj— Hj max X Tj

where A is the attack strength of cell "i", Hy is the number of handovers between cell "i" and neighbour "j", 7) is the number of cells in the tracking area that cell "i" belongs to, Dj^"\s the defence strength of cell "j", and ¾_ma is the highest number of handovers recorded by cell "j" within its own tracking area. 7/ is the number of cells in the tracking area that cell "j" belongs to. The inclusion of the number of handovers with the candidate recruit cell "j" H_ii} and of the "strongest" link within the TA of cell "j" H_J>max into the attack and defence strengths respectively serves to compare the benefits of changing TAs versus those of maintaining the status quo. For instance, by boosting its defence strength (e.g. in terms of an increased level of handovers within the TA), a high value of H_jiTnax makes it less likely that a cell will move out of a TA if the result of such a move will likely be more handovers across TA boundaries (because the record shows many handovers within the original TA of cell "j"). Essentially, an attack value is premised on the level of cross-TA border handovers, while a defence value is based on the level of handovers which remain within the particular cell's own TA. Other ways of characterising the level of handovers could be used in place of the value of Η_}ιΤηαχ, for example an average rather than maximum value. This would change the overall system dynamics but could still achieve the overall goals.

An optional parameter β allows a network operator optionally to tune the output. A value larger than 1 amplifies the result. In preferred step S34, the obtained At result is multiplied by a random number or parameter of e.g. between 0.5 and 1 .5 to avoid the output from being deterministic, to yield At'. The attack factor of cell "i" At' is then compared in step S35 with the defence factor of cell "j" Dj lo determine if the strength of the connection of cell "i" (in terms of handover levels) is greater than the strength of the connection to cell "j" (in terms of handover levels). In the situation where A (or Αή is determined to be greater than Dj, cell "j" is recruited to and joins the TA of cell "i" in step S36, at which point the process can terminate. The process is completed for this particular cell "i" in step S37, and the process returns to step S25 in Figure 4.

If in the contest between cells "i" and "j" (A/'> Dj) in step S35 is false, then a number of next steps are possible: for example cell "i" could simply maintain its designation to its TA as the attacking cell is not strong enough to recruit the target cell to join its TA, at which point the process goes straight to step S37. A network provider could also configure the process so that the process goes straight to step S42 and S37 on account of weakness of cell "i" relative to the defence value of cell "j"; this could be useful in certain circumstances, e.g. cells which are obviously suffering a significant level of stress in an environment of fast-changing UE movement patterns so that "doing nothing" (in which neither cell is re-designated) is unlikely to be a reasonable option. More preferably however, the process moves to the next phase after step S35 to a "counter-attack" process symmetric to the above, and again tests the relative strengths of the cells. This time, the attack value Aj O cell "j" is compared against the defence value Z¾ of cell "i". If A, is greater than Di, then cell "i" is recruited to and joins the TA of cell "j" at which point the process can terminate. Again, if A, is not greater than Di, then cell "j" could simply remain associated with its own TA. It will be appreciated that a number of variations and additional steps can be used to enhance the outcome or to suit the network provider's particular requirements in different circumstances. For example, if one cell has won the recruitment contest in a first round, and the other in the subsequent round, the respective strength values could be weighted to determine the outcome of which, if either, cell is re-designated to obtain a change in the TA boundary between the cells. Essentially however, methods according to preferred embodiments may be directed to discovering that a cell is participating in so many handovers with a "foreign" cell belonging to another TA causing so much cross-border handover stress that on objectively-verifiable criteria, it would benefit the network as a whole for the two cells in question to be designated to the same TA. Whether this is achieved by the cell in question leaving its TA, or recruiting the other cell to join its TA is determined by the process in step S33.

Preferably however, the process moves to a mirror or symmetric series of steps in which the attack factor of cell "j" is calculated and compared against the defence factor of cell "i", as set out in steps S38 to S41 . Again, an optional but preferred step S39 is included to multiply the obtained Aj with a random number or parameter, similar to that carried out in step S34. As noted above, this guards against a deterministic result, which can help prevent stalemate outcomes in the strength contest between the cells. In the event of either a successful attack or successful counter-attack the recruitment subroutine ends (step S37), otherwise it proceeds to a loop (restarting at step S40) to similarly consider each further neighbour in the aforementioned list until a successful attack or counter-attack happens, or until the list is exhausted. When this recruitment sub-routine ends at step S37, the process continues at step S25.

Advantageously, the distributed nature of this recruitment process allows cells that are under the most stress in terms of cross-border handovers to reduce their stress level by targeting specific handover partner cells for recruitment based on their own perception of the local environment. The TA reorganisation process results from local information gathering (as illustrated by the flowchart of Figure 3) and recruitment events (Figure 5) which supported by individual cells/APs, none of which use or require global knowledge (apart from the size of the TA). This is a key difference from a centralised and/or "brute force" optimisation method of the prior art that uses information about the whole system to identify an efficient solution. The advantages of distributed problem-solving of this nature are evident, in that less

"concentrated" computational power is required as they converge towards an approximation of the optimal solution progressively, through many small and simple steps. Furthermore, this solution scales well, especially in contrast with known optimisation approaches in which problems quickly become practically unsolvable as system size increases, because the complexity of the configuration space (i.e. the number of possible configurations) tends to grow exponentially with size. Distributed solutions tend to be more robust and can adapt to unforeseen situations, because they effectively decompose complex problems into a (very) large number of simple local ones (here: whether to transfer one cell from one TA to the other and in which "direction"). Advantageously, the process can run in real-time, enabling a timely response to surges and changes in UE movement patterns within a TA and more generally, in the network.

The process as described in the flowcharts could be deployed in a decentralised manner as described above or it could be implemented in a centralised way if so desired. The centralised aspects may be implemented in an OAM element co-located or in communication with the MME. Since the flowchart of Figure 4 entails knowledge at the level of TAs, this would typically be centralised. The part of the process shown in the flowchart of Figure 4 could be implemented in a different OAM element co-located or in communications contact with an AP, however it could be implemented in the centralised OAM element if necessary. The part of the process shown in the flowchart of Figure 5 could be implemented in either of the aforementioned OAM elements; if it were implemented in the decentralised AP OAM element then since the size of TAs is required then this would need to be obtained from the centralised OAM or from the MME.

The approach has a "rich becomes richer" aspect to it in that that smaller TAs or clusters of cells tend to be absorbed into larger ones since the strength of an attack is a function of the size of the attacker's TA. This aspect is desirable when seeking to reduce the number of small TAs which could lead to more frequent cross-TA handovers. However, one

consequence of operating this method is that, if left unchecked, a large TA would eventually absorb most if not all cells in the network into a single "universal" tracking area containing all cells in its stable state. This of course would remove all the benefits and indeed the very purpose of having TAs. An optimal TA size depends on the particular circumstances and various factors, but generally this would not comprise all the cells of the network. The size of TAs is known to affect a trade-off between the level of paging traffic and the level of TA update message traffic. Small TAs help reduce the level of broadcast paging traffic, but can mean that UEs frequently cross into other TAs and so increase the number of TA update events. Conversely, large TAs result in more paging, but mean that UEs change TA less often and so reduces the number of TA update messages. One means to help address this trade-off is to specify a predetermined size limit for a TA. As a variant, such a size limit could be dynamically adjusted in real-time according to, for example, the observed level of paging traffic or the observed level of TA updates, or both. As another variant, there could be more than one size limit applicable to different regions such as urban regions versus rural regions. According to implementations of the method, cells can also be "expelled" or "ejected" from the TA if the TA size exceeds its predetermined size limit. Cells are ejected until the TA falls back below its given limit. These ejected cells can be newly- recruited cells, or cells previously associated with the TA prior to the start of the recruitment exercise. In one implementation, cells which have the fewest handovers with any other member of the TA are ejected first, as this is likely to be least "stressful" both for the ejected cell and for the remaining members of the TA. The ejected cells can join a neighbouring TA, or in a particular implementation, the ejected cells can form a new TA. This can occur where the neighbouring TA is itself at its size limit, or else it can be planned or caused to occur by the network operator. This process can commence by the formation of a "single cell" TA, which might grow in size as other ejected cells join the TA during the recruitment process. As is known in LTE networks, the MME manages TAs, so it would be necessary for an OAM element in such a network to register the new TAs with the MME and they would be assigned new or suitable non-conflicting TA identities in the form of TA codes. During the next recruitment round, other cells (which have been ejected, or having lost to an "attack" by the single cell) may join the new TA; alternatively the single cell might itself be absorbed into another TA, having lost to an attack by another cell associated with that TA. A simpler approach to keep TA size within a predetermined limit involves making a check - prior to an intended recruitment - that doing so would not lead to the TA exceeding the maximum size. Step S27 of Figure 4 is an example of how this may be implemented. Preferably, the size limit can be defined in "fuzzy" (rather than in strict, absolute) terms, so as to deal with a possible issue of recruitment paralysis as no TA having reached the maximum size will be able to initiate the TA reconfiguration process.

As noted earlier the fact that the strength of an attack increases as a function of the size of the attacker's TA leads to the "rich becomes richer" behaviour resulting in a single "universal" TA. Another way to avoid this extreme behaviour is to make the strength of the attack (and defence) be functions of the TA's closeness to a desired target size, rather than always increasing with its absolute size. In this case size(TA) in Figure 5 steps S33 and S38 could be replaced with a function that rewards closeness to the target size, for example,

1/(1 +abs(Target_size - size(TA)))^Y, where Target_size and / are parameters set by the operator. The Target_size could then be set to a TA size that the operator knows to strike a good balance of the aforementioned trade-off between paging load and TA update traffic load, / values could be set experimentally or by inspection of the function itself or its impact on system dynamics, e.g. in simulation. Another example would be to use the size(TA) multiplier in Fig 5 steps S33 and S38 whilst this size is less than the desired target size, but to replace that multiplier with (Target_size + (Target_size - size(TA))) otherwise. These changes would preferably be applied to both the calculation of A/ and the calculation of Dj. However a variant would be to use different multiplier functions for each of A/ and Dj as a way for example to maintain a strong defence but not to seek a strong attack. The introduction of a desired target size would encourage the formation of TAs of that approximate size for large regions that might otherwise grow into a single large TA with a size that would otherwise only be curtailed by some possibly specified maximum size. The dynamic and continuous self-(re-)organisation process makes tracking areas capable of adapting their size and shape to changing movement patterns. In the exemplary network shown in Figures 1 and 2, the method can operate to reconfigure the TA layout shown in each drawing to and from the other. In particular, implementations of the method can form specific TA shapes configured to usefully mirror or approximate UE trajectories to avoid or to minimise cross-TA border handovers. As previously noted, a new TA can also be formed where this is the best response to the kind of UE movement and patterns in the particular case. Here, the drawings respectively represent UE movement and patterns (as depicted by the arrows associated with UEs 6) within a network topographical area at two times of the same day.

Figure 1 represents UE traffic conditions during a time when users are in the office buildings (B1 , B2) e.g. at 3.00 pm, which accounts for the "circular" shape of UE movement. Figure 2 represents UE traffic conditions during a time when users are leaving the office e.g. at 6.00 pm, hence the significant amount of linear movement along the road "R" as depicted by the arrows of the UEs 6'. As previously discussed, the TA shapes for each UE movement pattern are very different. Using methods according to preferred embodiments, it is possible to reconfigure the layout of Figure 1 to that in Figure 2 (and back again after the rush hour) by using the cell recruitment method described above - either in response to e.g. an increasing number of TA update messages being generated following cross-border handovers, or else as a planned measure during each week day or working day. The TA configuration of Figure 1 with its three TA regions "A", "B" and "C" can, by be reshaped completely by the inclusion of a new, fourth, TA region "D" shown in Figure 2, which has been generated by one or more cells previously belonging to another TA. In addition to the spontaneous creation of a new TA in the manner described above, a network provider can configure the system to trigger the generation of a new TA where e.g. such traffic patterns are known ahead of time. As shown, the cells of new TA "D" have been strong enough to win from TAs "A", "B" and "C" two, two and one cells respectively. A UE travelling through TA region "D" thus optimally requires no cross-border handovers to maintain a connection while traveling through this TA, which greatly minimises the network operational overhead and potentially improves the customer experience in respect of the UE. Region "D" has an elongated shape as the UEs travelling along the road "R" cause handovers which cross TA boundaries as they pass through each of the regions "A", "B" and "C". The proposed process causes cells previously associated with TA regions "A", "B" and "C" to challenge each other through the above-described strength challenges or contests between them in the manner described above, so that each TA suffering some kind of cell stress would be subject to the recruitment process in turn. Thus the cells self-organise in a seemingly organic way to generate and populate the new region in a way that follows UE traffic along the road, which passes through the previous three TA regions. Any cell which is not "sufficiently needed" (in terms of sufficient numbers of handovers with a TA "D" cell) along the UE trajectory will not win the strength contest and will hence not join cell "D". In short, TAs are capable of intelligently self-organising or self-configuring or -reconfiguring when the frequency of cross-boundary handovers between TAs increases, and again when the frequency of cross-boundary handovers between TAs decreases so that any default TA configuration can be resume e.g. once rush hour traffic has cleared. In this way, applications and implementations of preferred embodiments can be completely directed by UE movement, although network operators can include into the process any intermediary steps or controls as might be seen fit to obtain any desirable outcome.

The methods and configurations as described above and in the drawings are for ease of description only and not meant to restrict the methods or apparatus to a particular arrangement or order or process in use. It will also be apparent to the skilled person that various sequences and permutations on the methods and apparatus described are possible within the scope of this invention as disclosed.

Claims

1 . A method of configuring tracking areas in a cellular communication network comprising a plurality of cells, the cells being reconfigurably grouped into a plurality of reconfigurable tracking areas such that each tracking area has one or more cells associated therewith, the tracking areas being operationally separated by boundaries, the method comprising:

calculating a cell stress value in respect of one or more cells, the cell stress value for a particular cell being indicative of a level of handovers between said cell and a cell associated with a different tracking area; and

in respect of a first cell associated with a first tracking area and a second cell associated with a second tracking area, at least one of the first and second cells having been selected in dependence on the cell stress value calculation:

- calculating a first value based on the number of handovers between the first cell and the second cell during a monitoring period;

- calculating a second value based on the number of handovers between the second cell and another cell associated with the second tracking area during the monitoring period;

- comparing the first value and the second value; and

- using the result of the comparison in a decision to change the boundary between the first tracking area and the second tracking area.

2. A method according to claim 1 wherein the step of using the result of the

comparison comprises re-designation of the first cell to be associated with the second tracking area, or re-designation of the second cell to be associated with the first tracking area.

3. A method according to claim 2 further comprising implementing a threshold on the number of cells that may be associated with the first and/or second tracking area so as to prevent re-designation of the second cell to the first tracking area and/or to prevent re- designation of the first cell to the second tracking area if the threshold number is reached.

4. A method according to claim 1 wherein the step of using the result of the

comparison comprises creation of a new tracking area and re-designation of the first cell and/or the second cell to be associated with the new tracking area.

5. A method according to claim 1 wherein the step of using the result of the

comparison comprises a decision against changing the boundary between the first tracking area and the second tracking area.

6. A method according to any preceding claim further comprising

calculating a third value based on the number of handovers between the second cell and the first cell during a monitoring period;

calculating a fourth value based on the number of handovers between the first cell and another cell associated with the first tracking area during the monitoring period;

comparing the third value and the fourth value; and

using the result of the comparison in the decision to change the boundary between the first tracking area and the second tracking area.

7. A method of operating a network in which tracking areas are configured according to any preceding claim.

8. A method of operating a network according to claim 7, wherein a user entity is travelling along a user entity route within the network, the method comprising responding to a change in the user entity route to make a decision to change the boundary between a first tracking area and a second tracking area thereby changing the configuration of the first tracking area and/or the second tracking area.

9. A method according to claim 8 wherein the response to a change in the user entity route comprises responding to receipt of a pre-determined level of tracking area update messages.

10. A method according to claim 9 wherein the decision to change the boundary between a first tracking area and a second tracking area is made pre-emptively of a change in the user entity route.

1 1 . A cellular communication network comprising a plurality of cells, the cells being reconfigurably grouped into a plurality of reconfigurable tracking areas such that each tracking area has one or more cells associated therewith, the tracking areas being

operationally separated by boundaries, the network having a control module arranged to configure the first tracking area and the second tracking area by performing a method according to any of the preceding claims.