WO2001076096A1

WO2001076096A1 - A packet scheduler and method

Info

Publication number: WO2001076096A1
Application number: PCT/EP2001/003017
Authority: WO
Inventors: Peter Jonathon Legg; Stephen John Barrett; Stephen Hill
Original assignee: Motorola Inc.
Priority date: 2000-03-31
Filing date: 2001-03-19
Publication date: 2001-10-11
Also published as: AU2001239303A1; GB0007848D0; GB2360909B; GB2360909A

Abstract

A scheduler (7) for allocating data packets from a single queue (8) to power bins (9, 10, 11) associated with respective base stations (1, 2, 3) in a cellular communications network facilitates sharing of capacity between base stations and relieves loads on 'hot spots'. By reducing the value of maximum transmit power of an under-utilised base station in the calculation of intercell interference, more packets can be scheduled to an over-loaded cell (9) without having to increase transmit its power whilst still maintaining the target signal to noise ratio at a receiver (4, 5).

Description

A PACKET SCHEDULER AND METHOD

Field of the Invention

This invention relates to a packet scheduler and method therefor, and in particular to a packet scheduler for a cellular communication system for mobile communication.

Background of the Invention

In a cellular communication system remote terminals (typically mobile stations) communicate with a fixed base station. Communication from the mobile stations to the base station is known as uplink and communication from the base station to the mobile stations is known as downlink. The total coverage area of the system is divided into a number of separate cells, each covered by a single base station. The cells are typically geographically distinct with an overlapping coverage area with neighbouring cells.

As a mobile station moves from the coverage area of one cell to the coverage area of another cell, the communication link will change from being between the mobile station and the base station of the first cell, to being between the mobile station and the base station of the second cell. This is known as a handover or handoff.

All base stations are interconnected by a fixed network. This fixed network comprises communication lines, switches, interfaces to other communication networks and various controllers required for operating the network. A call from a mobile station is routed through the fixed network to the destination specific for this call. If the call is between two mobile stations of the same communication system the call will be routed through the fixed network to the base station of the cell in which the other mobile station currently is. A connection is thus established between the two serving cells through the fixed network. Alternatively, if the call is between a mobile station and a telephone connected to the Public Switched Telephone Network (PSTN) the call is routed from the serving base station to the interface between the cellular mobile communication system and the PSTN. It is then routed from the interface to the telephone by the PSTN.

A cellular mobile communication system is allocated a frequency spectrum for the radio communication between the mobile stations and the base stations. This spectrum must be shared between all mobile stations simultaneously using the system.

One method of sharing this spectrum is by a technique known as Code Division Multiple Access (CDMA) . In a Direct Sequence CDMA (DS-CDMA) communication system, the signals are prior to being transmitted, multiplied by a high rate code whereby the signal is spread over a larger frequency spectrum. A narrowband signal is thus spread and transmitted as a wideband signal. At the receiver the original narrowband signal is regenerated by multiplication of the received signal with the same code. A signal spread by use of a different code will at the receiver not be de-spread but will remain a wide band signal. In the receiver the majority of interference caused by interfering signals received in the same frequency spectrum as the wanted signal can thus be removed by filtering. Consequently a plurality of mobile stations can be accommodated in the same wideband spectrum by allocating different codes for different remote terminals. Codes are chosen to minimise the interference caused between mobile stations typically by choosing orthogonal codes when possible. A further description of CDMA communication systems can be found in ^Spread Spectrum CDMA Systems for Wireless Communications', Glisic & Vucetic, Artech house Publishers, 1997, ISBN 0-89006-858-5. Examples of CDMA cellular communication systems are IS 95 standardised in North America and the Universal Mobile Telecommunication System (UMTS) currently under standardisation in Europe.

Traditional traffic in mobile cellular communication systems has been circuit switched voice data where a permanent link is set up between the communicating parties. In the future it is envisaged that data communication will increase substantially and typically the requirements for a remote terminal to transmit data will not be continuous but will be at irregular intervals. Consequently it is inefficient to have a continuous link setup between users and instead a significant increase in packet based data traffic is expected, where the transmitting remote terminal seeks to transmit the data in discrete data packets when necessary. An example of a packet based system is General Packet Radio Service (GPRS) introduced to the Global System for Mobile communication (GSM) . Further details on data packet systems can be found in 'Understanding data communications: from fundamentals to networking, 2^nd ed.', John Wiley publishers, author Gilbert Held, 1997, ISBN 0-471-96820-X.

In a packet based system where a high number of mobile stations may require resources for packet transmissions at unknown and irregular intervals it is important for optimal utilisation of the limited resource to schedule the order and time for transmission of the individual packets. This becomes even more important when different data packets have different requirements with respect to delay, bit error rate etc. Therefore most packet based systems contain schedulers which control when the individual data packets are transmitted and therefore share the available resource, whether time-slots in a TDMA system or power and codes in a CDMA system. An introduction to schedulers can be found in 'Service discipline for guaranteed performance service in packet-switching networks' , Hui Zhang, Proceedings of the IEEE, volume 83, no. 10, October 1995.

However, known schedulers have been optimised for different environments than CDMA systems. For example, scheduling algorithms used for GPRS are optimised for a Time Division Multiple Access (TDMA) system and therefore not optimal for CDMA systems where codes and power must be shared.

This invention aims to provide a packet scheduler suitable for use in a CDMA system and which enables an efficient sharing of network resource. Summary of the Invention

According to a first aspect, the present invention comprises apparatus for scheduling queued data packets for transmission between a plurality of base stations, each serving a respective cell in a cellular communications network, and a plurality of mobile stations located therein, the apparatus including,

means for assigning an associated power bin to each cell into which at least some of the queued data packets are to be placed,

means for setting a maximum power level value for each power bin,

means for identifying a first power bin associated with an overloaded cell and second power bin associated with an underloaded cell,

means for reducing the maximum power level of the second power bin, thereby reducing intercell interference in the overloaded cell,

and means for allocating further data packets from the queue to the first power bin.

According to a second aspect, the present invention comprises a method for scheduling queued data packets for transmission between a plurality of base stations, each serving a respective cell in a cellular communications network, and a plurality of mobile stations located therein, the method including the steps of,

assigning an associated power bin to each cell into which at least some of the queued data packets are to be placed,

setting a maximum power level value for each power bin, identifying a first power bin associated with an overloaded cell and second power bin associated with an under-loaded cell,

reducing the maximum power level of the second power bin, thereby reducing mtercell interference in the overloaded cell,

and allocating further data packets from the queue to the first power bin.

In one embodiment, a maximum power level value (P_omax) is set for each power bin.

A total power level of each power bin resulting from the data packets which are destined therefor is calculated during a scheduling period, the calculation taking account of mtercell interference based on transmissions from each base station being set at a level of P₀max. A first power bin whose total power level is substantially equal to P_emax is identified and a second power bin whose total power level is less than P₀max is also identified. The maximum power level of the second power bin is set to a new, lower value P_omax (new) .

A re-calculation of the total power level of the first power bin is performed in the cell based on downlink transmissions associated with the second power bin being set at a level of

P_omax (new) . Subsequently further data packets can be allocated from the queue to the first power bin. In an alternative embodiment, the calculations and re-calculations of total power levels of power bins are performed based on uplink transmissions from mobile stations operating within the communications system.

Thus the invention allows re-distribution of network resource from an under-utilised base station to an overloaded one.

In a preferred embodiment the apparatus comprises a scheduler which is located within a radio network controller and all the steps of the method are performed by the scheduler. Brief Description of the Drawings

Some embodiments of the invention will now be described, by way of example only, with reference to the drawings of which:

Fig.l is a schematic block diagram of a cellular communications system operating in accordance with the invention;

Figs. 2, 3 & 4 are schematic diagrams illustrating the operation of a packet scheduler in accordance with the invention;

And Fig. 5 is a flow chart illustrating operation of a specific embodiment of the invention.

Detailed Description of the Preferred Embodiments

The following embodiments are described within the context of, the current approach for the standardisation of UMTS but it will be apparent that the invention is not limited to this application.

FIG 1 shows a schematic diagram of an embodiment of a CDMA communication system in accordance with an embodiment of the invention. The communication system has a number of base stations 1, 2, 3 each covering a geographical area and thereby defining a cell. A number of mobile stations 4, 5 are associated with the communication system and communicate to each other or to other systems via the base stations 1, 2, 3.

The base stations are connected to a common Radio Network Controller (RNC) 6. The RNC further provides gateways to other communication systems such as the fixed public telephone system (not shown) and contains a scheduler 7.

Each of the mobile stations 4, 5 have independent communication needs and communicate by use of data packets. The mobile stations 4, 5 may require different services and can for example be Internet browsers, telephones or data terminals. Each remote terminal may also request different services at different times.

The resource requirement for each individual mobile station may vary significantly over time so that a mobile station may sometimes require no transmissions and at other times require long transmissions at high data rate. The resource requirement for each mobile station from the communication network can thus vary significantly and in order to ensure that the available network capacity is used optimally an efficient scheduling of packets for the different mobile stations is required. This task is performed by the scheduler 7.

In a frame based communication system such as UMTS the communication is divided into discrete time intervals or frames and the communication resource is allocated on a per frame basis. In UMTS, packets to be transmitted are scheduled during one time frame and transmitted during a subsequent frame.

In CDMA systems the total data throughput is affected by the interference at a mobile station's receiver caused by transmissions from base stations other than its serving base station. A successful sharing of the available resources of the system requires a knowledge of these interference levels so that the signal to noise ratio for each connection is appropriate for the desired quality of service (QoS) . In order to perform an optimal allocation, transmissions from non-serving base stations must therefore be taken into account and scheduling in all cells must be done simultaneously.

Consider sharing the power in a single cell. The scheduler's task is to divide the available transmit power of the base station (BS) between data packets removed in sequence from the queue. For each transmission the signal to noise ratio (SNR) at the destination mobile station must equal that necessary to achieve the agreed QoS. The SNR at mobile station i is given by the following equation: SNR_t = ₍₁₎ aP₀(l - ,) /L, + /_{ιntercellιI} + P_lhermal

P₀ = BS transmit power φ_] = fraction of power assigned to user i

L, = path loss from BS to user i a = loss of orthogonality factor

I_ιnterceiι_,ι ⁼ intercell interference at user i

^_thermal ⁼ thermal noise power added before the receiver

To determine the transmit power to a mobile, equation (1) must be solved for φi .

_ SNR, {P_thL, + I_{merce!l l}L, + czP₀ ) P^l + SNR^)

Φ i Is equivalent to one or more data packets

The path loss to a mobile station, the loss of orthogonality factor and the thermal noise power can all be measured by known techniques and reported to the scheduler.

P_ocan be set by the network operator and so can a target SNR.

Intercell interference can be calculated from knowledge of transmit power of the other base stations in the network and from the path loss.

Thus, by substituting the relevant values into equation (2), a value for fractional power of every mobile station operative in a cell can be found. The total power transmitted (during a frame) from a base station will then be the sum of the individual Φ, values. (∑Φ, ) multiplied by P_o . However, the total transmit power of any base station is limited for practical reasons (eg. Depending on the capabilities of the power amplifiers) . Say that this limit is the same for all base stations and has a value P_omax. Hence, the number of packets that can be transmitted by a particular base station during a frame is similarly limited.

It will be appreciated, however, from inspection of equation (2) that if the intercell interference (for a particular cell) should decrease, then for a fixed target SNR, Φ, will also decrease and hence, so will ∑Φ₍ . This means that more packets can be scheduled to and transmitted from the base station serving that cell without exceeding P₀max.

Fig. 2 shows a schematic representation of the constituent parts of the scheduler 7. A single queue of packets 8 is allocated as necessary to one of three power bins 9, 10, 11 which relate respectively to base stations 1, 2 and 3. Packets labelled 1 are destined for transmission by base station 1, packets 2 are for base station 2 and packets 3 are for base station 3. A maximum value for the power bins 9, 10, 11 is set at P_omax.

A power bin represents the total available power at a base station, for either reception or transmission, which is to be shared by i users. The height of a data packet shown in the power bins of Figs. 2, 3 and 4 represents the amount of power required by a user per frame/scheduling period.

Fig. 3 shows the situation part-way through a frame/scheduling period. The scheduler has scheduled and calculated the values for Φ, for a number of packets destined for base stations 1, 2 3 and begun to fill up the power bins 9, 10, 11 accordingly. In order to calculate a value for intercell interference, the value for power transmitted by every base station is taken to be

P_omax. In Fig. 3 the power bin 9 for base station 1 is completely full with some packets in the queue 8 still remaining to be scheduled. On the other hand, power bins 10 and 11 are not full yet no more packets are required to be scheduled for base stations 2 and 3 during this frame. Thus base stations 2 and 3 have spare resource which could be utilised by the overloaded base station 1.

Without the benefit of the present invention, when a power bin of a certain base station is filled (it reaches P₀max) then further packets directed this base station must wait until the next scheduling period/frame.

At the end of the scheduling, base stations 2 and 3 will transmit at a lower power because there are not sufficient packets for the cells to reach P_omax (no packets queued for the cells remain in the single queue.) Therefore, there is some unused capacity in the network.

Furthermore, as packets are skipped once the power bin for a base station is filled, coverage of hot spots (which occur when there is a lot of high priority traffic in a single cell) is prevented.

The present invention can utilise the unused capacity and further accommodate hot spots as follows.

P₀max for power bins 10 and 11 are lowered to a new value P_omax (new) [chosen by one of several methods to be discussed below] . The scheduler 7 then re-calculates the values for Φ, for the base station 1 based on values for intercell interference determined from the adjusted values P₀max (new) . The sum of the new Φ₍ values will now have a value of less than unity. This means that more packets (remaining in queue 8) can be scheduled to base station 1 into power bin 9) yet still keeping total transmit power within the P_omax limit.

The re-configuration of the power bins at the end of the recalculation procedure is shown in Fig. 4. At the end of the frame, base station 1 transmits all the queued packets for its cell at a total power of P_σmax. Base stations 2 and 3 transit at a total power level of P₂and P₃ both being less than P_omax

(new). In Fig. 4, all packets in queue 8 have been successfully scheduled during the one frame.

Hence to accommodate hot spots, the invention permits the borrowing of downlink capacity from neighbouring cells.

One option for selecting which cells should have their maximum power reduced, is to choose those cells which generate the greatest interference to mobile stations lying within the hot spot cell.

In a further embodiment, the scheduler can be configured to redistribute any unused capacity within the network based on a re-calculation of the Φ₍ values (and of ∑Φ, ) in one or more neighbouring cells or all cells in the network.

A hot-spot is defined as a cell in which there is high priority traffic queued which cannot all be serviced if the network capacity is shared equally amongst all cells (P_omax equal in all cells). Hot-spots have limited lifespans, otherwise the network operator would deploy greater infrastructure in these areas to handle the greater demand. Examples include cells around a football ground, this cell becomes 'hot' at half-time and the end of play as people phone home; a stretch of motorway when there has been an accident and a queue of cars develops. It is of considerable benefit to an operator if hot-spots can be handled efficiently by careful radio resource management over a number of cells.

To handle a hot-spot it is necessary to be able to allocate packets from the single queue 8 which, conventionally is ordered in terms of transmission priority, wi thout skipping pa ckets . This is possible by borrowing capacity from other cells, in the manner described above with reference to Fig. 4. In order to choose the appropriate values for P_omax (new) for any underused base station or base stations, the scheduler may keep reducing P₀max of the appropriate base station (s) and carry out the re-calculation of Φ, values procedure until enough space has been created in the over-subscribed power bin to accommodate the queued packets. It is preferable to put a lower limit on

P₀max (new) so that the capacity of the reduced power bins is not disadvantaged, le there should be enough resource left available in the reduced bins so that their own packets which are already scheduled can be transmitted.

Different strategies may be employed in determining how to arrive at the best P_omax (new) value. For example, it is possible to reduce P₀max in cells for which the path loss from the base station to the mobile stations m the overloaded cell is smallest. This will require a smaller reduction than for cells which are 'further' (in the path loss sense) from the overloaded cell. In other words, it is best to borrow capacity from the neighbouring cells.

Alternatively, it is possible to reduce P₀max in proportion to the remaining available transmit power (designated Pr in Fig. 3) in the cells which have the greatest spare capacity.

As an alternative to this technique, the buffered traffic in queue 8 for each cell may be determined by the scheduler 7 prior to filling the power bins. Using this information the scheduler 7 can more accurately determine which cells will have spare capacity and therefore should have their P₀max re-defined.

Making the changes to a minimum number of base station P_omax values minimises complexity of the scheduling process. The process is simpler if only P_omax m one cell needs to be changed. Using equation (2) the scheduler 7 forms an equation for the sum of φi (including an additional packet [s] ) in the overloaded cell, and solves this for a P_omax (new) value of the single cell whose maximum power has been adjusted. If the scheduler were to adjust the P₀max in a number of cells simultaneously the mathematics is more complex and an iterative approach as outlined in Figure 5 (see below) is more appropriate. Following the accommodation of the additional packet (s) the scheduler recalculates the φi values allocated m cells other than overloaded cell (since the mtercell interference may have changed) . Alternatively, it may be less complex to recalculate the φi values for packets in a cell only when the scheduler is arranging for the addition of a packet to that same cell. It is also possible to calculate the new φi values for all packets in the one cell in one calculation.

Optionally, a safety margin may be introduced so that the sum of the desired recalculated φi values is ∑φi : = 1-margin.

Figure 5 shows a flow chart illustrating the adjustment of the maximum transmit power (P₀max) m one or more cells to accommodate the allocation of an additional packet to a cell "A" which is a "hot spot" . All steps are performed in the scheduler 7.

In Fig. 5, in step 12, P₀max is reduced for one or more neighbouring base stations. In step 13, the scheduler 7 recalculates ∑Φ, including an additional data packet in cell A.

In step 14 it tests the value of ∑Φ, for cell A to see if it lies within a pre-defined margin of value 1. If it does, then the scheduler 7 adds the additional data packet to cell A' s power bin (step 15) . (It may also go on to calculate Φ₍ values all other cells [step 6] ) . If not, then it checks to see if further changes in P_omax for neighbours is possible (step 16) .

If so, the process repeats. If not, the process ends. Employing the method for dealing with hot spots as described above has a limitation in that the overall capacity of the system can be less than if the capacity is shared equally amongst all cells. This compromise may be addressed by considering a hybrid method which employs a priority obedience factor, gamma. In such a hybrid method the extent to which maximum transmit power may be reduced below the nominal value of P ax is controlled by gamma. When gamma is set to 1, this is the hot spot method as described above and the transmit power P_omax

(new) may be set to zero if necessary. In addition, gamma values of between zero and 1 can be used where gamma controls either the maximum reduction of the transmit power below Pmax (i.e. this is a hard limit), or aspects of a probability density function for the transmit power reduction, for example, the standard deviation (i.e. this is a soft limit). In this configuration there is both an attempt to follow the strict priority order and also to achieve good utilisation of the network.

In an alternative embodiment with reference again to Fig. 4, the P_omax (new) values of the under-utilised power bins (eg bins 10 and 11) are set at P₂ and P₃ respectively. The recalculation of

Φ, values and further allocation of queued packets to power bin

9 is then performed as above. In general terms, in this embodiment which permits sharing of capacity among cells, the scheduler freezes the value of P_omax in a power bin which is unfilled, to the level reached at some pre-determined point during the frame and then considers re-allocation of the further queued packets to cells whose power bins are already full. A procedure for this embodiment is as follows. At the predetermined point in a frame, the scheduler sets P_omax (new) for each power bin to the filled power level in each bin. For power bins thus having P₀max (new) less than P_omax, then these levels are effectively "frozen" . Next, the scheduler 7 re-calculates Φ, values in cells which have P_omax (new) equal to P_omax (ie full bins). The scheduler then continues scheduling and adds more packets where possible to the power bins not "frozen". While the preferred embodiments have been described with respect to scheduling data packets on a downlink, it is to be understood that similar principles can be employed on an uplink, thus allowing system capacity sharing between cells. In this case, a power bin now represents received power at a base station, and this includes both useful power and intercell interference. The total uplink power resource in a cell (P_totai₎ is a constant based on the characteristics of the UMTS uplink. The capacity of the uplink is limited by the need to meet the signal to noise ratio for every transmission (SNR) . P_totai s set to such a level to ensure a high sustainable throughput.

At the beginning of each frame, a maximum useable power for resource scheduling in any cell (P_maX) is set to an appropriate level based on the mean ratio of the received power at a base station from the mobile station it serves to the received power from all other mobile stations. (This ratio is set a approximately 1:0.6). This ensures that there is adequate headroom within any cell to allow for the neighbouring cells to be fully loaded and thus use their power resource fully.

The invention as described above in the downlink case can be similarly applied in the uplink case in order to transfer power resource between cells whilst maintaining the Signal to Noise Ratio (SNR) and therefore the desired QoS. Again, the principle employed here is to reduce the P_raax in cells neighbouring a highly loaded cell. This means that while P_otai remains constant, the split between usable uplink power P_ma at the receiver of a highly loaded base station and intercell interference is altered by lowering the amount of allowed intercell interference. This reduces the interference expected in the highly loaded cell, allowing P_raay to be increased, and thus additional packets to be scheduled.

In a preferred embodiment, if power resource is transferred from one cell to another, then extra interference will be generated and will impact on the SNR at the cells which power has been taken from. Sufficient headroom (60% of the additional power requirement) must be available at all adjacent cells to enable the extra interference to be handled without reducing the SNR of already scheduled packets. Thus P_max must be reduced by say 60% of the additional power requirement at all adjacent cells.

P_πrnx can now be further reduced at adjacent cells to create the additional headroom for scheduling in the hot-spot cell. It is assumed that the interference from any cell is distributed equally between all neighbouring cells. The reduction of P_max by 1 unit in an adjacent cell will result in a reduction in interference and therefore an an available increase in P_raax by 0.6 /N in the hot spot cell, where N is the number of cells neighbouring the hot spot cell. The preferred embodiment implements the distribution of P_max reduction by further reducing P_max in at least one of the adjacent cells (for example, the most lightly loaded neighbour) until the required P_max increase in the hot spot cell is reached. Alternatively, P_max of one adjacent cell is reduced until the level of another adjacent cell is reached, then both are reduced at the same rate until the required P_max increase in the hot spot cell is reached.

When the above conditions are met and the required extra headroom is found, additional packet (s) can be scheduled; otherwise they must be skipped.

Claims

1. Apparatus for scheduling queued data packets for transmission between a plurality of base stations, each serving a respective cell in a cellular communications network, and a plurality of mobile stations located therein, the apparatus including, means for assigning an associated power bin to each cell into which at least some of the queued data packets are to be placed, means for setting a maximum power level value for each power bin, means for idendifying a first power bin associated with an overloaded cell and second power bin associated with an underloaded cell, means for reducing the maximum power level of the second power bin, thereby reducing intercell interference in the overloaded cell, and means for allocating further data packets from the queue to the first power bin.

2. A method for scheduling queued data packets for transmission between a plurality of base stations, each serving a respective cell in a cellular communications system, and a plurality of mobile stations located therein, the method including the steps of,

(a) assigning an associated power bin to each cell into which at least some of the queued data packets are to be placed,

(b) setting a maximum power level value for each power bin,

(c) identifying a first power bin associated with an overloaded cell and second power bin associated with an under-loaded cell,

(d) reducing the maximum power level of the second power bin, thereby reducing intercell interference in the overloaded cell, (e) and allocating further data packets from the queue to the first power bm.

3. A method according to claim 2 in which step (b) includes the step of setting a maximum power level value P_omax for each power bm; further including the step of calculating during a scheduling period, a total power level of each power b resulting from the data packets which are destined therefor, the calculation taking account of mtercell interference based on transmissions from each base station being set at a level of

P₀max, and in which step (c) includes the step of identifying the first power bm as one whose total power level is substantially equal to P_omax and the second power bin and one whose total and power level is less than P₀max; and in which step (d) includes the step of resetting the maximum power level of the second power bm to a new, lower value P_omax (new) , and including the further step of re-calculatmg the total power level of the first power bm based on downlink transmissions in the case associated with the second power bm being set at a level of P_omax (new) .

4. A method according to claim 3 in which the cell associated with the second power bm is one which contributes a high level of interference, compared with other cells m the system, to mobile stations located m the cell associated with the first power bm.

5. A method according to claim 3 and including the further step of repeating the steps of resetting, re-calculatmg and allocating until a pre-determmed limit for P_o (max) new is reached.

6. A method according to claim 3 including the further step of re-calculatmg the total power level of the power bm associated with at least one cell in the system other than the first cell after the re-settmg step.

7. A method according to claim 3 in which P_omax (new) is set to a level dependent upon the magnitude of the difference between P_omax and the total power level of the second power bin.

8. A method according to claim 4 in which P_omax (new) is set to the total power level for the second power bin.

9. A method according to claim 2 in which step (b) includes the step of setting a maximum power level value Pmax for each power bin where Pmax is set to a level based on a signal to noise ratio of a signal received at each base station, and in which step (d) includes the steps of reducing Pmax in proportion to the degree of overload of the overloaded cell in cells adjacent to the overloaded cell and further reducing Pmax of the power bin of at least one adjacent cell.

10. A computer program product comprising a medium on or in which is recorded a program which, when executed in a computer- controlled system, will perform the method of any of claims 2-9.