2.1.4          Centralised Measurement (CM)

DCA methods in the centralised measurement  category have knowledge of both the channel usage of other cells and also the interference environment.  It has the advantage of global knowledge using centralised network information making it capable of achieving optimum solutions.  It is also able to achieve lower reuse patterns and a higher SIR by using the measured information.  However, it also inherits the disadvantages of centralised and measurement methods: i.e., large amounts of signalling and the delay and accuracy associated with measurements and consequently it can be difficult to implement.

A fully centralised and measurement method is described in [66] called MAXMIN .  In MAXMIN the base station assigns a channel that maximises the minimum SIR of all the existing subscribers.  This method somehow requires all channels to be tested (e.g. using channel probing) and all the base stations to estimate the SIR of the existing subscriber.  A method may be required to test these channels without interfering with existing calls and hence is difficult to implement.

Having centralised information, the information exchange between the subscriber and its base station can be reduced, especially in DCA algorithms where both the subscriber and base station need to perform measurements (e.g. [57]).  The use of centralised information to reduce the quantity of measurements is implemented in [67].  Two algorithms utilising the centralised and measurement  information are proposed namely the Network Assisted Least Interfered (NA-LI) method and the Network Assisted DCA with Throughput Optimisation (NA-TO).  In both methods, the subscriber performs measurements to estimate the path losses from base stations (both serving and are nearby base stations) to itself and this information is sent to its serving base station.  The base station uses the channel usage of nearby base stations – presented in the form of a matrix as in (2.3) – and the estimated path loss to calculate the SNR for each channel.  NA-LI selects the channel with the highest calculated SNR.  The data throughput (bps) is assumed to be a function of the Block Error Rates (BLER) and the throughput for each active user is stored in the base station.  NA-TO uses the SNR information to estimate the data throughput of each channel and selects the channel that would give the maximum throughput for the active users including the new subscriber (i.e. it maximises the summation of all throughputs).  Simulation results show that NA-TO achieves a higher data throughput compared to NA-LI.

Figure 2.15: Microcell and macrocell in a two-layer hierarchical cellular system.

A DCA method is proposed in [68] for use in a hierarchical cellular system  consisting of macrocells and a microcells as described in Figure 2.15.  This method uses fuzzy logic to measure the degree of channel occupancy and the average time a subscriber spends in a cell such that a fast moving subscriber is assigned to a channel in a macrocell while a slow moving subscriber is assigned to a channel in a microcell.  In conditions of low channel occupancy, a slow moving subscriber can also be allocated to a macrocell.  This reduces the number of handoffs in the network.  A compatibility matrix is defined for the macrocells and each macrocell maintains a table of the channel usage of its interference neighbouring macrocells and microcells.  For the macrocell, the channel that does not violate the macrocell compatibility matrix and has the least number of cells (macro and micro) using it within its interfering neighbours is selected.  At the microcell level, the channel with a measured interference power below a threshold and also has the highest usage among other microcells within the same macrocell is selected.  This is to encourage channel packing.  Otherwise, the channel that cannot be used by the overlaying macrocell is selected (assuming the interference of this channel is confined within the macrocell and therefore does not interfere with other macrocells).  If no such channel exists, a channel that is available to its overlaying macrocell is selected (this reduces the capacity of the overlaying macrocell).  The blocking probabilities of this method out perform that of FCA under all traffic conditions.  However, no comparisons with other DCA methods are presented.

In a Broadband Fixed Wireless Access  network, DCA can be employed so that the frequency reuse is aggressive (i.e. frequency reuse of one).  A Staggered Resource Allocation (SRA) [30], described in the FCA section and in Figure 2.11 is used for BFWA utilising a frequency reuse of one.  An Enhanced Staggered Resource Allocation  (ESRA) is introduced in [69] where measurement and coordination among base stations (i.e. centralisation) is included in SRA to improve the packet throughput.  ESRA classifies the subscribers into different levels based upon the number of concurrent transmissions that it can tolerate in a cell.  To do this a SIR measurement is performed where the base station under consideration transmits a pilot tone while the others stop their transmissions.  The subscriber uses the pilot tone as an estimate of the signal strength.  The other base station then transmits while the base station under consideration is silent.  Using these signals the subscriber can estimate its receive SIR.  The classification is done periodically during transmission and if a subscriber continuously needs retransmissions, it will be reclassified to a lower level to reduce interference.  In a similar manner to that in SRA, each time frame is divided into six time slots and each sector is given a dedicated slot (1^st choice slot) for transmission (if additional slots are required, the sector can transmit in another predefined slot – 2^nd choice slot, etc. as in Figure 2.16).  ESRA further divides each time slots into six mini-slots (i.e. total of 36 mini-slots).  Each mini-slot allows different degrees of concurrent packet transmission and hence a subscriber that can tolerate six (level 6) concurrent packet transmissions transmits at mini-slot 6.

Figure 2.16: ESRA timeslots and mini-slots arrangement in a cell (BS = base station).

Figure 2.16 shows the timeslot and mini-slot arrangement in a cell with six base stations (as in Figure 2.11) and each base station can transmit only in the non-shaded mini-slots.  For example, a level 3 subscriber of base station 1 transmits at the 3^rd mini-slot of its 1^st choice slot and if it has extra packets, it can transmit them in the 3^rd mini-slots of its 2^nd choice and 3^rd choice slots (3^rd mini-slots of higher choice slots i.e. 4^th choice and above are not allowed for transmission).  This subscriber (level 3) can transmit at a lower mini-slot (e.g. mini-slot 1 or 2) if it is available and not forbidden (i.e. not shaded).  Although this level 3 subscriber will not affect other base stations if it transmits in a higher mini-slot (e.g. mini-slot 4), its packet may not meet the required SIR (as at mini-slot 4, there are a maximum of 4 concurrent transmission).  ESRA is shown to have a better SIR performance and packet throughput than SRA.