GB2552960A - Wireless communications networks - Google Patents

Abstract

Frequency channels are allocated to sub-networks (W, X, Y, Z) in a millimetre waveband wireless mesh communications network by first defining a graph comprising vertices and interconnecting edges. In this graph, the vertices represent sub-networks and the edges have values representing interference levels between them if they were to use the same frequency channel. Assuming they must use different channels, the chromatic value of the graph is then determined using a graph colouring algorithm, this value indicating how many unique frequency channels are required. If the number of channels required is greater than the number available, edges are removed from the graph and the chromatic value recalculated until this condition is satisfied. Finally, frequency channels are allocated to the sub-networks such that no directly connected sub-networks have the same frequency. Such mesh networks find application in providing backhaul links for small cell base stations. A sub-network may also be called a Personal Basic Service Set (PBSS).

Description

(54) Title of the Invention: Wireless communications networks

Abstract Title: Allocation of frequency channels to sub-networks in a mesh communications network (57) Frequency channels are allocated to sub-networks (W, X, Y, Z) in a millimetre waveband wireless mesh communications network by first defining a graph comprising vertices and interconnecting edges. In this graph, the vertices represent sub-networks and the edges have values representing interference levels between them if they were to use the same frequency channel. Assuming they must use different channels, the chromatic value of the graph is then determined using a graph colouring algorithm, this value indicating how many unique frequency channels are required. If the number of channels required is greater than the number available, edges are removed from the graph and the chromatic value recalculated until this condition is satisfied. Finally, frequency channels are allocated to the sub-networks such that no directly connected sub-networks have the same frequency. Such mesh networks find application in providing backhaul links for small cell base stations. A sub-network may also be called a Personal Basic Service Set (PBSS).

\.

FIGURE IS /26

FIGURE 1

2/26

FIGURE 2

3/26

FIGURE 3

4/26

FIGURE 4

5/26

FIGURE 5

6/26

7/26

FIGURE 7

8/26

Ο

110 / =

9/26

10/26

FIGURE 10

11/26

FIGURE 11

12/26

FIGURE 12

13/26

FIGURE 13

14/26

FIGURE 14

15/26

FIGURE 15

16/26

2οί

Μ

1

Μ

V

2ο4

ι

2»^

V

1/

2ο}

V

2«δ

1

Jo*5

21ο

FIGURE 16

17/26

301

302

303 τ

304

FIGURE 17

18/26

FIGURE 19

19/26

FIGURE 20

FIGURE 21

20/26

FIGURE 22

21/26

FIGURE 23

FIGURE 24

22/26

FIGURE 25

FIGURE 26

23/26

FIGURE 27

24/26

FIGURE 29

25/26

FIGURE 31

26/26

FIGURE 32

WIRELESS COMMUNICATIONS NETWORKS

The present invention relates to wireless communications networks, and in particular to wireless mesh communications networks.

BACKGROUND OF THE INVENTION

Figure 1 of the accompanying drawings illustrates a simplified example wireless mesh communications network which comprises a plurality of network node devices 10 interconnected by bidirectional wireless communications links 12. The network node devices 10 operate to communicate with one another, for the transfer of communications data therebetween. This type of network is known as a “mesh” io network because of the multiple connections between network node devices that defines a mesh of communications links 12.

Wireless mesh networks are strong candidates to provide data communications, for example for Internet access, or for backhaul data traffic from small cell base stations. This is primarily because such wireless mesh networks require no cabling and can be deployed and extended in a flexible manner. Transmission and routing of data packets in a wireless mesh network is affected by many factors, including wireless link quality. This is particularly the case with outdoor networks in which the link quality can be affected by many different outdoor factors such as weather or other signal attenuating and blocking factors. In addition, low latency and low packet drop are highly desirable in such networks, since consumers desire high quality, high speed services, particularly for the delivery of online content over the wireless network. Early wireless mesh networks employed Wi-Fi technology with omnidirectional antenna, but are no longer able to meet the data throughput rates (measured across the mesh from one edge node to another) of today's data traffic requirements. In addition, such techniques are subject to interference as the unlicensed 5GHz band becomes more congested. This has led to interest in high speed millimetre wave wireless networks, such as those operating in the 60GHz waveband, for example as defined in the Institute of Electrical and Electronic Engineers (IEEE) Standard IEEE 802.11 ad. Such networks offer much higher capacity than the Wi-Fi mesh networks by exploiting large carrier bandwidth and (steerable) directional antenna to give high signal to noise ratios.

In wireless networks, such as those implemented according to the IEEE 802.11 ad standard mentioned above, sub-networks are often defined in combination with the routing and packet forwarding definitions, with traffic being forwarded across the mesh via multiple sub-networks. In IEEE 802.11 ad a sub-network is called a Personal Basic

Service Set (PBSS), and comprises two or more radio transceivers (or stations, STA) sharing access to a radio carrier over a local area. There are a multitude of ways in which radio nodes can be assigned to PBSSs and also a multitude of ways in which traffic can be routed over the mesh network which is thereby created. Clearly, identification of the combination of mesh network design and routing tables that gives io the greatest performance (for example, measured using key performance indicators (KPI) such as mean latency or throughput) is non-trivial. A joint design of the mesh network design and the routing algorithms implies a centralised algorithm which precludes the use of decentralised components.

Accordingly, it is desirable to provide a technique that can address the complexity of mesh network design and routing methods, suitable for use in such high speed wireless communications networks.

SUMMARY OF THE INVENTION

According to one aspect of the present invention, there is provided a method of allocating frequency channels for sub-networks in a millimetre waveband wireless mesh communications network which comprises a first plurality of network node devices each having a transceiver unit which includes a beamforming antenna device, a second plurality of millimetre waveband wireless communications links which interconnect the transceiver units ofthe network node devices, the wireless mesh communications network having a plurality of sub-networks defined therein, the sub-networks using respective single frequency channels and including respective pluralities of transceiver units and at least one communications link interconnecting the transceiver units of the sub-network, the method comprising providing a predetermined number of available frequency channels for allocation to the sub-networks of the network; defining a graph having vertices and edges having respective edge values, the edges interconnecting pairs of adjacent vertices, in which graph the vertices represents sub-networks of the network, and the edge values represent interference levels between sub-networks when those sub3 networks use the same frequency channel; determining the chromatic value of the graph, the chromatic value being the minimum number of unique frequency channels required to colour the edges of the graph without any of the vertices being connected with two edges of the same colour; removing edges from the graph to provide a modified graph, and determining the resulting chromatic value of the modified graph, until the resulting chromatic value is less than or equal to the number of available frequency channels; and using the modified graph, allocating the available frequency channels to the sub-networks such that no adjacent sub-networks are allocated the same frequency.

According to another aspect of the present invention, there is provided a millimetre waveband wireless mesh communications network comprising a first plurality of network node devices each having a transceiver unit which includes a beamforming antenna device; a second plurality of millimetre waveband wireless communications links which interconnect the transceiver units of the network node devices in an existing physical topology; and a frequency channel allocation unit operable to provide a predetermined number of available frequency channels for allocation to sub-networks of the network, such sub-networks using respective single frequency channels and including respective pluralities of transceiver units and at least one communications link interconnecting the transceiver units of the sub-network concerned; define a graph having vertices and edges having respective edge values, the edges interconnecting pairs of adjacent vertices, in which graph the vertices represents sub-networks of the network, and the edge values represent interference levels between sub-networks when those sub-networks use the same frequency channel; determine the chromatic value of the graph, the chromatic value being the minimum number of unique frequency channels required to colour the edges of the graph without any of the vertices being connected with two edges of the same colour; remove edges from the graph to provide a modified graph, and determining the resulting chromatic value of the modified graph, until the resulting chromatic value is less than or equal to the number of available frequency channels; and use the modified graph, allocating the available frequency channels to the sub-networks such that no adjacent sub-networks are allocated the same frequency.

In one example, allocation of the available frequency channels to the sub-networks comprises, for each sub-network, determining which of the available frequencies does not conflict with a frequency allocated to another sub-network, and allocating one of the determined frequencies to the sub-network concerned.

In one such example, one of the determined frequencies is allocated using random selection.

In one example, the available frequency channels define respective pluralities of sub-bands, the method further comprising, following allocation of the available frequency channels, allocating respective frequency channel sub-bands to the sub10 networks.

In one example, each of the first plurality of network node devices includes a plurality of transceiver units having respective directions of operation.

In one such an example, the respective directions of operation are substantially mutually perpendicular.

According to a third aspect of the present invention, there is provided a frequency channel allocation unit for use in a wireless mesh communications network embodying the second aspect of the present invention.

BRIEF DESCRIPTION OF THE DRAWINGS

Figure 1 is a schematic block diagram illustrating a wireless mesh communications network;

Figure 2 is a schematic block diagram illustrating sub-networks in a first wireless mesh communications network;

Figure 3 illustrates a multi antenna network node device;

Figure 4 is a schematic block diagram illustrating sub-networks in a second wireless mesh network;

Figure 5 illustrates constraints on sub-network definition;

Figure 6 is a schematic block diagram illustrating a sub-network definition controller;

Figure 7 is a flow chart illustrating steps ion a method embodying an aspect of the present invention;

Figure 8 illustrates an example PBSS designer;

Figure 9 illustrates interaction of the PBSS designer of Figure 8 with a routing calculation agent and network nodes;

Figures 10 to 16 illustrate steps in the definition of a mesh network topology;

Figure 17 illustrate steps in a method of frequency channel allocation embodying an aspect of the present invention;

io Figures 18 to 22 illustrate frequency channel allocation in a first method embodying an aspect of the present invention;

Figures 23 to 27 illustrate frequency channel allocation in a second method embodying an aspect of the present invention and

Figures 28 to 32 illustrate frequency channel allocation in a third method embodying an aspect of the present invention.

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

Figure 1 illustrates an “ideal” grid-like mesh communications network is which the network node devices 10 are arranged in a regular, predictable grid. As such, organisation of the network node devices 10 into a usable, optimal or near optimal communications topology, and, where appropriate, into sub-networks (also known as Personal Basic Service Sets - PBSSs - in the IEEE 802.11 ad standard) is relatively straightforward. The grid structure allows simple rules to be applied for the subdivision without consideration of the traffic routing/forwarding algorithm of the mesh, and enables the sub-network definition and routing calculations to be taken in sequence in a single calculation entity or separate entities.

However, for “off-grid” networks in which the network node devices are physically arranged in a random or at least in an irregular manner, definition of a usable, optimal or near optimal communications topology, and sub-networks where required, is more difficult since, for example, any single network node device could communicate with any number of other, possibly physically adjacent network node devices, and/or belong to any one of a number of potential sub-networks.

Embodiments of the present invention provide techniques for defining a usable, optimal or near-optimal communications topology for a mesh communications network. Embodiments are also able to define appropriate sub-networks within a communications topology.

A simplified off-grid network is illustrated in Figure 2. A first network node device 20 is connected with a wider network 19 via an external connection 20E. The external connection 20E may be provided by a high speed wireless connection or by a wired connection such as an optical fibre connection.

The first network node device 20 is able to communicate with second and third network node devices 22 and 24 over respective wireless communication links 21 and 23. The second and third network node devices 22 and 24 are able to communicate with one another over a wireless communications link 25.

In embodiments of the present invention, the wireless communications links are millimetre waveband communications links. The millimetre waveband extends from approximately 25Ghz to approximately 300GHz. A preferred waveband in which embodiments of the present invention operate is the 60GHz waveband governed by the IEEE 802.11 ad standard, which extends from approximately 58GHz to approximately 66GHz.

A fourth network node device 26 is connected with the wider network 19 via an external connection 26E, similar to that of the first network node device 20. The fourth network node device 26 is able to communicate with fifth and sixth network node devices 28 and 30 via wireless communications links 27 and 31, and the fifth and sixth network node devices 28 and 30 communicate with one another via a wireless communications link 29.

Ί

The network of Figure 2 is able to be sub-divided into two sub-networks 33 and 34. The first of the sub-networks 33 comprises the first, second and third network node devices 20, 22 and 24, and the second of the sub-networks 34 comprises the fourth, fifth and sixth network node devices 26, 28 and 30. The first network node device 20 is defined as the control node for the first sub-network 33, and the fourth network node device 26 is defined as the control node for the second sub-network 34.

The control node for each sub-network 33 and 34 controls the channel on which the sub-network concerned operates and controls communications within the subnetwork concerned. When the first and second subnetworks are connected only by io the wider network 19, routing of data packets from one of the network node devices in the first subnetwork to a network node device in the second subnetwork brackets (or vice versa) is realised through the wider network 19, using the external links 20E and 26E.

In one possible example, the third network node device 24 and fifth network node device 28 are also able to communicate over a wireless communication link 35. In such a case, routing may take place over that communications link 35 or via the wider network 19. Such routing decisions are made independently of the definition of the first and second sub-networks 33 and 34.

Figure 3 illustrates schematically a multi-antenna, multi-direction network node device. The network node device 40 of Figure 3 has four transmitting and receiving antenna units 40A, 40B, 40C and 40D, and an external network connection 40E. Each antenna unit is operable to receive and transmit data packets over a beam forming directional antenna. Each antenna has a nominal central direction 41 A, 41B, 41C or 41D, and is able to communicate over a range of directions indicated in

Figure 3 by 42A, 42B, 42C and 42D. In transmission mode, each antenna unit is operable to receive data packets for transmission, to modulate those data packets to generate modulated signals, and to transmit the modulated signals from the antenna device in a direction determined by the wireless communications link. In reception mode, the antenna unit is operable to receive detected modulated signals from the antenna device, to demodulate such received signals, and to supply demodulated data packets for further processing and routing.

A switching unit 40S is provided and is operable to switch data packets from one antenna unit 41 A, 41B, 41C or 41D to another of the units for further transmission from the network node device concerned. The switching unit 40S operates in accordance with routing control instructions, and will not be described here in detail for the sake of conciseness.

Figure 4 illustrates a more complex off grid network in which the network node devices are multi-antenna, multi-direction node devices, such as described with reference to Figure 3. As such, each network node device has a plurality of communication directions. In Figure 4, each network node device is provided with io four possible connection directions, provided by respective transceiver units. Each transceiver unit includes a media access control (MAC) layer, a physical (PHY) layer and an antenna device. In effect, each network node device provides four bidirectional network node devices, connected together by an internal switching arrangement. Any number of network node devices may be provided with any number of transceiver units, as appropriate for the network concerned. For example, each network node device may be provided with two, three or four transceiver units, and include switch means to connect data packets between the transceiver units of the node.

The transceiver units of the network node devices of the example shown in Figure 4 serve respective directions which are substantially mutually perpendicular. It will be appreciated that the respective antennas of the transceiver units may be arranged to serve any set of chosen directions.

In Figure 4, a first network node device 52 is connected to an external network via an external connection 52E. The external connection 52 E may be a wireless connection or may be a wired connection such as an optical connection. The first network node device 52 has four antennas and associated transmission and reception units 52A, 52B, 52C and 52D, hereafter referred to as transceiver units 52A, 52B, 52C and 52D. The transceiver units 52A, 52B, 52C and 52D are operable to transmit and receive radio frequency signals in respective directions, using millimetre wavelength signals, for example in the 60GHz waveband according to the IEEE 802.11 ad standard. Each transceiver unit 52A, 52B, 52C and 52D includes a beamforming directional antenna element which is operable to steer the direction of transmission or reception within a predetermined range centred on the centre direction of the antenna element. This beamforming and directionality of the antenna enables each transceiver unit 52A, 52B, 52C and 52D of the first network node device 52 to communicate with different network node devices at different network in different spatial positions.

The first network node device 52 also includes a switching unit 52S (not shown for clarity, see Figure 3) which is operable to switch data packets between transceiver units 52A, 52B, 52C and 52D of the first network node device. The switching unit 52S enables data packets received by one transceiver unit to be transmitted from the io first network node device 52 from a different transceiver unit. As such, data packets can be routed appropriately through the first network node device 52.

Having multiple transceiver units connected by the switching unit enables the network node device to provide a plurality of node devices in a single unit. The arrangement of the respective directions of the antenna elements of the transceiver units combined with the beamforming directionality of the antenna elements enables the network node device to communicate with multiple different network node devices and enables the definition of a mesh network.

The sub-networks operate using a single operating frequency channel for each subnetwork. In previously-considered networks, these sub-network operating frequency channels are allocated by a central controller on a random or sequential basis. However, such allocation can lead to undesirable interference between subnetworks. Accordingly, an embodiment of the present invention is intended to allocate operating frequency channels to respective sub-networks in a manner that reduces such interference. An example technique for defining sub-networks is described below, and is included here for context, as one example of sub-network definition to which a technique embodying the present invention may be applied. However, the techniques embodying the present invention are equally applicable to frequency channel allocation to sub-networks defined by alternative definition methods. A technique embodying the present invention is applied to a sub-network definition generated prior to the allocation of frequencies; an example of such definition is described below.

The example network of Figure 4 will now be used to describe one example process of defining sub-networks within the overall network. Firstly, the possible connections will be set out, and then the definition of sub-networks described. It will be readily understood that the network topology of Figure 4 is merely exemplary, and is not reproduced to scale. The connections described are intended to represent the connections available for the exemplary network. Accordingly, the Figure 4 network, and the following description, is not to be considered limiting.

In the example of Figure 4, a total of six network node devices 52, 54, 56, 60, 62 and 64 are illustrated. Each of the network node devices 52, 54, 56, 60, 62 and 64 io includes a multiple number of transceiver units A, B, C and D and a switching unit S, as described above with reference to Figure 3 and to the first network node device 52. The provision of multiple transceiver units in multiple network node devices allows for many potential wireless connections to be made.

In the example of Figure 4, transceiver unit 52C of the first network node device 52 is able to communicate with transceiver unit 53A of the second network node device 54, with transceiver unit 56A of the third network node device 56, and with transceiver unit 60D of the fourth network node device 60 over respective wireless communications links 51,53 and 59.

Transceiver unit 54A of the second network node device 54 is able to communicate with transceiver unit 52C of the first network node device 52, with transceiver unit 56C of the third network node device 56, and with transceiver unit 62D of the fifth network node device 62 over respective wireless communications links 51, 55, and 57.

Transceiver unit 56A of the third network node device 56 is able to communicate with transceiver unit 52C of the first network node device 52 over wireless communications link 53. Transceiver unit 56C of the third network node device 56 is able to communicate with transceiver unit 54A of the second network node device 54 over wireless communications link 55.

Transceiver unit 60D of the fourth network node device 60 is able to communicate with transceiver unit 52C of the first network node device 52, and with transceiver unit 62B of the fifth network node device 62 over respective wireless communications links 59 and 61. The fourth network node device 60 also is connected with an external network via an external connections 60E, in a manner similar to that described for the first network node device 52.

Transceiver unit 62B of the fifth network node device 62 is able to communicate with transceiver unit 60D of the fourth network node device 60, and with transceiver unit 64D of the sixth network node device 64 over respective wireless communications links 61 and 63. Transceiver unit 62D of the fifth network node device 62 is able to communicate with transceiver unit 54A of the second network node device 54 over io wireless communications link 57.

Transceiver unit 64D of the sixth network node device 64 is able to communicate with transceiver unit 62B of the fifth network node device 62 over wireless communications link 63.

It will be appreciated that the exact nature of the available communications links in the network is not important to principles of the invention, and that the network shown in Figure 4 is shown merely by way of example in order to explain those principles. In the example of Figure 4, it is assumed that the wireless communications links described are the only wireless links available for the purposes of describing the principles of the present invention. The wireless mesh network shown in Figure 4 is defined by the existing physical topology of the network node devices. That is, the wireless communications links that are available are determined by the physical layout of the network node devices and the environment in which they are placed. For example, a specific location of a network node device may preclude communication in one or more directions for that device, due to local physical conditions. In addition, in certain examples of real environments, the communications links may be provided by reflective paths.

An example of a restriction on the definition of a sub-network is illustrated in Figure 5. In Figure 5, a first network node 80 is in communication with a second network node 82 over a communications link 81 defined between respective transceiver units

80B and 82D. The transceiver unit 80B and 82D of the first and second network nodes 80 and 82 respectively are initially defined as a sub-network, with either of the nodes performing the role of the sub-network controller. A transceiver unit 84D of a third network node 84 joins the sub-network, and communicates with the first network node 80 over the communications link 83. The communications link 83 is defined between the transceiver unit 80B of the first network node and the transceiver unit 84D of the third network node 84.

The transceiver unit 84D of the third network node 84 is not able to communicate directly with the transceiver unit 82D of the second network node 82. Accordingly, the second and third sub-network nodes 82D and 84D cannot be defined as the subnetwork controller for the sub-network shown in Figure 5. Therefore, the first sub10 network node 80B must be defined as the sub-network controller, since it the only controller that is able to communicate directly with all of the members of the subnetwork. Thus, the location of the third network node imposes a constraint on the definition of the sub-network with which the third network node is joining.

It will be readily appreciated that the sub-network shown in Figure 5 is very simplified, and is shown merely to illustrate a typical constraint on the definition of the sub-network.

Returning to the example network of Figure 4, there are several sub-networks that can be defined. A first sub-network 70 can be defined as including the first, second, third and fourth network node devices 52, 54, 56 and 60, using respective transceiver units 52C, 54A, 56A and 60D. For the first sub-network 70, transceiver unit 52C is defined as the control unit since transceiver unit 52C is the only one of the units able to communicate directly with each of the other members of the subnetwork.

A second sub-network 72 can be defined as including the second, third and fifth network node devices 54, 56 and 62, using transceiver units 54A, 56C and 62D.

The second network node device belonging to the second sub-network is designated as the sub-network controller, since this device is in direct communication with all of the other devices in the sub-network. However, transceiver unit 54A should not be placed in both the first and second subnetworks (otherwise very tight coordination is required between PCPs to allow the coexistence), and so this conflict must be resolved during the subnetwork definition process.

A third sub-network 74 can be defined as including the fourth, fifth and sixth network node devices 60, 62 and 64, using respective transceiver units 60D, 62B, and 64D.

It will be appreciated that the network layout shown in Figure 4 is merely exemplary, and is intended to illustrate the principles of the present invention, which may be applied to any suitable wireless mesh network.

In order to define the relevant sub-networks, sub-network definition functionality is provided in the network. This definition functionality may be provided by the switch units in any one or more of the network node devices, in a centralised controller 76, which may a software-defined network (SDN) controller, or distributed amongst an io appropriate number of the units.

Figure 6 illustrates a controller 90 for providing the topology and sub-network definition functionality. The controller 90 comprises a control unit 92, an input unit 94, a data storage unit 96, and output unit 98, and a user interface 100. Any or all of these units may be provided by dedicated hardware units, or may be provided by shared resources in an existing controller or processing unit of the network.

The input unit 94 is operable to provide data input for the control unit 92, as will be described below. The control unit 92 is operable to perform the required communications topology and sub-network definition operations, to store in, and retrieve data from, the data storage unit 96, and to output data via the output unit 98.

The user interface 100 is operable to provide user interface functions for the controller 90. The data storage unit 96 stores data relating to network physical topology, data traffic expectations, path loss modelling for the wireless communications links, and rules relating to the principles of the sub-network definition. The control unit 92 is operable to access and update this data in the data storage unit 96. Control unit 92 is also known as a topology/PBSS Designer, so called because it specifies the communications topology and composition of each PBSS in the mesh network (that is, which 802.11 ad STAs belong to each PBSS and which STA acts as the controller (PCP)), and also specifies the radio channel of each PBSS (since multiple radio channels are available in the 60 GHz band and radio channels can also be divided into sub-channels - exploiting radio channels reduces the impact of interference between PBSSs). Such radio channel assignment can be achieved in any suitable manner; for example, by using a specific allocation technique, or by allocating channels randomly, or by allocating the same channel to each PBSS.

Definition of sub-networks in a wireless mesh network will now be described with 5 reference to Figure 6 and the flow chart of Figure 7.

At step 101, the control unit 92 retrieves network topology data from the data storage unit 96. This network topology data may include information relating to physical locations of the network nodes in the network, the number of transceiver units per network node device, and the ranges of possible communications directions for io those transceiver units.

At step 102, the input unit 94 provides real-time data relating to the network to the control unit 92. This real-time data may include measurements from one or more of the network node devices. For example, the real-time data may include information showing which other network node devices have been detected by a given network node, signal strengths for wireless communications links between network node devices, and/or other relevant information.

At step 103, the control unit 92 retrieves data relating to the rules and principles by which the communications topology and sub-networks are to be defined.

At step 104, the control unit 92 combines the retrieved physical topology data, the 20 received real-time data and the retrieved rules data in order to define the communications topology, and where required a sub-network for the mesh communications network. For each such sub-network, a sub-network controller is chosen from the network node devices forming the sub-network concerned.

At step 105, the communications topology and sub-network definitions data are 25 stored in the data storage unit 96. The communications topology and sub-network definitions data are then accessible by routing functionality which determines routes through the network for specific data packet flows.

Steps 101 to 105 may be repeated at appropriate times, for example periodically.

In such a manner, a communications topology and sub-networks are able to be defined with lower impact on network performance than in previously-considered techniques.

Figure 8 illustrates an example topology/PBSS designer 110. The designer 110 5 includes a design unit 112, which runs multiple design algorithms/heuristics 112A,

112B, 112C. The design algorithms/heuristics 112A, 112B, 112C receive an input 114, which may include some or all of the following data:

• Physical topographical information (for example, node locations, possible antenna azimuths, STAs per Node) io · radio environment information (for example, this could be model based and comprise a path loss model to apply (line-of-sight, non-line-of-sight, path loss exponent), shadowing model, expected rainfall rate, small-scale fading model, or it could include radio measurements from field testing (including from the live network if the designer is executed on a network that is up and running) or estimated by an external tool (e.g. a radio planning tool) • objectives of design (e.g. robustness to failover or link performance degradation (e.g. from rain or foliage growth or other partial shadowing), max aggregate throughput, minimum latency, minimum mean latency).

The design algorithms/heuristics 112A, 112B, 112C use the input data to generate 20 respective PBSS designs. The designs capture which network nodes are connected (some may be left idle with no connectivity), and to which PBSS each node belongs. Each PBSS is assigned a radio channel of operation, and the network node within the PBSS that is the controller (the PCP) is identified.

A simulator unit 116 receives the PBSS designs from the respective design 25 algorithms/heuristics. The simulator unit is operable to calculate the link data rates of the mesh for links that interconnect network nodes in the defined PBSSs, to consider the likelihood of interference between links, and, given the expected traffic rates and characteristics at each traffic source to the mesh, together with the identity of the associated target network node, to assess the performance against the design objectives.

A comparison unit 118 weights the figures of merit produced by the simulator unit, and is operable to determine which PBSS design is best suited to meeting the predetermined criteria. The comparison unit 118 has an output 120 from which the chosen PBSS design is provided to a routing calculation agent.

In an alternative example, a single design algorithm/heuristic is operated by the design unit 112. The results of the design algorithm/heuristic are produced by the simulator unit 116, and used in future operations of the design algorithm/heuristic.

Figure 9 illustrates the interaction of the designer 110 with a routing calculation agent 122 and network nodes 124. As described above, the designer 110 provides the io routing calculation agent 122 with a definition of the communications topology and any PBSSs for the mesh network. The routing calculation agent 122 determines routing for data being transferred across the mesh network and PBSSs defined therein. The routing agent 122 provides routing data, communications topology, and the local PBSS designs to each network node 124 for use thereby in the IEEE

802.11 ad MAC operations and the routing of data packets being transferred across the mesh network.

Each network node 124 may include a plurality of transceivers 126. The transceivers 126 provide respective media access control/physical level functions (MAC/PHY) of the network node 124.

The designer and/or the routing agent(s) may be implemented using a software defined network (SDN) paradigm to control the network node devices.

Figures 10 to 16 illustrate communications topology definition in accordance with the principles of the present invention. Figures 10 to 15 show progress of defining a communications topology for an arrangement of nodes and transceivers in the network, and Figure 16 is a flowchart corresponding to steps in a method embodying an aspect of the present invention.

In Figures 10 to 15, a group of six nodes is illustrated, the nodes being labelled A, B, C, D, F, and G respectively. Each node comprises four transceiver units (STA), labelled A, B, C and D respectively for each node as shown in Figure 3. A particular transceiver unit is labelled using the notation NODE(STA). For example, the transceiver A of node A is labelled, and referred to, as A(A) in the description below.

It will be readily appreciated that the network may include any appropriate number of nodes, and that any given node may include any appropriate number of transceivers.

A method embodying an aspect of the present invention uses first and second metric values (E1, E2) for each transceiver and communications link connected thereto in the network to determine the communications topology for the mesh communications network. Respective first metric values are calculated for each communications link in the network. In one example, the first metric value for a link is the reciprocal of the link data rate for the link, and is assumed to be known a priori.

The first metric values (E1 ij) are determined (step 201) for each link (between transceivers i and j) in the network. In Figures 10 and 11 these first metric values are shown adjacent the links concerned; in Figures 12 to 15, these metric values have been omitted for the sake of clarity. It is assumed that the first metric value (E1 ij) for a link is applicable in each direction of the link.

In the example, the first metric values are as follows:

Link	First Metric Value (E1i,j)
A(B)-B(D)	1
A(B)-C(D)	5
A(A)-D(C)	2
B(B)-C(D)	2
B(A)-D(B)	4
B(B)-F(C)	3
B(A)-G(C)	6
D(B)-F(D)	5
F(A)-G(B)	1

The communications links of the network define a plurality of paths through the network from the transceivers to an arbitrary destination transceiver. At step 202, the first metric values are used to determine respective cost values for each path through the network. The cost values are an aggregation (for example, sum) of the respective first metric values for the communications links making up the path concerned. Also at step 202, a cost value V1 is selected for each transceiver from the cost values for the paths connected with the transceiver concerned. In one example, the cost value V1 represents the shortest path distance from the transceiver concerned to the arbitrary destination transceiver. In one example, the destination transceiver is that which provides a gateway (GW) node for the mesh network. A gateway node is a node which connects the mesh network with communications infrastructure, such as a fibre optic network. In the example shown in Figures 10 to 15, the gateway node is node A. In this example, it is assumed that io there is the same cost for switching from one transceiver at a node to another transceiver at that node, and so each transceiver at a node has the same cost value. It will be appreciated that, in other examples, the transceivers at a given network node device may have different respective selected cost values.

In the example shown, the selected cost values for each node are:

Node	Cost Value (V1)	Links to gateway node
B	1	A(B)-B(D)
C	3	A(B)-B(D) + B(B)-C(D)
D	2	A(A)-D(C)
F	4	A(B)-B(D) + B(B)-F(C)
G	5	A(B)-B(D) + B(B)-F(C) + F(A)-G(B)

The resulting values of first metric value E1 and selected cost values V1 are illustrated in Figure 10.

At step 203, a second metric value is calculated for each link. The second metric value (E2ij) for a link between transceivers i and j relates to the first metric value (E1ij) is given by:

E2ij = E1i,j-|V1i-V1j| where E2ij is the second metric value for the link interconnecting transceivers i and j, E1 ij is the first metric value for the link interconnecting transceivers i and j, V1i is the cost value for transceiver i and V1 j is the cost value for transceiver j.

The second metric values E2, for this example, are shown below, and are shown in the format E1 ;E2 for each link in Figure 11:

Link	E2i,j
A(B)-B(D)	0
A(B)-C(D)	2
A(A)-D(C)	0
B(B)-C(D)	0
B(A)-D(B)	3
B(B)-F(C)	0
B(A)-G(C)	2
D(B)-F(D)	3
F(A)-G(B)	0

At step 204, for each transceiver, a link is determined as a candidate link in dependence upon the value of the second metric value. If two or more links have the same second metric value, and are able to be selected as candidate links, then the link is identified as the candidate link in dependence upon the first metric value.

In one example, the link with the lowest second metric value is identified as a candidate link. If two or more links have the same lowest second metric value, then the link with the lowest first metric value is identified as the candidate link.

io A candidate link is then defined as a selected link forming part of the mesh network communications topology where the link concerned is identified in step 204 by both transceivers at respective ends of the link concerned (step 205). Figure 12 illustrates this initial selection of first selected links that form part of the communications topology of the mesh network. In this case, links A(B)-B(D), A(A)15 D(C), B(B)-C(D), B(A)-G(C) and F(A)-G(B) are chosen, as indicated by the dotted lines in Figure 12. Link B(B)-C(D)is selected by virtue of the first metric value for that link being lower that the first metric value of the other candidate link B(B)-F(C).

For each transceiver having a connected link, the remaining links connected to that transceiver are removed from further consideration. In one example, this is achieved by setting the second metric value for the link to infinity (step 206).

If the most recent iteration of step 204 results in at least one further link being chosen to be included in the topology, then the method returns to step 203. In the example, such a repeat of steps 203 and 204 results in selection of link D(B)-F(D) (as illustrated in Figure 13).

If additional links are not added, then the original second metric values are reinstated (step 207).

For each transceiver that remains unconnected in the communications topology, a link connected thereto is selected as a second selected link in dependence upon the second metric value. If two or more links have the same second metric value, and io are able to be selected as candidate links, then the link is identified as the candidate link in dependence upon the first metric value. In one example, the second selected link is that link having the lowest second metric value (step 208). If there are two or more links having the lowest second metric value, then the link having the lowest first metric value is selected. Alternatively, a different metric can be used to decide which link to select in step 208, in the case where a given transceiver can connect with more than one other transceiver. In one example, a link is selected as a second selected link on the basis of minimum occupancy (i.e., if an unconnected transceiver can connect to another transceiver which is connected to one other transceiver, or a different transceiver which is connected to two other transceivers, the first is selected). The second selected link is then added to the communications topology (step 209). In the present example, the transceiver F(C) is not connected, but has an available link (B(B)-F(C)). As such, step 209 results in link B(B)-F(C) being chosen to be added to the communications topology.

Steps 208, 209 and 210 are looped to ensure that all unconnected transceivers are connected to the communications topology. In one example, the steps are performed for unconnected transceivers in order of distance from the destination, or gateway, node.

If any parts of the resulting communications topology are of the form one transceiver to more than one transceiver, then for the transceivers which are connected to the central transceiver, all other links are removed from consideration (step 210).

The final communications topology is shown in Figure 15, and it will be appreciated that links A(B)-C(D), and B(A)-D(B) are not included in the final communications topology.

In order to determine appropriate subnetworks (PBSSs) in the communications 5 topology, a PBSS controller for each PBSS must be chose. In the case of PBSSs with two members, either can be the PBSS controller. In one example, the default is to use the one closer to the gateway (i.e., fewest hops away). However, another selection method could be used.

In the example communications topology of Figure 15, several PBSSs are defined, io as shown below:

PBSS	Members	Controller
1	A(B), B(D)	A(B)
2	A(A), D(C)	A(A)
3	B(B), C(D), F(C)	B(B)
4	B(A), G(C)	B(A)
5	G(B), F(A)	F(A)
6	D(B), F(D)	D(B)

The case where there are more than two members in a PBSS can only occur as a one-multiple structure, i.e., a three member PBSS would have one transceiver connected to the other two, which are not connected to each other, and likewise a four member PBSS would be in the form of one transceiver connected to the other three, none of which are connected to each other (and so on). This is forced to be the case by the techniques described above, and the controller is chosen as the transceiver which is connected directly to all of the others.

In an optional example, the designer operating to compare the selected links selected to an existing PBSS design, which includes some or all of the nodes in the population.

For a given existing PBSS having one transceiver forming part of the topology, and if all other transceivers in that PBSS are as yet unconnected (i.e., according to the link selections in the loop between step 204 to step 206) then each link in the PBSS concerned is automatically selected in the new communications topology, following step 206.

Valid PBSSs should not contain more than one transceiver from the same node. In 5 addition, each pair of nodes should be directly interconnected a maximum of once (even if the connectivity between the various transceivers on these nodes theoretically would enable such connectivity). These constraints can be checked each time a link is about to be added, such that the link is not added if these conditions are violated.

io Following definition of the PBSSs, any communications link which connects two transceivers in the same PBSS may be included in the communications topology.

Such a technique enables the algorithm to run in dynamic mode, where a small change to the node or link population such as a single node being added or removed is handled in such a way as to minimise the required re-designation of PBSSs, whilst also adapting to the change.

Each PBSS is assigned a single frequency channel in which to operate. All communications within a PBSS take place over the assigned frequency channel, under the control of the assigned PBSS controller.

In order to reduce and mitigate the effects of radio communications interference 20 between PBSSs, particularly between those that are geographically adjacent one another, a method embodying an aspect of the present invention provides a technique for allocating frequency channels to PBSSs of a mesh wireless communications network. An underlying assumption is that radio transmissions generate order of magnitude less interference to each other when they employ different radio channels, compared to when the same channel is used.

In an embodiment of the present invention, a frequency allocation unit operates to allocate respective frequency channels to PBSSs in a manner intended to reduce interference. The frequency allocation unit may be provided by the controller 92 of Figure 6, or my any appropriate unit, agent or functional element. The frequency allocation unit performs a method as follows, and as illustrated in the flowchart of Figure 17.

The unit represents the network in graph theoretic form (step 301) in which each vertex of the graph is a PBSS and the weight of each edge between vertices corresponds to the interference between each pair of PBSSs (if the PBSSs were to use the same frequency channel), assuming a schedule where the air interface is shared in time equally between all the links in the PBSS and each link operates for an equal time in each direction.

As is well known, an important property of a graph is the chromatic number, that is io the minimum number of colours required for vertex colouring without vertices joined by an edge having the same colour. In accordance with the principles of the present invention, the lowest weight edges are removed (step 302) from the graph until the remaining graph has chromatic number equal to the number of frequency channels available. The chromatic number may be determined by one of the algorithms described below with respect to step 303.

A search technique is used to find a number of edges which must be removed for the remaining graph to have a chromatic number equal to or less than the number of frequency channels. It is necessary to the “less than” condition to include situations where a removing a single edge reduces the chromatic number by more than one.

In a first example, the lowest weight edge is repeatedly removed from the graph until the graph meets the chromatic number requirement. In a second example, a binary search technique is used to determine the fewest number of edges that must be removed for the chromatic number constraint to be met. In a third example, edge removal following example one may be constrained by only considering removing edges that are connected with vertices that have more edges connected thereto than there are frequency channels to be allocated. For example, with two frequency channels to be allocated, only edges connected to those vertices with three or more edges connected thereto are considered for removal from the graph. These three example edge-removing techniques are described in more detail below.

Following the reduction of the number of edges, a suitable graph colouring algorithm is used (step 303). One example algorithm is the DSATUR graph colouring algorithm, as described in, for example, J. Riihijarvi, M. Petrova, and P. Mahonen, “Frequency allocation for wlans using graph colouring techniques, in Proceedings of the Second Annual Conference on Wireless On-demand Network Systems and Services, ser. WONS '05. Washington, DC, USA: IEEE Computer Society, 2005, pp. 216(222, and in D. Brelaz, “New methods to color the vertices of a graph, Commun. ACM, vol. 22, no. 4, pp. 251-256, Apr. 1979.

Such an algorithm always chooses the lowest number colour of the options, io Accordingly, the resultant solution tends to be uneven in terms of frequency usage.

In particular, this second issue is not ideal in a practical implementation of a wireless mesh communications network, mainly due to interference which is not accounted for explicitly within the graph (for example, because of the process of cutting low cost edges from the graph). In addition, an uneven use of the frequency channels is likely to lead to a greater than necessary adverse effect of this unaccounted-for interference.

Accordingly, in an embodiment of the present invention, an additional step (304) is introduced after the colour designation in which, for all the vertices in a random order, all of the candidate colours (i.e., colours not used by a neighbour vertex (PBSS)) are identified and a new colour is chosen at random from these (which may or may not be the original colour).

Figures 18 to 22 illustrate a first frequency channel allocation process in accordance with the principles of the present invention, for four sub-networks W, X, Y and Z, for the case where there are two frequency channels to be allocated. The method of

Figures 18 to 22 repeatedly removes the edge having the lowest weight until the chromatic number of the graph is less than or equal to the number of frequency channels available for allocation.

The exemplary graph shows the four sub-network W, X, Y and Z interconnected by a plurality of edges. Each edge has an edge weight associated therewith, and these edge weights are shown adjacent the edges in Figure 18 to 21. It will be appreciated that the graph shown in Figures 18 to 22 is merely exemplary.

Figure 18 shows the initial graph including all vertices and edges. The Figure 18 graph has a chromatic number greater than two (the number of frequency channels available), and so one edge is removed. In this example, the edge having the lowest edge weight (indicating the lowest potential interference level) is removed - the edge between vertex W and vertex Z.

The resulting graph (Figure 19) also has a chromatic number greater than two, and so another edge is removed, this time the edge between vertex X and vertex Z. The resulting graph again has a chromatic number greater than two, and so another edge is removed. In this example, the edge with the lowest weight of this remaining is that io connecting vertex X and vertex W. The removal of this edge gives the resulting graph shown in Figure 20. Once again, this graph has a chromatic number greater than 2, and so a further edge (that connecting vertex X and vertex Y) is removed.

Figure 21 shows the resulting graph having a chromatic number of two, thereby meeting the requirement that the chromatic number is less than or equal to the number of frequency channels available for allocation.

Figure 22 illustrates the frequency channel allocation resulting from step 203, for the example graph. Vertices (sub-networks) W and Z are allocated a first frequency channel, and vertices (sub-networks) X and Y are allocated a second, different, frequency channel.

Figures 23 to 27 illustrate a second frequency channel allocation process in accordance with the principles of the present invention, for four sub-networks W, X,

Y and Z, for the case where there are two frequency channels to be allocated. The process of Figures 23 to 27 uses a binary search technique.

Figure 23 shows the initial graph including all vertices and edges. The Figure 23 graph has a chromatic number greater than two (the number of frequency channels available). Accordingly, it is necessary to remove at least one edge from the graph in order to reduce the chromatic number of the graph. In this example technique, half the number of edges are removed, that half being those edges having the lowest edge weights.

The resulting graph is shown in Figure 24, in which the edges between vertices W and X, W and Z and X and Z are removed. This resulting graph has a chromatic number of three, and so further edges need to be removed.

In this example technique, half of the total number of remaining edges are again removed; the removed edges having the lowest edge weights. In the example shown, there are three edges remaining, and two are removed, because, in this example, it has been chosen to round up from a “half” edge. Alternatively, the technique could round down and leave the edge concerned. In the present example, the edges connecting vertices W and Y and X and Y are removed, and this leaves io the edge connecting vertices Y and Z, as shown in Figure 25. The chromatic number of this graph is two, and so the graph meets the requirement.

In the example technique, however, removed edges may be reinstated in order to determine the least number of edges that need to be removed in order that the graph meets the chromatic number constraint. The edges are reinstated (and possibly then removed) according to the same binary search method, working between the greatest known edge count that meets the chromatic number constraint and the smallest known edge count that fails the constraint. In the example, these count values are 1 and 3, respectively, when the algorithm has reached Figure 25. The algorithm therefore considers a count of two edges for the next assessment.

In the present example, the edge connecting vertices W and Y is reinstated, giving a chromatic number of two (Figure 26). Reinstating the next edge (the X-Y edge) is not attempted because it is already known that this graph has a chromatic number of three, and so the edge is not reinstated (the binary search has concluded). The resulting coloured graph is shown in Figure 27.

Figures 28 to 32 illustrate a third frequency channel allocation process in accordance with the principles of the present invention, for four sub-networks W, X, Y and Z, for the case where there are two frequency channels to be allocated.

The exemplary graph shows the four sub-network W, X, Y and Z interconnected by a plurality of edges. Each edge has an edge weight associated therewith, and these τι edge weights are shown adjacent the edges in Figures 28 to 31. It will be appreciated that the graph shown in Figures 28 to 32 is merely exemplary.

Figure 28 shows the initial graph including all vertices and edges. The Figure 28 graph has a chromatic number greater than two (the number of frequency channels available), and so one edge is removed. In this example, the edge having the lowest edge weight (indicating the lowest potential interference level) is removed - the edge between vertex W and vertex Z.

The resulting graph (Figure 29) also has a chromatic number greater than two, and so another edge is removed, this time the edge between vertex X and vertex Z. The io resulting graph (Figure 30) again has a chromatic number greater than two, and so another edge is removed.

In this example, at this stage only the edges connected with vertex Y are considered for removal from the graph, since the number of edges connected with vertex Y (three) exceeds the number of frequency channels to be allocated (two). The edge from this consideration group with the lowest edge weight connects vertex X with vertex Y, and so this edge X-Y is removed.

Figure 31 shows the resulting graph having a chromatic number of two, thereby meeting the requirement that the chromatic number is less than or equal to the number of frequency channels available for allocation.

Figure 32 illustrates the frequency channel allocation resulting from step 203, for the example graph. Vertices (sub-networks) W and Z are allocated a first frequency channel, and vertices (sub-networks) X and Y are allocated a second, different, frequency channel.

As an alternative to the heuristic technique described, an exhaustive checking technique may be used to establish the chromatic number of the graph. Such a development would significantly increase the computational load of the technique, but would result in increased accuracy.

The result of a method embodying the principles of the present invention is a frequency channel distribution that removes or reduces the effects of channel interference between sub-networks (PBSSs).

In a practical implementation of a method embodying the present invention, it is desirable to use full antenna patterns, rather than approximations. It is anticipated that, owing to reflections of objects in real environments, interference measurements not calculations will be needed in practice.

Reduction of interference may also be achieved by improving the planning of the communications topology so as to mitigate the interference arising from the leakage between adjacent frequency channels. This may be achieved by modifying final step of the method, so that preference is given to unoccupied channels which are not adjacent to the channels of any PBSSs with which an edge is shared.

In one practical example, the allocation of frequency channels is carried out in a socalled ‘half-band mode’, in which the network operates with a number (‘n’) of frequency channels/bands. Each of the frequency channels may operate in two subbands, or may operate as a full-band, on a PBSS by PBSS basis.

In such an example, this means that typically all PBSSs are set to half-band mode, using a single sub-band for communications. Such operation may be changed to full-band mode making use of the whole of a frequency channel/band (for example, according to current traffic conditions). The technique for allocation of frequency channels/bands is then adapted such that it mitigates interference even when PBSSs may arbitrarily shift to full-band mode.

This is achieved by running the frequency channel allocation technique with n colours, and then for each of n subsets (i.e., corresponding to each channel) running the method again with two colours, to determine which of the two sub-bands should be used. Thus, the first application of the method determines the full-band frequency allocation, and the second application determines which of the two half-bands should be used by a PBBSS. It will be readily appreciated that any appropriate number of sub-bands may be provided for each frequency channel.

Claims (9)

CLAIMS:

1. A method of allocating frequency channels to sub-networks in a millimetre waveband wireless mesh communications network which comprises a first plurality of network node devices each having a transceiver unit which includes a

5 beamforming antenna device, a second plurality of millimetre waveband wireless communications links which interconnect the transceiver units of the network node devices, the wireless mesh communications network having a plurality of subnetworks defined therein, the sub-networks using respective single frequency channels and including respective pluralities of transceiver units and at least one io communications link interconnecting the transceiver units of the sub-network, the method comprising:

providing a predetermined number of available frequency channels for allocation to the sub-networks of the network;

defining a graph having vertices and edges having respective edge values,

15 the edges interconnecting pairs of adjacent vertices, in which graph the vertices represents sub-networks of the network, and the edge values represent interference levels between sub-networks when those sub-networks use the same frequency channel;

determining the chromatic value of the graph, the chromatic value being the

20 minimum number of unique frequency channels required to colour the edges of the graph without any of the vertices being connected with two edges of the same colour;

removing at least one edge from the graph to provide a modified graph, and determining the resulting chromatic value of the modified graph, until the

25 resulting chromatic value is less than or equal to the number of available frequency channels; and using the modified graph, allocating the available frequency channels to the sub-networks such that no adjacent sub-networks are allocated the same frequency.

2. A method as claimed in claim 1, wherein allocation of the available frequency channels to the sub-networks comprises, for each sub-network, determining which of the available frequencies does not conflict with a frequency allocated to another subnetwork, and allocating one of the determined frequencies to the sub-network

5 concerned.

3. A method as claimed in claim 2, wherein one of the determined frequencies is allocated using random selection.

4. A method as claimed in any one of the preceding claims, wherein the available frequency channels define respective pluralities of sub-bands, the method io further comprising, following allocation of the available frequency channels, allocating respective frequency channel sub-bands to the sub-networks.

5. A millimetre waveband wireless mesh communications network comprising:

a first plurality of network node devices each having a transceiver unit which includes a beamforming antenna device;

15 a second plurality of millimetre waveband wireless communications links which interconnect the transceiver units of the network node devices; and a frequency channel allocation unit operable to:

provide a predetermined number of available frequency channels for allocation to sub-networks of the network, such sub-networks using

20 respective single frequency channels and including respective pluralities of transceiver units and at least one communications link interconnecting the transceiver units of the sub-network concerned;

define a graph having vertices and edges having respective edge values, the edges interconnecting pairs of adjacent vertices, in which

25 graph the vertices represents sub-networks of the network, and the edge values represent interference levels between sub-networks when those sub-networks use the same frequency channel;

determine the chromatic value of the graph, the chromatic value being the minimum number of unique frequency channels required to colour the edges of the graph without any of the vertices being connected with two edges of the same colour;

remove edges from the graph to provide a modified graph, and determining the resulting chromatic value of the modified graph, until

5 the resulting chromatic value is less than or equal to the number of available frequency channels; and allocate, using the modified graph, the available frequency channels to the sub-networks such that no adjacent sub-networks are allocated the same frequency.

10

6. A network as claimed in claim 5, wherein allocation of the available frequency channels to the sub-networks comprises, for each sub-network, determining which of the available frequencies does not conflict with a frequency allocated to another subnetwork, and allocating one of the determined frequencies to the sub-network concerned.

15

7. A network as claimed in claim 6, wherein one of the determined frequencies is allocated using random selection.

8. A network as claimed in any one of claims 5 to 7, wherein the available frequency channels define respective pluralities of sub-bands, and the frequency channel allocation unit is operable, following allocation of the available frequency

20 channels, to allocate respective frequency channel sub-bands to the sub-networks.

9. A frequency allocation unit for a millimetre waveband wireless mesh communications network mas claimed in any one of claims 5 to 8.

Intellectual

Property

Office

Application No: GB1613940.4 Examiner: Dr Stephen Bevan