METHOD AND APPARATUS FOR DETERMINING A FREQUENCY RE-USE PLAN IN A CELLULAR COMMUNICATIONS SYSTEM
This invention relates to determination of a frequency re-use plan in order to optimise the performance of a cellular communications system configured to carry both voice traffic and packet data traffic.
The invention has particular applicability to GPRS (general packet radio services) and EDGE (enhanced data rates for GSM evolution) based on the GSM system (global system for mobile communications).
Cellular communications systems generally include a mobile services switching centre (MSC) coupled to a public switched telephone network (PSTN), a plurality of base stations (BS) and radio telephone subscriber units often referred to as mobile stations. Usually, several base stations are under the control of a base station controller (BSC) which in turn communicates directly by land-line or microwave link with an MSC. Several base station controllers may report to one MSC. An operations and maintenance centre (OMC) is connected to the MSC and comprises a centralised facility which supports the day to day management of the cellular system. It also facilitates network engineering and planning.
Each of the plurality of base stations generally defines a geographic region or cell proximate to each base station to produce coverage areas. The coverage area of any given cell may partially overlap with that of one or more of its neighbours. Cell sizes range typically from 200 metres in diameter in urban areas to 120 kilometres in rural areas. Each base station comprises the radio frequency components and the antenna for communicating with the mobile stations. The communication link from a base station to a mobile station is called the downlink. Conversely, the communication link from the mobile station to the base station is called the uplink.
Multiple access techniques permit the simultaneous transmissions from several mobile stations to a single base station. The GSM system uses time division multiple access (TDMA), in which a communications channel consists of a time slot
in a period train of time intervals over the same frequency. Each mobile station is allocated one specific time slot for communication with a base station in a repeating time frame.
Standard GSM has a total of 124 frequencies available for use in a network. Most network providers are unlikely to be able to use all of these frequencies and are generally assigned a small sub-set of the 124. Typically, a network provider may be assigned 48 frequencies for providing coverage over a large area such as Great Britain, for example. As the maximum cell size is approximately 120 kilometres in diameter, 48 frequencies would not be able to cover the whole of Great Britain. To overcome this limitation, the network provider must re-use the same frequencies over and over again in a frequency re-use pattern. (Frequency re-use means the use of the same carrier frequency in different cells). When planning the frequency re-use pattern, the network must take into account how often to use the same frequencies and must determine how close together the cells are, otherwise co- channel and/or adjacent channel interference may occur. The network provider will also take into account the nature of the area to be covered. This may range from a densely populated city (high frequency re-use, small cells, high capacity) to a sparsely populated rural expanse (large cells, low re-use, low capacity).
Co-channel interference occurs when RF carriers of the same frequency are transmitting in close proximity to each other and the transmission from one RF carrier interferes with the other RF carrier. Adjacent channel interference occurs when RF source of a nearby frequency interferes with the RF carrier. Interference levels within a given cell may be monitored and in a typical cellular communications system, each mobile station performs real time measurements and reports them back to its serving base station. For example, in the GSM system, each mobile station measures the downlink signal quality i.e. the bit error rate, and the downlink signal strength of a signal received from its serving base station in whose cell it currently resides and also the downlink signal strength from neighbouring base stations serving neighbouring cells. All these so-called measurement reports are periodically reported back to the serving base station. For any given cell, it is possible to measure (at each mobile station) the interference levels caused by the
environment during a call. These levels are usually expressed as the ratio of the received signal level from the wanted source (carrier level, C) to the interference received level (interference level, I), or C/l and are expressed in dB. The distribution of C/l ratios of a given cell depends on the locations of the mobile stations in communication with the base station serving the cell and on the locations of the interfering sources, i.e. on cellular planning and frequency re-use.
The goal of cellular planning is to choose cell sites and system parameters in order to economically provide continuous coverage and to support the required traffic density.
Known methods of frequency planning use measurement reports which are analysed to generate an ideal frequency plan. An intermediate step in this process is the development of carrier-to-interference (C/l) matrices which indicate the degrees of interference which would occur between pairs of cells if they were to operate with co-channel or adjacent channel frequencies. Such a method which checks for interference between pairs of mobile stations and base stations in the system is described in US-A-4736453. Therein is disclosed a matrix which displays such results as are well known in the art. A simplified example of a C/l matrix is shown in Figure 1. As seen in Figure 1 , the entries in the matrix indicate the relative impact, defined as a penalty value (or performance metric) if any pairs of the cells A- D have co-channel frequencies. For example, the relative impact on cell A if cell C is made co-channel would be 5, where the 5 the percentage proportion of measurement reports which indicate that the received signal strength of a neighbour cell is greater than, for example, 15dB below the serving cell signal level. Thus, 5% of the measurement reports collected in cell A report cell C as being stronger than 15 dB below the level of cell A.
Currently frequency re-use planning tools have been found to give good results for systems carrying voice traffic (such as GSM) but such methods are not necessarily optimum for a system which supports both circuit switched voice traffic and packet switched data traffic.
GPRS sends packets of information to and from mobile stations using internet protocol and has been developed to complement existing circuit switched services such as voice and SMS (short message service). GPRS is sub-layer within the GSM network and is always connected. However, subscribers are only charged for the amount of data transmitted rather than call time. Information to be transmitted over GPRS is divided into packets, each being labelled with a header and then sent into the GPRS layer of the GSM network. A single message can travel to its destination in several packets by various routes and also may be mixed with packets of data from other messages. Hence, more than one user can occupy a single radio channel at a time, so maximising throughput (bits per second) and so spreading the cost of the resource between several users.
In contrast, in circuit switched voice calls, a specific physical link is reserved from caller to callee which no-one else can use for the duration of the call.
Whereas perceived quality of a voice call may be an appropriate criterion to use in planning a frequency re-use pattern in the circuit switched networks, it may not necessarily give optimum results for a system supporting packet switched transmissions as well. Further, no existing frequency planning tools account for queuing delays and transmission delays. Thus, network operators need to deploy frequency re-use plans which optimise the services for both types of user, i.e. voice and packet data, so as to maximise the return on their investment. Failure to satisfy the needs of data or voice users will lead to customer dissatisfaction and ultimately lost revenue.
As GPRS operates over existing GSM networks, it is subject to the same frequency constraints. GPRS mobile stations will have the capability to use up to 8 time slots on a radio carrier and in so doing achieve high data throughput rates. For some internet protocol applications, the higher throughput rates achievable with GPRS mobile stations equate to high quality of service. The probability of a GPRS mobile station being assigned multiple time slots is dependant upon the availability of sufficient radio resources. Consequently, in periods of traffic congestion, GPRS mobile stations are unlikely to be assigned multiple time slots. The quality of service
experienced by GPRS data users will also be related to the radio characteristics and queuing delay suffered by data packets traversing the communications system. Thus, the task of the network provider is to achieve the quality of service requirements within operating constraints. This task is further complicated by the requirement of network providers to meet the service requirements of the existing GSM subscribers, a group that will typically be responsible for a high proportion of business revenues.
The present invention consists of, in a first aspect, a method of determining a frequency re-use plan in a cellular communications network including a plurality of base stations each having a cell associated therewith and each cell serving a plurality of mobile stations, the method including the steps of;
(a) setting an initial frequency re-use plan for the network,
(b) setting a target throughput value for every one of said mobile stations,
(c) collecting measurement reports and network data for each cell,
(d) analysing said measurement reports and network data for each cell,
(e) from said analysis, calculating predicted throughput values for each mobile station,
(f) calculating performance metrics for all cells related to the differences between the target throughput value and the predicted throughput values,
(g) summing the calculated performance metrics, (h) re-setting the initial frequency re-use plan,
(i) and repeating steps (c) to (h) until the sum of the calculated performance metrics reaches a minimum value.
In a second aspect, the present invention consists of apparatus for determining a frequency re-use plan in a cellular communications network including a plurality of base stations each having a cell associated therewith and each cell serving a plurality of mobile stations, the apparatus including;
means for setting an initial frequency re-use plan for the network, means for setting a target throughput value for every one of said mobile stations,
means for collecting measurement reports and network data for each cell, means for analysing said measurement reports and network data for each cell and for calculating predicted throughput values for each mobile station, means for calculating performance metrics for all cells, related to the differences between the target throughput value and the predicted throughput values, and for summing the calculated performance metrics, and means for resetting the initial frequency re-use plan and re-calculating the performance metrics until the sum of the calculated performance metrics reaches a minimum value.
According to a further aspect of the invention, there is provided a computer program product comprising a medium on or in which is recorded a program which, when executed in a computer system, will perform the methods recited herein.
Throughput is defined as the inverse of a session delay for a given traffic input (in bits). A session delay may include transmission delays and queuing delays. A traffic input value (no. of bits) will be a constant value for any given frequency plan. Session delays will vary from one frequency plan to another with a minimum session delay being optimum.
Optionally, the performance metrics may be modified to include a second metric, derived, in accordance with known methods, from voice quality measurements. Such methods rely on minimising the frame erasure rate. This second metric, once calculated for each cell, is then added to the performance metric derived from the predicted throughput values to give a composite performance metric. The composite performance metrics for all cells are then summed and minimised.
Some embodiments of the invention will now be described, by way of example only, with reference to the drawings of which;
Figure 1 is a table of penalty values associated with a frequency re-use plan as is known in the art, and
Figure 2 is a schematic block diagram of a communications system adapted to determine a frequency re-use plan in accordance with the invention.
Figure 2 shows a communication system adapted to support GPRS communications. Several base stations 1-4 provide areas of coverage i.e. 5-8 for a multiplicity of mobile stations 9-16 which are distributed throughout the cells 5-8. The base stations are controlled by base station controllers 17-18 which in turn are linked to a mobile switching centre 19. The MSC 19 is connected to a PSTN 20 and an operations and maintenance centre (OMC) 21. An intelligent optimisation system block (IOS) 22 is linked to the OMC 21 and to the base station controllers (BSC) 17, 18.
The system of Figure 2 is capable of supporting GSM calls and GPRS data transmission in accordance with known procedures. Additionally, the IOS block 22 is adapted to generate an optimum frequency re-use plan, in accordance with the invention, based on measurement reports from the mobile stations 9-16 and on network data provided by the base stations 1-4, base station controllers 17, 18 and OMC 21.
In operation, an initial frequency re-use plan is set by the IOS 22 and notified to the OMC 21 which commands the base stations 1-4 via their controllers 17, 18 to set their frequencies accordingly. Also, a target throughput value for the mobile station 9-16 is decided upon by the IOS 22.
Whilst in a call, each mobile station 9-16 monitors the received signal levels from its serving base station (for example mobile station 9 is served by base station 1 ) and the received signal levels from its neighbouring base stations (base stations 2, 3, 4 in the case of the mobile station 9). These measurements are reported to the serving base station. The measurement reports are then collected by the IOS block 22 via the base station controllers 17, 18 for further analysis. Other network data, to be described below, is also collected by the IOS block 22 from the OMC 21 and base controllers 17, 18 and used to generate an optimum frequency re-use plan.
The IOS block 22 calculates a predicted data throughput value for each mobile station 9-16 in each cell for a given frequency re-use plan.
The first step in the procedure is to extract Measurement Report (MR) data from cells in the system. Using the MR data, a prediction of the Carrier to Interference (C/l) ratio between a server cell and its interfering, neighbour cells is carried out. This is done in the IOS block 22 employing known techniques.
The next stage in the procedure is for the IOS block 22 to convert the predicted C/l values into Radio Link Control (RLC) block error rates (BLER). In order to perform this conversion, in this preferred embodiment, the following data is employed, ( such data being supplied to the IOS block 22 by the appropriate system elements, ie. the BSC's 17. 19 and the OMC 21 ):
(i) Coding Scheme Switching Points - these are the received radio signal levels (RRSL) at which a mobile station is instructed to change to a different coding scheme (CS) for RLC block reception (or transmission), (ii) C/l Look Up Table - this look up table allows the predicted C/l values to be converted into RLC BLER values. The BLER selected is dependent on the coding scheme switching points that have been deployed.
Given the coding scheme switching points and the C/l distribution extracted from the Measurement Report Data, the CS distribution in the server cell under consideration can be determined.
Next, given the BLER distribution in the server cell and the associated CS distribution, a mobile station data throughput is predicted. In this embodiment , derivation of the throughput employs the following inputs: (i) "lambda": Temporary Block Flow (TBF) arrival rate - (or call arrival rate) this is determined from system statistics supplied by the base stations 1-4. (e.g. Number of downlink seizures for GPRS communications ) This is a measure of the degree of congestion in the server cell under consideration.
(ii) "c": Effective number of GPRS radio resources - this is used to determine the GPRS capacity in the server cell under consideration. (ie. no. of available time-slots). This information is supplied by the OMC 21. This gives an indication of queue size.
(iii) "mu": TBF size distribution -this gives an indication of the queuing characteristics of the server cell under consideration, ie. the number of successful calls having a certain "byte size." This information is provided by the base stations 1-4 and OMC 21.
(iv) "pm": Multi slot distribution - this distribution relates to the spread of GPRS mobile station "classes" in the server cell under consideration. (ie. how many time slots a mobile station is capable of using- different mobile station classes are capable of utilising different combinations of multiples of Uplink or Downlink timeslots). In the preferred embodiment, this information is supplied by the OMC 21 from channel request information.
(v) Queuing Delay range - this is a fixed, pre-determined time period which specifies the range over which the probability of queuing delay is determined. It is set by the IOS block 22.
The throughput values in the server cell in question will vary from one mobile station to another and there will be a range of values which is captured (in the IOS block 22) using a probability distribution. The probability distribution for the throughput, TQ, can be derived from throughput distribution Look Up Tables.These tables are derived from network data supplied by the system elements of Fig. 2 and from knowledge of the inter-relationships between the relevant factors learnt from operating the system.
Thus for a given server cell configuration and performance characteristic, the throughput distribution Look Up Table will return values such that :
p(TQ) = LUT(TQ,Clx,lambda,c, mu, pm)
where p(TQ) is the probability that a throughput value is equal to TQ.
Clx -is based on the measurement reports and is a descriptor which describes the predicted C/l characteristics of the server cell in question for a given frequency re-use plan. In the preferred embodiment the average C/i value in the server cell is used for this parameter entry.
It is recommended that a maximum slot capability of 2 is employed. Thus in this case, pm becomes a fractional value representing the proportion of mobile stations with at least 2 timeslots capability.
In order to restrict the dimensions of the throughput value Look Up Table to a manageable size, it may be necessary to "discretize" the input variables as follows :
c - integer steps of 1 and with a maximum value of 8 (the maximum number of
TDMA slots per RF carrier) mu - steps of 0.1 with a maximum value of 6 pm - steps of 0.1 with a maximum value of 1. lambda - integer steps of 1 with a maximum value of 48
TQ - steps of 1e2 with a maximum value of 100e3 bits/s mobile station user throughput
An analytical model could be employed in place of a multidimensional throughput value Look Up Table. However, it is preferred that the LUT values are generated using a simulator.
Once the IOS block 22 has performed the throughput value analysis as described above, it then proceeds to calculate a performance metric (or "penalty value") for the server cell in question, given an initial frequency re-use plan and then does the same for all the cells in the system, each being considered in turn as a "server" cell.
Thus, a performance metric is defined whose value is dependent on the data performance characteristics. Data performance is assessed by considering the predicted throughput values relative to the target throughput value. In this embodiment, the target throughput is chosen given the PDP (packet data protocol) Context requested Quality of Service.
Thus a performance metric, Q, is defined as;
Q = f (Target Throughput - TQ).D.p(Ta) over all possible values of TQ
f - is a utility function which relates to the probability of unacceptable Quality of Service for a user for a given throughput TQ . Thus, for example, where there is a high probability that a throughput value TQ is much lower than the target, f returns a very large value reflecting user dissatisfaction.
D - is the data volume in the cell, which can be measured in kilobytes. In this embodiment, D is computed using statistics (e.g AIR_DL_BLKS) known to the OMC 21.
∑AIR_DL_DATA_BLKSb,„ - 22 + ∑AIR_DL_DATA_BLKSbm -32
(data _adjustment) ^
£> =
1024 0.01
data_adjustment = 0.1 - this parameter specifies the chosen relative value of data in comparison to voice.
D is essentially a measure of potential revenue for the network provider.
The value of Q can be computed for cases where an unassigned interferer, neighbour cell is assigned a co-channel frequency with the server cell, or where an unassigned interferer, neighbour cell is assigned an adjacent-channel frequency with the server cell or where an unassigned interferer cell is assigned
a frequency which is neither coincident with nor adjacent to the server cell's allocated frequency. (Q0).
Qo constitutes a base penalty value beyond which no further gains from the frequency plan can contribute to the data throughput.
By searching for the frequency re-use that minimises the sum of the performance metrics for all assessed cells, an optimal frequency re-use plan is achieved. Minimisation of the sum of the performance metrics can be accomplished by employing one of any suitable known algorithms. Such algorithms are known to those skilled in the art and operate by searching for an optimum setting of multiple input variables. Thus, the IOS 22 repeatedly resets the frequency re-use plan which is implemented by the OMC 21 and BSC's 17,18, then it re-calculates the performance metrics and applies the appropriate minimisation process until a minimum is reached. At this point, the current corresponding frequency re-use plan setting is the best one achievable.