WO2011095884A1

WO2011095884A1 - Medium access control in wireless sensor networks

Info

Publication number: WO2011095884A1
Application number: PCT/IB2011/000208
Authority: WO
Inventors: Thi Hong Hanh Le
Original assignee: University Of Cape Town
Priority date: 2010-02-08
Filing date: 2011-02-08
Publication date: 2011-08-11
Also published as: ZA201206813B

Abstract

A Wireless Sensor Network (WSN) (1 ) and a method of scheduling communication in a WSN including a plurality of sensors (3) spaced apart over a sensing field (4) is provided. Each sensor (3) in the sensing field (4) is grouped into one of a plurality of sensor clusters (7) and each cluster is identifiable by means of a cluster identifier. A time slot during which sensors in a particular cluster area will be in an active state and capable of transmitting and receiving data can be determined from the cluster identifier of each cluster. Communication between the sensors (3) is controlled by means of a combination multiple access control layer technique comprising both a contention-based and schedule-based protocol.

Description

MEDIUM ACCESS CONTROL IN WIRELESS SENSOR NETWORKS

FIELD OF THE INVENTION This invention relates to wireless sensor networks (WSNs). In particular, but not exclusively, the invention relates to medium access control of the sensors operating within a WSN.

BACKGROUND TO THE INVENTION

WSNs typically consist of a plurality of autonomous, spatially distributed devices positioned at a number of nodes spread out over a geographical area, often referred to as a sensing field. The network uses the sensors to monitor and capture data relating to environmental or other physical quantities::

Each node in a WSN typically employs one or more sensors, limited storage means, processing capability, and an independent, power source, usually in the form of a battery. In the remainder of this specification the term "sensor" should be widely construed to include an autonomous unit incorporating all of these features.

WSNs often include a large number of sensors which, due to the limited, resources of their batteries, need to use their power sources optimally in order to prolong their useful lifetimes. The sensors often employ a variety of energy conserving techniques, for example, communication protocols that allow them to revert into a hibernating (also sometimes referred to as sleeping) state when they are idle. Sensors in a WSN may also be mobile and often cannot participate in a particular WSN for a long period of time. In such a WSN the sensor topology is dynamic and requires periodical reconfiguring, leading to high additional communication and computational overheads.

A simple way of transmitting data from a sensor to a single base station, commonly referred to as a sink, is via broadcasting directly from the sensor to the base station. For this purpose sensors are also typically equipped with radio transceivers or other wireless communication devices. The problem with this model is that it is not energy, efficient as a base station may be geographically far from the sensor node. Transmission over larger distances consumes considerably more energy.

One of the main constraints in WSNs is therefore the limited energy resource of the sensors. A base station, on the other hand, is normally more powerful and typically has a continuous power supply. The base station also typically has to perform more advanced data processing, aggregate data from multiple sensors, and transmit data to other processing stations. To send data from sensors to the base, station, a technique referred to as multi-hop routing is often used in WSNs to reduce the power consumption in the sensor nodes and prolong the WSN lifespan. Multi-hop routing implies that information gets transmitted ("hops") from one sensor node to a next until it reaches the information destination. Each "hop" is therefore over a shorter distance than would be the case for direct transmission and generally consumes less energy. This means that, on aggregate; more sensors will remain powered for a longer period of time.

The most common routing solution is to form a tree, rooted at the base station, along which data may be transmitted. However a single base station/route means a lack of routing redundancy which implies that a single failure could result in failure of a substantial portion of the network or, worst case scenario, the whole network. If there is more than one sink or base station in a system then the WSN needs to deploy a certain routing algorithm at the sensors to determine the routes along which information must be sent to reach the sink or base station as efficiently as possible. Such a tree structure is generally referred to as a minimum spanning tree.

The applicant is aware of a number of approaches that have^; been implemented for data gathering and transmission from sensors to a base station. As mentioned above, one such approach is the direct communication scheme. A problem with this scheme is that it consumes a large amount of energy with every transmission, especially when sensor nodes are far from the base station. This leads to the sensors exhausting their batteries quickly. Once a sensor has exhausted its available power resources it will die. The network as a whole will, however, still function albeit without the dead sensor(s).

A second approach for routing data to a base station is referred to as Low Energy Adaptive Clustering Hierarchy ("LEACH"). LEACH partitions sensor nodes into a number of clusters, each with a icluster head. Sensors in the same cluster take turns to be the head of the cluster and send data to their cluster head. Cluster heads then send their aggregated and compressed data packets to the base station to reduce the overall energy consumption.

In a multi-hop routing model, sensor nodes form a tree structure to multi-hop fused data to the base station. As mentioned above, each routing hop is short so the energy consumed per hop is low. However the trade-off of this model is that the delay from the source of the information to the base station is high (> log₂/V where N is the number of sensor nodes). This could make the scheme unsuitable for use in delay-sensitive applications.

A further problem encountered in conventional WSNs is that of data collision. Data collision occurs when two or more nodes attempt to transmit data to the same node or the base station at the same time. As collisions generally necessitate a re-transmittal of data by at least one of the transmitting sensors involved, they result in severe energy wastage. The applicant is aware of a number of multiple access protocol techniques that may be used to prevent data collision. These techniques can broadly be divided into so-called "contention-based" and ''schedule-based" protocols. Examples of schedule-based protocols include time division multiple access ("TDMA"), frequency division multiple access ("FDMA"), code division multiple access ("CDMA") and spatial division multiple access ("SDMA").

In TDMA, each node has a dedicated time slot within which it is allowed to transmit data so that there is no interference and no packet loss. This protocol may, however, result in long delays which are directly proportional to the number of nodes in the network. Moreover, the network has to reschedule the time slots every time a node joins or leaves the network. In -FDMA, each node is allocated a dedicated frequency at which it is allowed to operate. Because specialized frequency operation often results in increased energy requirements, this protocol likewise has its limitations. An additional complication with this protocol is that it is problematic for a sensor to select an operating frequency and at the same time communicate this choice to its neighbouring nodes, CDMA on the other hand is a technique that is often used for intra-cluster communication, in other words, communication inside a cluster between a cluster head and its non-head nodes. CDMA; adds additional computation overhead to each node, as nodes have to both encode and decode signals. As a result, CDMA reduces bandwidth available for information exchange. It is arguable that SDMA may be applied to a WSN due to unplanned deployment of sensors.

In contention-based protocols, such as carrier sense multiple access with collision avoidance ("CSMA/CA"), sensors wishing to transmit data first have to sense (listen to) the carrier channel for a predetermined amount of time before transmitting so as to check for activity on the channel and prevent exposed/hidden terminal problems. IEEE802.11 is a well-defined CSMA/CA protocol and is widely used today. In contrast to schedule-based approaches, contention based protocols can be used in a distributed fashion. However a contention-based protocol such as IEEE802.11 wastes a lot of energy in idle listening which could account for up to- 50% - 100% of the energy required for receiving.

A schedule-based approach, such as TDMA, is generally considered to be more efficient than a contention-based approach in terms of energy saving by switching sensors into sleeping mode. However it does not use the transmission media effectively as each timeslot is dedicated to a sensor and therefore cannot be used by others should it be idle. Moreover a schedule- based approach requires a central node to manage and broadcast the schedule to other members, causing high delay and un-scalability. On the other hand, contention-based protocols can be used in a distributed fashion but contention-based protocols such as IEEE802.1 1 waste a lot of energy in idle listening.;

In its PCT international patent application published under publication number WO2009/112937, the applicant discloses a WSN which combines a localization scheme and distributed routing for obtaining improved energy consumption and delay performance. This is achieved by partitioning the sensing field into cluster areas, each bordered by adjacent radii originating at the base station of the sensing field and arcs on concentric circles centred at the base station. One of the nodes in each cluster is assigned as a dynamic head sensor and holds a routing table of the head sensors associated with adjacent clusters. The routing table is used to determine the path along which information may be transmitted to other nodes or the base stations. In the remainder of this specification the WSN configuration of WO2009/1 12937 will be referred to as "S-Web".

S-Web self-organizes sensors into clusters based on their geographical location without requiring the sensors to have a Global Positioning System (GPS) or actively locate themselves. In addition S-Web enables sensor nodes to route data packets to any destination efficiently while consuming low energy in a decentralized manner. Initially, the Base Station (BS) will send beacon signals for every a degree angle. Sensors that receive the beacons at time slot /^' will measure their signal strength to determine their relative distances to the BS. If T is a predefined distance (which is inversely proportional to the received signal strength),-all sensors that receive beacon signals at angle order ?/^' (=i*a) with signal strength of 6j*T (within sector j) will be in the same group/cluster, denoted as (βϊ, 6j). Figure 1 illustrates a sensor network of 100 nodes distributed in the area of [100X100]. The root of the coordination (0;0) is at the top left corner. The base station is located at a position of (50;150). Scanning angle s is 10 degree and threshold T is 35. Sensors in a cluster will choose a cluster head which has the highest remaining energy to act as a router for the cluster. This task is done locally between sensors in the same cluster. The role of head sensor may be rotated amongst sensors in the same cluster to balance the load and prolong the lifespan of the cluster. Each head sensor has a routing table of size [3x3] which points to head sensors of surrounding clusters. When a sensor needs to transmit data to the base station or to other sensors, if it is a non-head sensor, it will transmit the data to the head sensor of its cluster. The head sensor will then forward the data packet to the head sensor of its neighbouring cluster which is closer to the destination than itself. Closeness to the destination is determined by the difference in angle order and signal strength (distance) order between the neighbouring cluster and destination cluster (or base station), as opposed to that between the current (transmitting) sensor and destination. Even though S-Web does not aim to form a minimal spanning tree from a sensor to the base station or destination sensor, it generally produces short transmission routes. One of the distinctive features of S-Web is that data can be transferred between any sensors without going through the base station and while using only local network knowledge. This prevents bottlenecks at sensors near the base station, makes the WSN more scalable and also prolongs the average lifetime of sensors in the network.

In; the remainder of this specification the terms "node" and "sensor" will be used interchangeably.

OBJECT OF THE INVENTION

It is an object of this invention to provide a wireless sensor network which will at least partly alleviate the problems with existing wireless sensor networks mentioned above.

SUMMARY OF THE INVENTION In accordance with this invention there is provided a method of scheduling communication. between sensors in a wireless sensor network, the. wireless, sensor network including a plurality of sensors spaced apart over a sensing ^■ field with each -sensor grouped into one of a plurality of sensor clusters and each cluster being identifiable by means of a cluster identifier, the method comprising the steps of dividing each period of a periodic control signal configured to control the sensors into at least four time slots; assigning one of the four time slots of the periodic control signal to each cluster, the assigned time slot being determinable from the cluster's identifier; allocating the time slot assigned to each cluster to each sensor grouped within it; configuring sensors to be in an active state during which they may transmit and receive data during the time slots of the periodic control signal allocated to them, and in an inactive state outside them; and configuring sensors to temporarily switch to the active state in alternative time slots of the periodic control signal not allocated to them if they are required to transmit information to other sensors to which such alternative time slots have been allocated. Further features of the invention provide for the method to include the step of configuring the sensors to contend media based on the concept of Carrier Sense Multiple Access with Collision Detection, so as to prevent data collision and use available resources more efficiently, and the steps of allocating the cluster identifiers to the clusters so that each identifier includes ·:. at least first and second spacial orders represented by integer numbers, and associating each of the four time slots of the periodic control signal to each of the four odd and even combinations of the first and second spacial orders. The invention also provides a wireless sensor network comprising a plurality of sensors spaced apart over a sensing field with each sensor being grouped into one of a plurality of cluster areas spanning the sensing field, characterised in that each cluster has associated therewith a cluster identifier, the cluster identifier being communicated to each sensor within its cluster area, in that the network includes a periodic control signal each period of which is divided into at least four time slots, in that each of the clusters and : sensors grouped within it is allocated one of the time slots so that the time slot allocated to each sensor can be determined from the identifier of the cluster area into which it is grouped, and in that sensors are configured to switch into an active state during which they are capable of transmitting and receiving data during the time slot of the periodic control signal allocated to the cluster area into which they are grouped and into an inactive state outside it. Further features of the invention provide for each cluster identifier to include at least first and second integer spacial orders; and for each of the four time slots to be associated with one of the four odd and even combinations of the first and second spacial orders and allocated to the sensors accordingly. Still further features of the invention provide for sensors to be further configured to calculate the time slot of the periodic control signal during which a destination sensor will be in an active state from the destination sensor's cluster identifier and to temporarily switch to the active state and transmit data to it during such calculated time slot; and for communication between the sensors to be controlled by means of a combination multiple access control layer technique comprising both a contention-based and schedule-based protocol.

Yet further features of the invention provide for each cluster area to be bordered by two arcs of two adjacent concentric circles about a common centre and two adjacent radii originating at the centre, the radii being positioned at predefined angular intervals defining angular orders between them and the concentric circles being at predefined distance intervals from the common centre defining distance orders between them; for the angular and distance orders to correspond to the first and second integer spatial orders, respectively; for the common centre to be a base station configured to transmit the angle orders to the sensors; and for the sensors to in turn be configured to receive and store the angle orders they, receive from the base- station and to. determine and store their own distance orders based on the strength of the signal they receive from the base station. Further features of the invention provide for each cluster to have a dynamic head sensor; for the head sensor to be periodically chosen as the sensor in the cluster with the highest residual energy level of all the sensors in the cluster,. the identity of the head sensor being communicated to the other, non-head, sensors in the cluster; for each head sensor to include a routing table containing information of the head sensors of surrounding clusters, the information including cluster identifiers and sensor identifiers of the head sensors of such surrounding clusters; for non-head sensors to be further configured to transmit information to and receive information from the head sensors associated with their clusters; for the head sensors to in turn be configured to transmit information to and receive information from head sensors in other clusters and to transmit information to and receive information from the base station; and for head sensors to be further configured to send information to other head sensors or the base station via a route determined as a function of the difference in first and/or second spacial orders between the head sensors in the routing table associated with an originating head sensor and a destination sensor or the base station, as the case may be.

A still further feature of the invention provides for the schedule-based protocol to be a time division multiple access protocol and the contention- based protocol to be Carrier Sense Multiple Access with Collision Avoidance.

The invention also provides a wireless sensor network comprising a plurality of sensors spaced apart over a sensing field, wherein each sensor is grouped into one of a plurality of cluster areas spanning the sensing field with each cluster area being identifiable by means of a cluster identifier, the wireless sensor: network being characterized in that communication between the sensors is controlled by means of a combination multiple access control layer technique comprising both a contention-based and .schedule-based protocol. .. BRIEF DESCRIPTION OF THE DRAWINGS

The invention will now.be described, by way of example only with reference to the accompanying representations in which: Figure 1 is a diagrammatic representation of the S-Web scheme according to the prior art;

Figure 2 is a diagrammatic representation of a wireless sensor network in accordance with a first embodiment of the invention; Figure 3 is a table exemplifying a cluster distribution in a wireless sensor network in accordance with the invention;

Figure 4A is . a routing table of a head sensor In a wireless sensor network with the cluster distribution as shown in Figure

3;

Figure 4B is a routing table of a second head sensor in the

Wireless Sensor Network of Figure 2;

Figure 5 is a graph illustrating time scheduling for the clusters allocated to four different time slots;

Figure 6 is a diagrammatic representation of a wireless sensor network in accordance with an alternative embodiment of the invention;

Figure 7 is a graph illustrating time scheduling for the clusters allocated to the two different time slots according to the embodiment of Figure 6;

Figure 8 is a table indicating communication performance results between sensors and the base station in accordance with S-Web;

Figure 9 is a graph showing the number of rounds against the percentage of node deaths for S-Web, Direct and SHORT when sensors send data to the base station; Figure 10 is a graph showing the average number of contending messages in each timeslot for different numbers (loads) of concurrent messages in a WSN according to the invention;

Figure 1 1 · is a graph showing the average number of attempts that a node has to contend before being able to send a message successfully in a WSN according to the invention;

Figure 12 is a graph showing the total time it takes to deliver the messages (load) in a WSN according to the invention; and

Figure 13 is a graph showing the number of undelivered messages that occur in various networks as the network load increases.

DETAILED DESCRIPTION WITH REFERENCE TO THE DRAWINGS

In the embodiment of the invention shown in Figure 2, a wireless sensor network ("WSN") (1 ) includes a base station (2) and a plurality of sensors (3) spaced apart over a sensing field (4). The sensing field (4) is divided into concentric circles (5) centering at the base station (2) at predefined distance intervals (7) and is further divided into sectors by radii (6) at predefined angular intervals {a). The sensing field (4) is thus made up of cluster areas (7), each defined between two arcs on two adjacent concentric circles (5) and two adjacent radii (6) with each cluster area (7) uniquely identifiable by means of an angle order ( ?,·) and distance (signal strength) order (Sj) where /^' and j are integer values. Each of the sensors (3) therefore falls within a cluster area (7) and all the sensors (3) within a specific cluster area (7) form a cluster (8) of sensors. Initially, at the setup of the WSN, the base station sends beacon signals for every a degree interval at a time. Each a degree interval therefore corresponds to an angle order β,. Each sensor that receives the beacon signal at time slot / then measures the strength of the beacon signal it received from the base station and determines its relative distances to the base station as a function of the signal strength. The predefined distance T is inversely proportional to the received signal strength. All sensors that receive beacon signals at angle order ?,· (=i*a) with signal strength of δ ^*Τ (within sector j) are therefore in the same cluster, denoted as (β,, δ). In practice certain boundaries will naturally be set up within which a measured signal strength will be considered to fall within a specified cluster. It should, however, immediately be apparent that there may be numerous other ways of assigning angle and distance orders to the sensors. .: The sensors (3) in a cluster (8) then select a cluster head sensor which has the highest remaining energy to act as a router for that cluster (8). The process .of selecting the cluster head is done locally between the sensors (3) in the same cluster (8). When data is transmitted to sensor nodes outside of the cluster from where it is sent, or the base station, the non-head sensors in a cluster first send the data to the cluster head sensor which, in turn, sends it - further along a path towards its destination. It should be appreciated that the cluster head sensor will expend more energy by sending data over potentially longer distances than the non-head sensors in the same cluster. As a result, the role of cluster head sensor is periodically exchanged between the nodes in the same cluster to balance the load and prolong the lifespan of the cluster as a whole.

Figure 3 shows a table representing a typical cluster distribution wherein each cell in the table contains the angle order ( ?), distance order (6) and a node identifier (nodelD) of the cluster head sensors of each of the clusters (8) in the sensing field (4). The distance order of each head sensor therefore represents its distance from the base station. Each head node further has a routing table (RT) which includes information on the head nodes of surrounding clusters. For example, sensor myNode (PmyNode, SmyNode) has a table of surrounding neighbouring head node identifiers and their cluster information (β₅, 6_S). As shown in Figures 4A and 4B, the routing tables include the node identifiers of the head nodes of its surrounding clusters. The routing table- in Figure 4A corresponds to head node number "1" in the table of Figure 3 and the routing table in Figure 4B corresponds to head node number "3" in the table in Figure 3.

The Following is an example algorithm that may be used to insert a : neighbouring sensor (s) into the routing table of myNode of size 3x3.

Procedure lnsert(Sensor myNode, Sensor sj {

//calculate the column index: col

angle_dist = s - βηιγΝο β!

if (angle_dist < 0)col = 2;//right sector (bigger angle)

else if (angle_dist == 0) col = 1; //same angle

else col = 0;//left sector (smaller angle)

//calculate the row index: row

signal_dist = δ_ε-δ myNode!

. if (signal dist < 0)row =2;//bigger signal order

else if (signaljdist == 0) row = 1;//same signal order

else row = 0;//smaller signal order (stronger signal)

RT[col][row] = s;

}

end Insert

In addition to the above, cluster areas are assigned time slots during which they may listen for and receive information. In the sensor network (1 ) of Figure 2, each time window is divided into four equally long time slots. Each time slot corresponds to one of four odd/even combinations of the angle order ( ?,·) and distance order (<£/) namely (even β; even 6), (odd ?; even £); (odd ?; odd rf), and (even /?; odd £). Clusters belonging to the various time slots are indicated in Figure 2 by 4 different types of shading. A node can therefore determine its neighbouring node's wakeup interval by knowing its identifier which, in the current example, includes the scanning angle and distance orders. This can, for example, be done by means of the following Pseudo-code: Procedure findTimeSlot ( Sensor x) {

//find time slot or window that node x wakes up to listen

ίί((βχ % 2 = 0) && (δχ % 2=0))

Slot = t + 4*n^*R;

if (@x % 2 = 1) && (δχ % 2= 0))

Slot = t + (4*n+1)^*R;

if ( x % 2 = 1) && (δχ % 2=1))

Slot = t + (4^*n+2)*R;

if(( x % 2 = 0) && (δχ % 2=1))

Slot = t + (4*n+3)*R;

}

end findTimeSlot

With the above time slot allocation, a sensor node wakes up at its assigned timeslot to listen for incoming data and sleeps (hibernates) at other times in order to conserve energy. This is similar to a schedule based protocol. An important distinction between the invention and other known technologies is that a node can wake up at its next hop neighbour's time slot to contend the media if it has data to send to it, which in turn is similar to contention based protocol. By simply having knowledge of another cluster's angle and distance orders, a transmitting sensor can determine the exact time at which a destination sensor (cluster) will be active (awake) and able to receive data, thus alleviating the need for it to have knowledge of wakeup schedules of other nodes or for such wakeup schedules to be broadcast from a central source.

Figure 5 illustrates an example of time scheduling for the clusters allocated to^; the four different time slots. For convenience, this method of time allocation- and signal contention will be referred to as "Dartboard" in the remainder of this specification. It should be appreciated that the scheduling could be achieved by dividing a periodic control signal, which is visible to all sensors, into the four time slots and allocating one of the time slots to each cluster and the sensors within it. Each time slot may correspond to one of the four odd/even combinations of the angle order ( ?,·) and distance order (<5 ) and may be allocated accordingly.

Figure 6 shows an alternative embodiment of the invention. Features corresponding to those described with reference to Figure 2 above are indicated with like numerals. In this embodiment of the invention the time window (and control signal) during which cluster .head nodes may transmit information is divided into two slots as illustrated by the black (20) and white (22) areas, respectively. Clusters having the sum of their angle and distance orders equal to an odd number are illustrated as black and are configured to listen and receive incoming data at times (f + (2n-1)^*R) where f represents an initial time, n represents an integer number, and R, represents the average time taken to transmit a data packet. Likewise, clusters having the sum of their angle and distance orders equal to an even number are illustrated as white and are configured to listen and receive incoming data at times (t + 2n*R). As explained with reference to Figure 2 above, by having knowledge of a receiving cluster's angle and distance order, a transmitting cluster can determine the exact time at which the receiving cluster will be able to receive data. The sending cluster may then wake up at such time to transmit data, even if during a time slot when it would not normally have been active. Figure 7 illustrates an example of time scheduling for the clusters allocated to the two different time slots according to the embodiment explained with reference to Figure 6. For convenience, this method of time allocation and -signal contention will be referred to as "Checkerboard" in the remainder of this specification.:

In both the above embodiments, when a sensor needs to transmit data to the base station or to another node, it first transmits the data to the head sensor in its cluster, unless it is the head node itself. The head node then inspects its routing table and identifies its neighbouring cluster head node which is closer to the destination than itself. By using the angle and distance orders of the identified neighbouring cluster head node, the transmitting head node can then determine the exact time at which the receiving head node will be active and capable of receiving the data. If the transmitting head node is not in an active time slot itself at the calculated : time, it will wake up, transmit the necessary data: and switch to its hibernating state again after the transmission is complete.

Nodes therefore wake up during their allocated time window and listen or sense the carrier medium for incoming data. The configuration of the invention can therefore be described as a combination between schedule- based (TDMA) and contention based protocols.

The following is a Pseudo-code example for finding a next hop neighbour that is closer to the destination node. In the example, dest represents the destination node, myNode represents the present (transmitting) node, RT is the routing table associated with myNode, β_χ is an angle order of node x and δ_χ is a distance order of node x.

Procedure findNextHop(Sensor myNode; Sensor dest) {

//find next hop in the routing table of current

//node: myNode to the destination sensor dest next = null; //next hop

min_diSt = | fidest -

//equal to the number of hops from myNode to dest

if (min_dist == 0)

next = myNode ;//my Node is the head of dest cluster

else {

//find a sensor -in the routing table that closest

//to dest in terms of angle β and signal δ

or

next = s that has minimum

//next has the lowest number of hops to dest end findNextHop

Routing data from a node to the base station is done differently because the base station has the same angle order with every node. Therefore a node myNode forwards a data packet to its neighbouring cluster head X which has signal order 6_X smaller than the current node SmyNode- The neighbour node X should be in cluster (β myNode, 0_myNode-i) where /^' = 1 , 2, ....

If a head sensor does not have a neighbour in the direction of the base station, it transmits data directly to the base station. However if the head sensor is far from the base station, its energy will be depleted quickly as a result of such action.

The WSN nodes therefore determine next hop routing using local network knowledge, as opposed to having to receive this information from the base station or other external routing calculator. This, decentralized, way of routing results in energy saving. It should also be appreciated that the nodes in the network can send data to any destination rather than just to the base station. Unlike with conventional schemes, where nodes generally have the same wakeup schedules (this is, for example, the case in S-MAC where nodes ideally wake up at the same time), the WSNs according to this invention reduces the number of nodes contending for the media at the same time due - to them having different wakeup schedules. As will be seen from the results provided below, the proposed combination-MAC approach results in fewer data collisions. In addition, sensors in the proposed system can function without having to know and/or broadcast wakeup schedules of all other nodes. This is because this information can be deduced from their identification (i.e. angle and signal strength orders) and a time frame (for example the state of the periodic control signal). Moreover the proposed combination-MAC reduces overhearing overhead due to incoming packets aiming for only nodes that wake up in that timeslot (not in other timeslots). Performance evaluation

To evaluate the S-Web scheme, 100 sensors randomly distributed in a (100 : : X 100)ΊΤΊ² area with the base station located at (50; 150), were evaluated. All nodes were allocated the same initial energy of 1 Joule.

In the first part of this evaluation, S-Web was evaluated in terms of routing, under the assumption that the media is collision-free. In the second part, the performance of a combination-MAC of S-Web in accordance with the invention, taking collisions into account, was evaluated.

(i) S-Web Performance in terms of routing

S-Web routing was compared with direct communication between sensors and the base station, referred as "Direct" and a minimum spanning tree routing algorithm rooted at the base station, which is referred to as "SHORT". The scanning angle a was selected as 10 degrees and the signal strength threshold T as 30 m, in order to keep the total number of clusters smaller than 30 percent of the total number of nodes. The table in Figure 8 summarises communication performance results between sensors and the base station. From Figure 8 it is evident that S-Web consumes the least energy whereas "Direct" consumes the most energy at 2619μ J per message (where each message consists of 2000 bits). S-Web also transmitted messages with a lower average number of hops and energy consumption per message than SHORT does. As expected "Direct" has the lowest message hop count because data is sent directly to the base station. To evaluate the WSN lifespan, random sensors that have data to send were simulated using one round as a unit of measure. A round was defined as having been completed when 1000 messages had reached their destinations. The table shown in Figure 9 shows the number of rounds against the percentage of node deaths for S-Web, Direct and SHORT when sensors send data to the base station. As can be seen from Figure 9, S-Web outperformed both the other- strategies and showed the longest overall: lifespan. It is particularly noteworthy that S-Web outperformed SHORT both in terms of energy consumption and delay.

(ii) Combination MAC performance of S-Web in accordance with the invention In order to evaluate the performance of the MAC layer, different numbers of concurrent messages (load) ranging from 1 to 50 were sent from random nodes to random destinations.

Figure 10 shows the average number of contending messages in each timeslot for different numbers (loads) of concurrent messages. As expected, the higher the load in the network (in other words the more concurrent messages that had to be transmitted) the more messages have to contend for available transmission resources. As can be seen in Figure 10, both embodiments of the proposed combination-MAC (in other words Dartboard and Checkerboard) of the invention resulted in a lower number of contending messages than S-MAC. This can mainly be ascribed to the fact that all the nodes in S-MAC have the same wakeup schedule.

Figure 11 shows the average number of attempts that a node has ίσ contend before being able to send a message successfully. S-MAC shows the highest number of attempts due to the higher number of contenders that contest to transmit their messages at the same time. It is estimated that S-MAC, Checkerboard -and Dartboard have approximately 100%, 50% and 25% of their nodes active (awake) at the same time, respectively.

Figure 12 indicates the total time it takes to deliver the messages (load) in the network. Both Checkerboard and Dartboard again outperforms S-MAC in this regard by delivering the information to their destination nodes-faster.

Figure 13 shows the number of undelivered messages: that occur in the various networks as the network load increases. A message will, for example, be considered to be undelivered if a node was unable to transmit the message for more than a predefined number of consecutive attempts (i.e. 10). It is again clear that S-MAC works well only when the load in the network is low. As can be seen from Figures 10 to 13, Dartboard consistently outperforms both Checkerboard and S-MAC schemes. Energy consumption was not plotted in any of the cases because of the high number of messages that were undelivered. It is, however, expected that the more attempts needed to deliver the messages (as shown in Figure 11 ), the higher the overall energy consumption in the network will be. The invention therefore provides a WSN that uses spatial TDMA for each cluster in a scalable and self-determined manner based, for example, on S- Web. Each node has the capability of determining wakeup periods of other - nodes based on their cluster identifiers (angle and distance orders). When awake, a node has to sense the carrier and contend the media based on the idea of CSMA/CD to prevent data collision and use the media more efficiently. The preliminary results show that the proposed combination -MAC approach of the invention results in lower delays than S-MAC. It also saves more energy consumed in overhearing by minimizing the number of active nodes at a time while not introducing extra schedule broadcast overhead. The invention addresses challenging problems for WSNs at both routing and MAC layers and allows WSNs to not only prolong their life spans, but also ·. ^'-^■- minimize energy consumption for each node_; In addition, it also minimizes delivery delays.

It should be appreciated that the above examples are by way of example only and that numerous changes and modifications may be made to the invention without departing from the scope thereof. In particular, it is envisaged that the invention may also be applied to networks in which sensors are divided into clusters in a Cartesian coordinate plane as opposed to the Polar coordinate plane described in the above examples. In such a network the identifier of each cluster may have a first spacial order which could correspond to a horizontal spatial order and a second spacial order which could correspond to a vertical spatial order. Of importance is that the sensing field is divided into first and second spatial orders which may both be represented by odd and even integer numbers which are allocated to clusters and the sensors within them. The invention therefore provides a combination approach between contention and schedule based protocols. Using spatial division, each cluster is assigned a timeslot during which sensors grouped within it will be in an active state. Within each timeslot, nodes wishing to transmit data to an active node have to perform CSMA/CD procedures. In doing so, a node can be inactive outside of its own designated timeslot and therefore consume less energy when compared to traditional CSMA/CD. In terms^; of delay, the combination technique implies lower delays than would otherwise have been the case had TDMA been used on its own.

Claims

CLAIMS:

1. A method of scheduling communication between sensors (3) in a wireless sensor network (1 ), the wireless sensor network including a plurality of sensors (3) spaced apart over a sensing field (4) with each sensor (3) grouped into one of a plurality of sensor clusters (7) and each cluster being identifiable by means of a cluster identifier, the method comprising the steps of: (a) dividing each period of a periodic control signal configured to control the sensors (3) into at least four time slots;

(b) - assigning one of the four time slots of the periodic control signal to each cluster (7), the assigned time slot being determinable from the cluster's identifier;

(c) allocating the time slot assigned to each cluster (7) to each sensor (3) grouped within such cluster;

(d) configuring sensors (3) to be in an active state, during which they can transmit and receive data, during the time slots of the periodic control signal allocated to them and in an inactive state outside them; and

(e) configuring sensors (3) to temporarily switch to the active state in alternative time slots of the periodic control signal not allocated to them if they are required to transmit information to other sensors to which such alternative time slots have been allocated.

2. A method as claimed in claim 1 which includes the step of configuring the sensors to contend media based on the concept of Carrier Sense Multiple Access with Collision Detection so as to prevent data collision and use available resources more efficiently.

A method as claimed in claim 1 or claim 2 which includes the steps of allocating the cluster identifiers to the clusters (7) so that each identifier includes at least first and second spacial orders represented by integer numbers. and associating each of the four time slots of the periodic control signal to one of the four odd and even combinations of the first and second spacial orders.

A wireless sensor network (1 ) comprising a plurality of sensors (3) spaced apart over a sensing field (4) with each sensor (3) being grouped into one of a plurality of cluster areas (7) spanning the sensing field (4), characterised in that each cluster area (7) has associated therewith a cluster identifier, the cluster identifier being communicated to each sensor (8) within such cluster area, in that the network includes a periodic control signal each period of which is divided into at least four time slots, in that each of the clusters (7) and sensors (8) grouped within it is allocated one of the time slots so that the time slot allocated to each sensor (3) can be determined from the identifier of the cluster area (7) into which it is grouped, and in that sensors are configured to switch into an active state, during which they are capable of transmitting and receiving data, during the time slot of the periodic control signal allocated to the cluster area (7) into which they are grouped and into an inactive state outside it.

A wireless sensor network (1 ) as claimed in claim 4 in which each cluster identifier includes at least first and second integer spacial orders and in which each of the four time slots is associated with one of the four odd and even combinations of the first and second spacial orders and allocated to the sensors (3) accordingly.

A wireless sensor network (1 ) as claimed in claim 4 or claim 5 in which sensors (3) are further configured to calculate the time slot of the periodic control signal during which a destination sensor will be in an active state based on the destination sensor's cluster identifier and to temporarily switch to the active state and transmit data to it during such calculated time slot.

A wireless sensor network (1 ) as claimed in any one of claims 4 to 6 in which communication between the sensors is controlled by means of a combination multiple access control layer technique comprising both a contention-based and schedule-based protocol.

A wireless sensor network (1 ) as claimed in any one of claims 4 to 7 in which each cluster area (7) is bordered by two arcs (5) of two adjacent concentric circles about a common centre (2) and two: adjacent radii (6) originating at the centre (2), the radii (6) being positioned at predefined angular intervals (a) defining angular orders between them and the concentric circles being at predefined distance intervals (7) from the common centre (2) defining distance orders between them, and in which the angular and distance orders correspond to the first and second integer spatial orders, respectively.

A wireless sensor network (1 ) as claimed in claim 8 in which the common centre (2) is a base station configured to transmit the angle orders to the sensors (3), the sensors in turn being configured to receive and store the angle orders they receive from the base station (2) and to determine and store their own distance orders based on the strength of the signal they receive from the base station.

A wireless sensor network (1 ) as claimed in any one of claims 4 to 9 in which each cluster (7) has a dynamic head sensor, the head sensor being periodically chosen as the sensor (3) in the cluster with the highest residual energy level of all the sensors (8) in the cluster, the identity of the head sensor being communicated to the other, non-head, sensors in the cluster.

11. A wireless sensor network (1 ) as claimed in claim 10 in which each head sensor includes a routing table containing information of the head sensors of surrounding clusters, the information including cluster identifiers and sensor identifiers of the head sensors of such surrounding clusters.

12. A wireless sensor network (1 ) as claimed in claim 10 or claim 11 in which non-head sensors are further configured to transmit information to and receive information from the head sensors associated with their clusters, head sensors in turn being configured to transmit information to and receive information from, head sensors in other clusters and to transmit information to and receive information from the base station (2). .

: ^. , · ^■. · -

13. A wireless sensor network (1) as claimed in any one of claims 10 to 12 in which head sensors are further configured to send information to other head sensors or the base station (2) via a route determined as a function of the difference in first and/or second spacial orders between the head sensors in the routing table associated with an originating head sensor and a destination sensor or the base station, as the case may be.

14. A wireless sensor network (1 ) as claimed in claim 7 in which the schedule-based protocol is a time division multiple access protocol and the contention-based protocol is Carrier Sense Multiple Access with Collision Avoidance.

15. A wireless sensor network (1 ) comprising a plurality of sensors (3) spaced apart over a sensing field (4), wherein each sensor (3) is grouped into one of a plurality of cluster areas (7) spanning the sensing field (4) with each cluster area (7) being identifiable by means of a cluster identifier, the wireless sensor network (1 ) being characterized in that communication between the sensors (3) is controlled by means of a combination multiple access control layer technique comprising both a contention-based and schedule-based protocol.