Energy efficient wireless sensor network and a method for arranging communications in a wireless sensor network

Description

TECHNICAL FIELD

[0001] The invention relates generally to wireless sensor network technology. First of all, the invention relates to the optimization of the overall energy consumption of a wireless sensor network, mainly by finding communication protocols and systems that provide an effective balance between the requested information transfer and the energy used to send and receive information.

STATE OF THE ART

[0002] A wireless sensor network (WSN) is a special case of wireless networks, its peculiarity basically includes very strict requirements for minimizing the physical size and energy consumption of node devices and the specialized roles performed by the nodes: the vast majority of nodes are sensors that collect information and relay it to specific data sink nodes, of which there are only a few, and can act as gateways that pass the collected information to other networks or systems. Wireless sensor networks often have an ad-hoc structure, meaning that nodes can appear and disappear or move from one part of the network to another, and the network itself has to adapt to successive changes in topology and connectivity, often forcing multi-routing. hop. Transmission rates in sensor networks are typically low, at least compared to the Mbit / s data rates in communication networks between computers. In general, actuators can also be viewed as nodes; according to the adopted convention, the term "sensor" in wireless sensor networks should be understood as real sensors and actuators.

[0003] Wireless sensor networks and their nodes are generally known from many prior art publications. US 2004/0100917 A1 discloses a method for selecting a coordinating device, the purpose of which is to ensure that no part of the ad-hoc wireless sensor network becomes disconnected and to minimize the overall amount of energy required to ensure communication through the network. The solution disclosed therein is based on the distribution of the initiating message over the network after which each node applies a random delay before transmitting the "coordination status request" message. A subsequent publication, US 2003/0152041 A1 discloses certain functions at the general level of wireless sensor networks including hierarchical allocation of nodes to multiple levels of nodes and keeping the node asleep at any time other than node initiation or task completion.

[0004] US 2002/0044533 A1 discloses the drawbacks of relying on the accurately known spatial position of each node, forcing each node to have a Global Positioning System (GPS) receiver. As a more advantageous alternative, this publication discloses a system in which each node learns about the full set of other nodes it can communicate with, but maintains active communication with only a subset of these nodes, which translates into a way to configure and maintain the network topology independent of position. Another publication CA 2 311 245 A1 relates to the division of nodes into two hierarchical levels such that each higher level node manages a cluster of neighboring lower level nodes, and more power-consuming "trunk" communication is only necessary between higher level nodes.

[0005] WO 01/69279 discloses a wireless sensor network in which each node has its own locating device and the nodes are able to exchange both location information and reconnaissance data. The system is mainly intended for military reconnaissance purposes. The subsequent publication WO 01/26329, which belongs to a very large family of related patent applications, discloses a great many details which, at the time of writing this specification, are viewed as the known art of wireless sensor networks. Another prior art publication, US 6,208,247 B1, focuses on the physical implementation of nodal wireless sensor network devices.

[0006] A major source of problems on the path to minimized power consumption in wireless sensor networks has traditionally been communication protocols that determine the amount and nature of wireless transmission between nodes. The fact that the network has to dynamically adapt to the appearance and disappearance of nodes and other changes in the network topology means that the communication protocol must include enough procedures to recognize currently available communication possibilities with other nodes and rules for determining the order in which currently connected nodes will communicate with each other. The communication protocol should have a certain degree of scalability, which means that it should facilitate cost-effective communication, no matter how many nodes there are in the network. Moreover, the communication protocol should provide a certain required minimum level of throughput, i.e. the amount of information that can be transferred to a specific location via the network in a given unit of time. The most essential part of the communication protocol is the MAC (Medium Access Control) part.

[0007] Known wireless sensor network protocols include Sensor-MAC (also known as S-MAC), Timeout-MAC (T-MAC), and IEEE 802.15.4 Low Rate Wireless Personal Area NetWork (LR-WPAN). Of these, S-MAC is described in W. Ye, J. Heidemann and D. Estrin: "Medium access control with coordinated, adaptive sleeping for wireless sensor networks, ACM / IEEE Trans. Networking, Vol. 12, No. 3, pp. 493-506, June 2004. It uses the general window structure in the CSMA (Carrier Sense Multiple Access) MAC scheme. Nodes are set to periodically be awake and asleep, which significantly reduces energy consumption compared to conventional CSMA solutions. The S-MAC window consists of a beacon type synchronization transmission, a fixed length (300 ms) of active time for data exchange, and a sleep time to the end of the window. Each node wakes up at the beginning of the window, and each node that wants to send data performs CSMA / CA channel access in accordance with RTS / CTS (Request To Send / Clear To Send). The window length is a defined and static MAC parameter ranging from 500 ms to 10 s. In the latest implementation, the S-MAC protocol runs a 10 s network scan every 2 minutes. As you can see, the network scan consumes a large amount of energy.

[0008] The protocol is similar to S-MAC, but energy efficiency is improved by dynamically setting the length of the active period. The node goes to sleep if it does not receive any information from the channel within 15 ms. Contrary to S-MAC, the window length is up to 610 ms. TMAC performs an occasional network scan of 610 ms. This short scan time keeps energy consumption to a minimum.

[0009] The IEEE 802.15.4 LR-WPAN standard belongs to the IEEE 802.15 family of WPAN standards and is described in "Wireless Medium Access Control (MAC) and Physical Layer (PHY) Specifications for Low-Rate Wireless Personal Area Networks (LR-). WPANS) ”, IEEE Std 802.15.4 2003 Edition. It uses Access

- 3 channel CSMA / CA with the optional use of a super frame structure that resembles the S-MAC window structure. The LR-WPAN includes orchestrators that provide synchronization services and conduct data over the network, and devices that can only communicate with the coordinators. Coordinators control their superframe structures by sending navigators at the beginning of each superframe. The navigator is followed by a CAP (Contention Access Period) during which network nodes can send data and requests to the coordinator using CSMA / CA. The optional, dedicated GTS (Guaranteed Time Slots) job type after the CAP reduces contention and delay. The CAP and GTS may optionally be followed by a time of inactivity until the next navigator, during which the nodes may switch to sleep mode. The navigator element period and superframe length vary from 15.4 ms to 252 s, which enables a swap between delay and power consumption. The LR-WPAN nodes perform a periodic network scan on a set of RF channels. Each channel is received for the duration of the beacon or until receipt of a predetermined number of beacons. The least preferred scan time exceeds 67 minutes when the maximum duration of the navigator and all 16 channels defined for the 2.4 GHz frequency band are scanned. This corresponds to the energy of several million transmissions of navigation elements. As you can see, the energy used to scan the network can be very significant,

[0010] A further improvement of the above-described LR-WPAN is known as ZigBee and is described online on the official website of the ZigBee alliance (http.7 / www.zigbee.org). It includes network and security layer definitions and application profiles, and supports access control lists, packet refresh timers, and some encoding standards.

[0011] Another suggested protocol is SMACS (Self-Organizing Medium Access Contro! For Sensor Networks) and is described in the publication by K. Sohrabi, J. Gao, V. Ailawadhi and GJ Pottie: "Protocols for selforganization of a wireless sensor network, IEEE Personal Communications, Vol. 7, No. 5, pp. 16-27, October 2000. It allows nodes to discover their neighbors, based on links, and sets a schedule! transmission and reception without the need for head nodes. The network uses P2P topologies with FDMA (Frequency Division Multiple Access), where each link operates on different RF channels. The SMACS protocol relies on wireless fixed nodes. However, the algorithm has an extension for mobility management thanks to the EAR (Eavesdrop-And-Register) algorithm, which enables the interconnection of mobile nodes in the area of ​​fixed wireless nodes. SMACS enables quite high energy efficiency due to the scheduled data exchange windows. The disadvantages of this solution are the need to ensure high parameters for each network node and to handle limited mobility.

[0012] LCA (Linked Cluster Architecture) is a solution that has been known for more than twenty years. It was first described in the scientific publication by D. Baker and A. Ephremides: "The architectural organization of a mobile radio network via a distributed algorithm, IEEE Trans. Communications, Vol. 29 No. 11, pp. 1694-1701, November 1981. Improves scalability by organizing the network into a set of clusters, each with a cluster header that acts as a local controller. The remaining nodes are either regular or gateway nodes, both of which communicate directly with a plurality of cluster headers. LCA uses MAC TDMA (Time Division Multiple Access) with dedicated time slots for each node. The regular data transfer is suspended periodically due to the steered phase to implement a distributed clustering algorithm during which neighboring nodes are discovered and the logical function of the node is selected. In LCA, the nodes register

-4i store information about its intermediate environment, which makes the protocol quite scalable. A disadvantage is that LCA uses a global TDMA frame structure which requires a global clock. Also, since nodes are only aware of their neighboring nodes, multi-hop routing is not supported.

[0013] Another protocol is LEACH (Low-Energy Adaptive Clustering Hierarchy), described in W. Heinzelman, A. Chandrakasan and H. Balakrishnan: "An application-specific protocol architecture for wireless microsensor networks," IEEE Trans. Wirełess Communications, Vol. 1, No. 4, pp. 660-670, October 2002. It is a dedicated MAC allocation protocol with a cluster topology. LEACH extends the network hierarchy with a base station that acts as the core of the network. The cluster headers and the base station only use direct communication. Thus, a star topology is used on two hierarchical levels. The LEACH protocol improves overall energy efficiency by allowing virtually all nodes to transmit short distances, with cluster headers communicating on more energy only with the base station. Consequently, the cluster headers have a shorter battery life than the other nodes, which limits the overall network life. To distribute energy consumption more evenly, LEACH proposes a random rotation of the cluster headers. A disadvantage of this is that the scalability of the network is limited by the range and parameters of the base station.

[0014] All the known suggested protocols have drawbacks that make them unsuitable for use as communication protocols for a very low power consumption wireless sensor network. Many of them require too many transmissions to be considered a true low-energy solution. In the case of others, the time it usually takes a node to scan the network is far too long. One problem with many other advanced protocols is that they require complex hardware on each node, for example for repeatedly measuring and storing multiple RSS values! (Received Signal Strength Indicator).

[0015] Moreover, US-B1-6304556 discloses an Ad-hoc wireless network using node clusters.

SUMMARY OF THE INVENTION

[0016] The object of the invention is to provide a communication solution for a wireless sensor network that will make it possible to reduce the overall energy consumption of nodes at a level where they will not necessarily require batteries or replenished or replaced energy sources, but will be able to obtain the necessary energy from secondary sources. such as vibration, ambient light, temperature variations and the like. Another object of the invention is to provide a communication solution that allows efficient network management, for example avoiding excessively long scan times. Another object of the invention is to provide a communication solution that allows the selection of the transmission power and / or the determination of the transmission distance without having to measure the received signal power level or a similar explicit value.

[0017] In order to achieve the objects of the invention, several aspects and points of view have been considered. The preferred communication resource allocation system consists of a combination of frequency channeling and time slot partitioned channel access, the latter term being a combination of MAC window multiple access and time partitioning, and further comprising a combination of random access and bookable windows. The energy management is preferably based on transmitting predetermined beacons on at least two consecutive power levels such that

The receiving device was able to deduce what is the minimum necessary power level for transmission on a given wireless connection. Unnecessary network scanning is reduced by the forwarding of so-called idle time-marking beacons to the next valuable reception time. The order of operation of the role that requires the most energy among the network nodes can be circulated.

[0018] The wireless sensor network according to the invention is characterized by the functions described in the descriptive part of claims 11, concerning wireless sensor network.

[0019] Additionally, the invention relates to a method for providing communication in a wireless sensor network, the method characterized by the functions recited in the descriptive part of independent claim. 20, regarding the method.

[0020] The wireless sensor network according to the invention has a topology that resembles a group of connected stars. Multiple subnodes communicate with the master node, thereby forming a cluster. Communication between clusters takes place thanks to connections between the main nodes of the clusters. Each cluster has a specific communication frequency used during the transmission of information between subnodes and the cluster master node. Each cluster uses a multiple access scheme thanks to the cluster specific time slots, including both random access time windows and bookable windows. In addition, a network specific signaling frequency that is used in transmitting network beacon signals is provided.

[0021] There are two types of beacons. Each head node transmits network beacons on a network specific signaling frequency according to a specified schedule. In addition, each head node transmits the cluster beacons on a cluster specific communication frequency according to a specific schedule. Most preferably, the beacon transmission consists of two instances of a navigation frame consecutively, each instance being transmitted at a different (but defined) power level. The receiving node learns which transmit power level is the lowest at which it can still receive the beacon frame correctly and uses this information to judge the distance between itself and the node that sent the beacons and / or the default power level it should use to transmit information in the direction opposite to the node.

[0022] Each of the clusters has a so-called access cycle, which is a consecutive period of time with a defined time for active communication (TDMA-based) and a relatively long idle time. The division into communication period and idle period is generally known as division into MAC time windows. If a new subnode wants to join a specific cluster or if the head node of another cluster wants to configure a connection with the head node of a given cluster, it must first perform a network scan, i.e. receive continuously or according to a specific reception schedule until it knows the time and frequency used for transmission in a specific cluster. The required continuous reception period can be relatively long if the idle period is continuous and takes up most of the access cycle. It is preferable to transmit "idle beacons during the idle period, informing at least when the next communication time will come." If each root node transmits beacon signals at regular intervals ("beacon periods), the continuous receipt of one beacon for a period of time should ensure that the time and frequency of all neighboring clusters are known. The root node may wish to scan the network for the entire duration of the navigation signal, while the subnode (or the low energy reserve root node) only needs to

Continuously receive, until it receives another beacon transmission that will be either an actual beacon transmission or an idle one. In any event, a useful maximum time limit for network scanning is one beacon period.

[0023] To work as a master node, the node device needs more expandability than when operating in a sub-node mode. Therefore, it is logical to provide two kinds of nodal devices, some of which would connect to the wireless sensor network as sub-nodes, while others would take over the responsibility of the main nodes. On the other hand, since working as a master node requires more energy than working in a subnode mode, and since the available energy resources can be roughly equal for all node devices (especially when the nodes draw energy from their environment), it is preferable that the nodes can circulate the responsibility of the root node among itself. This is only possible when there is "a surplus of nodes that are able to act as the head node. The wireless sensor network according to the invention may comprise a mixture of two kinds of nodes, or alternatively only one type of nodes, which are able to act as head nodes if necessary.

[0024] The new functions which are regarded as characteristics of the invention are set out in particular in the appended claims. The invention itself, both in terms of its structure and method of operation, with additional objects and advantages, will be better understood from the following description of specific embodiments, read in conjunction with the accompanying drawings.

[0025] The embodiments of the invention disclosed in this specification are not to be construed as limiting the applicability of the appended claims. The verb "includes", as used in the description, means that it admits without limitation the possibility of other unquoted functions. The functions set out in the appended claims may be combined as required, unless otherwise stated.

DESCRIPTION OF DRAWING FIGURES [0026]

Fig. 1 shows an exemplary wireless sensor network topology according to an embodiment of the invention, Fig. 2 shows the features of a wireless sensor network protocol according to an embodiment of the invention, Fig. 3 shows an access cycle and superframe structure according to an embodiment of the invention, Fig. 4 shows a schematic energy distribution used for various purposes, Fig. 5 shows various possibilities for optimizing the navigation speed, Fig. 6 Fig. 7 shows other possibilities for optimizing navigation speed, Fig. 7 shows still other possibilities for optimizing navigation speed, Fig. 8 shows frame structures of a wireless sensor network according to an embodiment of the invention, Fig. 9a shows aspects of network scanning according to an embodiment of the invention, Fig. 9b illustrates aspects of a node association according to an embodiment of the invention, Fig. 10 shows a data transfer in a wireless sensor network according to an embodiment of the invention, Fig. 11 shows an exemplary node device according to an embodiment of the invention, Fig. 12 shows a software architecture in a node device according to an embodiment of the invention, Fig. 13 shows certain root nodes involved in minimizing delay 14 shows the transmission schedules of the nodes shown in Fig. 13.

DETAILED DESCRIPTION OF THE INVENTION

[0027] Various aspects of the invention will be discussed in detail below.

NETWORK TOPOLOGY

[0028] Fig. 1 is a graphical illustration of a topology of a wireless sensor network 100, according to an embodiment of the present invention. Multiple nodes acting as root nodes are provided, represented by black circles, for example, nodes 101 and 111. Each of the root nodes has one or more subnodes 102 or 112, represented by little white circles, which communicate directly with the head node. Together, the head node and subnodes that communicate directly with a given head node constitute cluster 103 or 113. Communication between the clusters occurs through peer-to-peer peer connections between the head nodes. Multi-hopping is supported, which enables communication between any pairs of nodes in the network.

[0029] Some of the nodes may act as exit nodes, meaning that they are information users (while other nodes are only information producers) and may provide gateway connections to other systems and / or other networks. The exit nodes are shown as large white circles, for example the exit node 104. It is preferable - but not necessary - for the exit node to act as a root node in a network topology. The outlet node can be e.g. an actuator or data concentrator that collects information produced by sensor-type subnodes and transfers it to the main processor. It is possible for the exit node to send requests to the network and collect only the information that is currently of interest to the actuator, user or other party. To reduce congestion on the transmission paths leading to the exit nodes, it is advantageous to build a tendency to sum data in the internal network. The exit node is not excluded from the ability to produce information; in other words, the division into producing nodes and using nodes need not be final. In addition, keep in mind that the exit nodes are not essential for wireless sensor network operation; it is also possible for the outlet junction to protrude! on the web only from time to time to collect the information gathered.

[0030] Unlike the cells of cellular radio systems, the cluster is not ultimately dedicated to a given area; and there are no contraindications whatsoever to ensure extensive or continuous geographic coverage. The ability to communicate is only necessary in the areas where there are nodes, and on the other hand the nodes are adapted to provide the required communication capability without external configuration. The number of subnodes in any cluster may change dynamically, new clusters may arise, old ones may be dissolved or split, and the "backbone network" of connections between the main nodes may change its topology, depending on which node device is selected for head node. The wireless sensor network according to an embodiment of the invention thus configures itself automatically and dynamically adapts to changes such as the appearance or disappearance of head nodes and subnodes, changes in the physical location of nodes, changes in signal propagation conditions between nodes, and so on. Further differences from cellular radio systems relate to the MAC protocol according to an embodiment of the invention which will be described in detail below.

Regarding the capabilities and functionality of nodal devices, it can be assumed that there are so-called RFDs (Reduced Functionality Devices) that are only able to act as a subnode and FFDs (Full Functionality Devices) that can act both as subnodes and and head nodes. The designations "RFD and" FFD refer to the IEEE 802.15.4 LR-WPAN standard, so their use in this case should be viewed as a description of an exemplary parallelism, without the requirement that the RFD and FFD of the networks according to the invention are exactly the same as in the cited standard. . The functions that the FFD must have outside of the RFD functions are mainly related to data routing and summation, which will be described in detail later. There must be at least one FFD in each cluster. If there are more, the FFDs can delegate to each other the responsibilities of working as the cluster root node. In the extreme, all node devices in the network may be FFD units.

MEDIA ACCESS CONTROL IN A STACK OF PROTOCOLS

[0032] A preferred protocol stack that is suggested to be used in a wireless sensor network according to an embodiment of the invention includes five layers as shown in Figure 2. At the bottom of the stack is a physical layer 201 that provides methods for bit-exchanging in the wireless medium. Above, a MAC layer 202 is provided, which will be described in detail later. Going upstream from MAC layer 202, a routing layer 203 is provided which manages multi-hop data communication between head nodes. Layer 203 is not used as a subnode. Transport layer 204 performs data fragmentation during transmission and defragmentation during reception, and typically also has the task of handling overall communication between the data source and destination. Application layer 205, which in sensor networks is closely related to the sensor application, depends on the logical function of the node. In a subnode, the application layer consists of the sample acquisition and preprocessing functions. In the master node, data fusion and resource distribution functions are included.

[0033] Details of the MAC layer 202 are shown on the right side of Fig. 2. SAP (Sen / ice Access Points) provides functional interfaces to access MAC protocol and physical layer services. Management procedures 211 are performed on demand to establish and maintain the network topology and manage the data transfer. Signaling of vertical management with routing 203 and physical layers 201 is handled by MSAP (Management) 212 and PMSAP (Physical Management SAP) 213, respectively. User data processing functions are grouped into frame 214 control, queue control 215, error control 216 and access channel 217. Access to the higher layer (routing 203) is possible via DSAP (Data SAP) 218 ​​and the transceiver of the physical layer 201 via PDSAP (Physical Data SAP) 219. An exemplary implementation of the error control function 216 is to remove the 16-bit CRC (Cyclic Redundancy) Check) on successful acknowledgment, and retransmission if confirmation is requested but not received within the specified time.

[0034] MAC frames are built (while transmitting) and dismantled (while receiving) by a frame building function which also manages the addressing. The usable payload of user data is queued by priority. Error control function 216 performs detection, acknowledgment, and retransmission. The channel access function 217 controls the transmission and reception of frames on the RF channels using the access cycle time and frequency assignment function. The power control function selects either a high or a low TX power level according to the transmission distance. Frame priorities are controlled by the access priority function.

[0035] Even though the traditional concept of Open Systems Interconnection (OSI) modeling implies that the protocol layers must be relatively independent of each other so that they can only communicate with each other via well-defined interfaces between the layers to minimize power consumption, more the preferred solution is to consider the possibility of using the protocol stack as a whole. The most important decisions that directly relate to energy savings are made on the MAC layer so that the remaining layers can be designed to adapt to and use the MAC layer peculiarities to the maximum extent possible. For example, other protocol layers should generate as little overhead traffic as possible, and such overhead traffic should adapt to data rate, delay, frame size, and other constraints imposed by the MAC layer. The restrictions will primarily affect the routing protocol. It is not possible to constantly change and update the long routing tables. The optimal paths to all possible destinations cannot be kept in memory all the time for all root nodes. Most preferably, a reactive routing protocol will be used which stores in memory a sequential hop address for each necessary destination.

ACCESS CYCLE AND WINDOW STRUCTURE

[0036] The base unit of time in a cluster transmission schedule is the access cycle. In Fig. 3, the access cycle 301 includes a superframe 302 and an idle period 303. The relative length of superframe 302 has been exaggerated for the purpose of illustrating more clearly in Fig. 3; although the length of the access cycle 301 is a system parameter and may vary depending on the requested bandwidth and the delay value, superframe 302 typically occupies a smaller relative portion of the superframe than shown in Fig. 3. For example, the length of an access cycle 301 may be from 1 to 10 seconds, while the suggested superframe length 302 is 260 ms.

[0037] The super frame 302 includes a number of windows. The invention does not limit the number of windows in a superframe, although the preferred value is 13 windows each 20 ms in length in the exemplary 260 ms superframe shown. According to an embodiment of the invention, the first window 311 of the super frame is a cluster navigation window in which the cluster root node transmits the cluster beacon. Of the remaining windows, it is preferable to designate several of them as random access windows or contention windows, shown in hatched form in Fig. 3, an example of which is box 312. Since the ALOHA protocol is a very well known example of ordinary contention based transmission, these windows can be also called ALOHA windows. An example of the number of ALOHA windows is four. The remaining windows, for example the windows 313, are bookable windows. The most preferred embodiment of the invention includes both ALOHA and bookable windows in a superframe, the invention generally does not exclude superframes which only have windows of one type, in addition to the cluster navigation window. Bookable windows are superior to ALOHA windows in terms of energy efficiency as only the assigned bookable windows have to be received, which argues in favor of having a superframe with more bookable windows than ALOHA windows. It is possible to confidently predict a transmission event in the ALOHA boxes, which means that they must be received in full.

[0038] The planned use of the ALOHA windows and the bookable includes mapping the high priority traffic to the bookable time windows in the MAC protocol, ensuring certainty and consistent bandwidth. For normal priority traffic, used

- 10 are ALOHA time windows. ALOHA windows do not require reservation, and frame collisions can only occur as a result of contention with other nodes. Acknowledgments and fast MAC retransmissions can be used for two traffic classes, which enables the exchange between confidence and energy consumption. The traffic priority and confidence classes as well as their mapping to time windows are presented in the table below.

<td rowspan="2">Traffic priority and reliability class</td><td colspan="2">Time slots</td><td colspan="2">Confirmations</td>

<td>bookable</td><td>ALOHA</td><td>Included</td><td>Off</td>

<td>High priority, high confidence</td><td>X</td><td></td><td>X</td><td></td>

<td>High priority, normal confidence</td><td>X</td><td></td><td></td><td>X</td>

<td>Normal Priority, High Confidence</td><td></td><td>X</td><td>X</td><td></td>

<td>Normal priority, normal confidence</td><td></td><td>X</td><td></td><td>X</td>

Ό039] Data frames are arranged in two queues for high-priority and normal traffic. The frames in the queue are arranged based on the confidence classes such that the high confidence class is ahead of the normal certainty class. In addition, ALOHA time windows may be used for high priority traffic when the high priority traffic queue becomes too long. Bookable time slots are assigned to nodes based on node priorities. Head nodes are assumed to transmit more relevant data than sub nodes and have high priority. Normal priority nodes are usually assigned to subnodes.

[0040] In general, it is possible to decide to transmit the cluster beacon elsewhere within the super frame than the very beginning. The start of a super frame with a cluster beacon, however, has some advantages. The other nodes can easily synchronize with the superframe window structure when it starts with the cluster beacon. In addition, since the cluster beacon preferably includes more recent window allocation information related to the bookable windows, it is advantageous for other nodes to receive this information before the windows used for data exchange occur.

[0041] Additionally, it has been found advantageous to perceive each window such that it consists of a first half and a second half. Regarding the cluster beacon 311, the first half 321 is for transmitting the beacon frame cluster at the first power level (in this case, the high power level) and the second half 322 is for transmitting a substantially identical copy of the same cluster beacon frame at the second power level (in in this case, low power level). The use of different power levels serves to estimate the distance between nodes and the energy necessary to accomplish a transmission as detailed further below. Half of the ALOHA windows and the reserved windows are assigned to uplink and downlink transmissions. In this example, first half 323 of the bookable window 313 is half uplink and second half 324 is half downlink.

[0042] By ensuring that uplink and downlink halves (or more generally uplink and downlink instances) follow each other very quickly and in this order, it facilitates the selection of transmit power for the downlink based on the transmit power used for the uplink transmission. The node transmitting the uplink will select the uplink transmit power based on how well it is able to receive beacon signals from the node to which it will transmit as detailed later. The selected uplink power is most preferably reported in the field header included in the uplink transmission. The node that receives the uplink transmission reads the header field value and selects the appropriate downlink transmit power. The time proximity of the uplink-downlink transmission pair ensures that the signal propagation conditions most likely remain the same.

[0043] Furthermore, the timing of the uplink and downlink windows - and the fact that the suggested window length is greater than the actual time it takes to transmit a frame - triggers a possible high-speed transmission scheme where a negative acknowledgment (or no acknowledgment even if there was when requested), in the event of an uplink transmission, initiate an instantaneous replay of the uplink transmission in the remainder of the time originally assigned for the downlink transmission. If the transmit power can be increased, it is usually advantageous to use more transmit power for immediate repetition than for the original uplink transmission.

[0044] The use of downlink windows is not always needed; at least if the main function of the network is to transfer data in a substantially one direction, from subnodes towards mouths. Retaining the routing information, however, necessitates in most cases that downlink transmission is at least possible, although the downlink capability need not be symmetrically equal to the uplink capability. In addition, acknowledgments are required after the data frames to guarantee reliable communication. In such cases, it is preferable to combine the acknowledgment and any downlink data into a common frame. Hence, the downlink portion does not significantly restrict channel utilization and energy efficiency. Creating downlink data links upon request would most likely reduce energy efficiency significantly. Furthermore, efficient channel utilization is not a major problem in WSN as the data transfer rates are very low.

[0045] A node that should receive data in a bookable window or that needs to check the ALOHA window to find out if it contains a transmission does not necessarily need to keep its receiver on for the entire duration of the window. If transmission is present, the receiver can turn off as soon as the frame has been fully received (assuming the window length is greater than the frame length). Since each window is most likely arranged to transmit at the very beginning of the window given that it is empty, it suffices for the receiving node to listen for a specific time window before concluding that no transmission will be made. The length of such a time window can be expressed as t + p, where p is the time required to transmit the frame and t is the additional marginal time to compensate for timing and swing crystal tolerance errors. For example, values ​​for t = 100 microseconds and p = 256 microseconds can be used. Calculations of cost and energy exchange can be made: by using a more accurate (and more expensive) crystal oscillator, a lower t-value can be selected, which reduces energy consumption.

[0046] In addition to transmitting the cluster beacons and broadcasting the downlink in specific windows of the super frame 302, the root node transmits network beacons over a network channel. In order for the head node to have only a single radio transmitter, it is preferable to schedule the transmission

- 12 network beacons during the idle period 303. In the embodiment shown in Fig. 3, the root node transmits a so-called active network navigation signal 331 once during each access cycle 301. In this case, the transmission of the active network beacon 331 is scheduled to take place at the very end of each access cycle such that the active network beacon 331 is immediately replaced by the transmission of the cluster beacon at the beginning of the super frame. In addition, the root node transmits several so-called idle network beacons 332 during the remainder of the idle period 303. In this case, the number of idle network beacons per cycle is one, but may be zero or more than one. The beacon 333 is the time from the start of the network beacon to the start of the next network beacon. If the root node transmits network beacons at constant intervals, navigation period 333 is a well-defined constant value, and the reciprocal thereof may be called navigation rate. If the network navigation signals are transmitted at varying intervals, the average navigation period and the corresponding average navigation speed may be calculated.

[0047] The target of so-called idle beacons (shown above due to idle network beacons) can be analyzed more closely. The general assumption about wireless networks is that keeping the number of transmissions per cycle of time as small as possible leads to the greatest energy savings. However, this is not entirely true. The realization of a certain number of "unnecessary navigation transmissions allows to keep the time necessary for network scanning, which may translate into much greater energy savings necessary to carry out these transmissions. Contrary to popular belief, a radio transceiver consumes much more energy in receiving than transmitting. An exemplary node device according to an embodiment of the invention consumes 20.07 mW of power when transmitting at -20 dBm, 30.68 mW when transmitting at 0 dBm, and 44.98 mW when receiving when the microcontroller of the node device is in an active state. Additionally, it should be remembered that the beacons are short and follow a well-defined schedule, while the network scan is generally continuous and continues indefinitely until either it produces a result or time lapse, and that potentially there are many more. subnodes that need to scan the network, rather than the head nodes performing the beacon.

[0048] Figure 4 shows a schematic breakdown of the energy consumed by a wireless communication node into initial energy 401, data exchange energy 402, and network maintenance energy 403. Of these, initial energy 401 comprises scanning the network and receiving beacon signals to detect nodes in range and associate the network, which is exchanging transmissions necessary to connect the node to the network. The time interval 411 in which the initial energy 401 is consumed may be designated as the node start-up period. The data exchange energy 402 consists of the transmissions and receptions that are necessary to transmit the high-layer data payload and acknowledgments. Network maintenance energy 403 is consumed during the transmission and reception of beacon signals and during periodic network scans to update information on the network topology. The period 412 that a node consumes data exchange energy 402 and network maintenance energy 403 is the node life during which the node is available to actively participate in the operation of the network.

[0049] The vertical dimensions of the segments 401, 402 and 403 are given by way of example and do not relate to any quantities or actual energy or power information. The overall energy consumption is highest in the initial 411 node because when the node! will get sufficient knowledge about

The communication schedule and the required power levels can effectively apply various types of energy limitation strategies during the lifetime 412 of the node.

[0050] The relative importance of the start-up energy 401, the data exchange energy 402, and the maintenance energy 403 of the network to the energy budget of the entire network depends on the nature of the network. In a high dynamic network with a very short node life, it may happen that the starting energy 401 will dominate. In typical cases, however, the life of a node may be months or even years, in which case the relative importance of the starting energy 401 remains negligible.

[00511 In order to mathematically model the energy consumption of a grid, certain definitions are needed. Energy E._tx used to send the data frame is (L Ί ~ L_fHey + T_{£ i} + - P * (1) where

Lf = frame length in bits

Ei = energy required to exchange one data bit between the node controller and the transmitter and the accusative

T._st = length of the transient start time of the transceiver

R = radio data rate in bits / second and

Ptx = power consumption in transmit mode when the node controller is also in active mode.

[0052] We assume that the average power consumption during the transient start-up period of the receiver transmitter is the same as the Ptx. If there are several possible values ​​for the transmission level, they directly affect the Ptx which in turn is viewed as multiple Etx values. For example, we assume a high level of transmit power and a low level of transmit power, leading to different values ​​of Et ^ high) and Etx (iow). Frame receiving energy Er_x this

<img file="PL1829291T3_D0001.tif" />

(2) where

Ti = length of the listening period, a

P._ra = power consumption in receive mode when the node controller is also in active mode.

[0053] If the network scan always results in successful reception of the next available beacon, the longest network scan time is substantially equal to the reciprocal of the beacon rate fb (it can also be assumed that the root node will always perform a network scan for the entire beacon period while a subnode may terminate the scan as soon as it finds a suitable cluster, where "suitable" means, that the cluster root node is close enough and the cluster does not appear to be overloaded.) Thus, the scanning energy of the network Ens can be expressed as

<img file="PL1829291T3_D0002.tif" />

[0054] In general, the above expression is an approximation as it does not take into account the energy used to transmit the data received from the radio transceiver to the node processing unit.

-14 The difference is negligible as the transceiver must be in standby mode (not receive mode) when data is sent to the processing unit. The time required to load one frame of received data is approximately 1 ms.

[0055] For the purposes of further calculations, we assume that data exchange operations may have a higher priority than beacon transmissions, so that the beacon rate does not affect the data exchange energy. The node start-up sequence includes a network scan (which may translate to receiving several network beacons until a suitable cluster is found, but assumed to be limited to one beacon period), receiving a network beacon, receiving a cluster beacon, transmitting a frame related and receipt of confirmation. We assume that the related frame is transmitted at a low transmit power level as above. The expression for the initial energy Es of the node is - E_ns + + E_{fx {low}} . (4)

[0056] This shows that the initial node energy is inversely proportional to the speed of the beacon fb. Taking the exemplary values ​​Lf = 256 bits, Ei = 2.3 nJ / bit, Tst = 250 ps, ​​R = 1 Mbps, Ptx (high) = 30.68 mW, Pugowj = 20.07 mW, Ti = 300 ps and P = 44.98 mW, we get the initial energy node at the level of 4.6 mJ for a beacon speed of 10 Hz and 45 mJ for a speed of 1 Hz.

[0057] The maintenance of the network is carried out continuously throughout the life of the network. Thus, the most convenient issue is to assume the energy consumption in periods of 1 s, which corresponds to the average power consumption. The average holding power Pm of the network is defined as the sum of the scanning power Pns of the network and the exchange power Pb of the beacon. In addition to the energy of network scanning, P_ns depends on the speed of the network scan operation required to enable proper data routing in the network. It depends on the dynamics of the network and is defined as the network scan time T. Since the head nodes use longer communication distances and more complex data routing, it is assumed that they scan the network more frequently than the subnodes. Thus, the network scan time is defined separately for head nodes (Ts &lt;h)) and sub nodes (Tstsj). Pns is the average value per node that takes into account the number n_s subnodes on each head node. Thus, the power of network scanning is obtained thanks to

F and f) ⁽⁵⁾

I + Π_Β 's (s)

[0058] To obtain tangible results, T can be established_S.(h&gt;, Ts (s) and n_s at an exemplary value level assumed, for example 100 s and 500 s for network scan times and n, respectively_s = four subnodes for each head node. Additionally, an access cycle time T of 4 s can be assumed. In Fig. 5, curve 501 shows the network scan power Pns which drops sharply as the beacon rate fb increases. The decrease from 160 pW to 16 pW is obtained with an increase in the speed of the navigation signal from 1 Hz to 10 Hz.

[0059] The beacon exchange power is consumed by transmitting and receiving the beacon signal. In addition to the beacons transmitted at rate fb, the cluster beacons are transmitted once per access cycle the length of which is Tac. For simplicity, it is assumed that the cluster beacons are only received from one root node. As discussed above, for power control reasons, it is recommended that beacon frames always transmit two races, the first time using

- 15 for the high power level and the second for the low power level. Thus, the power Pb consumed during the beacon exchange can be modeled as follows

Ε + E

P 'b "4,"

14- n_s '1 1 ζ + - 4

2E rx (6) ac / 'ac

[0060] The variations in Pb and fb are plotted in Figure 5 as a curve 502. Contrary to the network scanning power, the beacon exchange power Pb can be minimized by minimizing fb. The beacon exchange power Pb drops from 71 pW to 23 pW as fb drops from 10 Hz to 1 Hz. As can be seen from the figure, the holding power Pm of the network, plotted as curve 503, has a minimum value (77 pW) at a beacon speed of 5.6 Hz. For low speeds below 1 Hz, P_m it typically doubles when the transfer speed is halved. This effect is reversed at high beacon rates above 10 Hz.

[0061] Next, the effect of the network scan time and the number of subnodes for each head node on the power consumption for maintenance may be considered. Obviously, increasing the network scan time limits the network scanning power and shifts the minimum P_m to slower speeds of beacons. As an example, when fb is 1 Hz, P._m increases from 55 pW to 833 pW, as Ts (n) drops from 500s to 20s. On the other hand, an increase in n_s reduces the average power per node consumed for beacon transmission. As subnodes typically have longer network scan times than master nodes, power consumption for network scanning is also reduced. When fb is 1 Hz, the increase of n_s from 0 to 8 limits the Pm from 500 pW to 150 pW.

[0062] The average power consumption of a node in a wireless sensor network without scanning the network may be, for example, 100 pW. Thus, the energy consumption for network maintenance has a very significant impact on the entire service life of the network. From the description below, it can be concluded that the node has a long service life (i.e. the starting energy remains negligible in the overall energy consumption of the node). The optimal beacon rate fb * is determined by the constraint of network maintenance power relative to the beacon speed. The optimization function can be written as rx

P._m + n_s n_s

--- + —a

TT 's (s + n_s

2E 4 — a

T • ac (7) where Eb = Etx (hi₈h) + Etx (iow). It can be shown that there is a unique minimum value at fb * which is obtained by setting dPm / dfb = 0 in (7). This is ^ bn

__ | --S (8)

[0063] Thus, the optimal beacon rate is determined based on the network parameters Ts (h), T_S.(s) in_s and radio parameters Eb and Prx.

[0064] Fig. 6 shows the change of fb * in T_S.(h), with a value of n_s equal to 0 (601), 1 (602), 4 (603) and 8 (604). The relationship of Ts (s) / T_S.(h} is set to level 5. For the given parameters, the optimal beacon speed V is in the range 2-21 Hz Increasing the number of subnodes per head node increases the optimal beacon speed, but it also increases the power consumption of the head node, overall network power consumption drops due to shorter network scan time.

[0065] Fig. 7 shows fb * as a function of T_S.(h) when the ratio Ts (s&gt; / Ts ^} equals 2 (curve 703), 5 (curve 702) and 10 (curve 703). As can be seen, the limitation of the ratio T_s&lt;s) / T₅(h) increases the optimal rate of the beacons as network scanning operations are performed more frequently. When T_s&lt;h) is 100 s and T._S.(_S.) / Ts (h) drops from 10 to 2, the optimal speed of the beacons increases from 5.5 Hz to 9.2 Hz.

[0066] As a summary of this section, it can be seen that the beacon speed of the wireless sensor network has a very significant impact on the maintenance power consumption, which is defined as the sum of the beacon exchange power and the network scanning power. The speed of the beacons is optimized in relation to the required network scan time, determined on the basis of the network dynamics and the data routing algorithm. Estimates show that the network maintenance energy ranges from less than 100 pW to even a few milliwatts, which is mainly influenced by the speed of beacons. Referring to the example discussed above, the optimal beacon speed for a typical application and a stable network is about 5 Hz. In the case of a dynamic network, which requires frequent network scans, the optimal speed of beacons is 20 Hz. The optimization function used does not take into account collisions in the network channel that become noticeable at higher beacon speeds and dense networks. Thus, it is preferable that the beacon speeds are below 10 Hz. Most importantly, the information on optimizing the speed of beacons is generic and can be applied to previous wireless sensor network solutions, such as LP-WPAN, S-MAC and T-MAC. As a general rule, in solutions such as LR-WPAN, additional beacon signals sent during the coordinator's inactivity can effectively reduce the network scan time, if the time of beacons would be otherwise long, more than 1s. In dense and large networks, the optimal beacon speed will be slightly lower than in the above description due to the higher collision probability.

TIME SYNCHRONIZATION

[0067] Although the clusters are synchronized with each other, the possibility of extending the reference time of the network among the nodes is an advantageous solution, especially when the measurement of values ​​or other data needs to be timestamped, which helps to avoid confusion when measuring data from other sources or that were sent via different paths are combined and summed up. For example, timestamps can deduce the speed and direction of a moving target's movement, which translates into measurement results on different sensor nodes at different time instances. Additionally, this may require information on the location of at least some nodes; Site awareness is easily obtained, for example, by programming specific fixed nodes to know their coordinates and convey them in data messages to the exit nodes.

[0068] A natural requirement of time synchronization is that it must be achieved with the minimum possible number of transmitted and received control messages, i.e. with the minimum possible energy used for spreading the reference time.

[0069] According to an embodiment of the invention, the network beacon signals and the cluster beacons are for propagating the reference time across the wireless sensor network with a desired accuracy, e.g., one microsecond. We assume that there is one node in the network that has exceptionally good access to an accurate external reference time. Typically, such a node is or may be an exit node

-17 is a node with a dedicated reference time, equipped with an accurate clock, GPS receiver or a similar type of time source.

[0070] The network beacons and the cluster beacons may have different roles during reference time propagation. According to one suggestion, a network beacon frame comprises an absolute total time expressed in, for example, 48 bits, which, given an exemplary resolution of one microsecond, may represent an overall expressible time range of about nine years. The cluster beacon frame may contain a so-called short time that only includes a few (e.g. 16 bits, meaning a time span of 65 milliseconds at microsecond resolution) of the least significant bits of the entire time. The timestamp included from the beacon frame should represent the time during which the beacon frame has been transmitted.

[0071] In addition to the proper time stamp, it is preferable to include in the beacon signal transmission an indicator of how accurate the reference time is assumed to be. A simple indicator is the time hop field, which value is zero when a node is acting as a time reference source and increments by one for each node that is transmitting a time reference. A more accurate indicator may also include an indication of how much time has elapsed since the transmitting node received the reference time from elsewhere closer to the reference time source since the offset occurs in all crystals that act as local clocks in the nodes. If a node is able to select from a plurality of received reference time transmissions, it should always select the one whose index shows the highest possible reference time accuracy.

[0072] At least two basic alternatives are available that can be used as a strategy for introducing a received reference time to use. According to a simple alternative, the node simply receives the beacon transmission, reads the time stamp and - if the accuracy indicator meets some acceptance criteria - simply uses the reference time as an indicator of the exact time the beacon was received. Even if the use of the reference time involved adding a few default delays, this simple alternative will inevitably lead to some time inaccuracies as the time required for the transmission to propagate through the air and by receiving and transmitting equipment, including MAC processing, does not is exactly known. However, the accuracy achieved is in many cases sufficient if the network does not include any applications that could rely on exact synchronization.

[0073] According to a more elaborate and more precise alternative, the first node transmits a reference time in the cluster beacon at a time t0. The second node receives this cluster beacon and signs the receive time as ti according to its own clock. The second node stores the times to and ti but does not yet update its clock with the reference time. In the case of an uplink transmission window, the second node transmits the data frame of the first node at a time t2 which the second node is rewriting. The first node marks reception of the data frame as ta and transmits the few bits of the least valid time stamp ta to the second node in an acknowledgment frame. At this point, the second node knows all times are, tt, ta and ta and can calculate the propagation time tpropagation as follows f - ~ ^ 0)⁺ ^ 3 ~~ tpropagation £ '

[0074] At this point, the second node may update its own time t as follows:

-18t = t⁺ Him " ^f1)⁺ tpropagation '(¹ θ)

[0075] It can be seen that the transit delays are typically less than 1 millisecond. Thus, for the transmission of time ts in the acknowledgment frame, it is sufficient to reserve approximately 12 to 16 bits; 12 bits mean a possible time range of +/- 2 ms, while 16 bits mean +/- 32 ms. The maximum 16 bits reserved for this purpose does not constitute frame overbooking. However, if such a reference time distribution rule is chosen, the reservation should be made in downlink frames. To avoid complications, it is preferable that all nodes of the wireless sensor network always follow the same selected reference time distribution principle.

[0076] If the original time reference time source was an exit node, the reference time typically propagates down the network hierarchy in the form of a tree in the opposite direction to data propagation up through the exit node. If the optimal path (due to factors such as overall energy level, head node loading, link quality) between node X and the exit node is long and node X requires a more accurate reference time, it can be obtained from a nearby GPS-equipped node (or other suitable equipment), listening at regular (network or cluster) intervals for transmission of the beacons of such a node. Node X must therefore remain aware of the time and frequency of both the GPS-equipped node and the adjacent root node through which it transmits data towards the exit node. To keep the overall power consumption to a minimum, it is not recommended to increase the network scan.

FRAME STRUCTURES

[0077] Fig. 8 illustrates a preferred frame structure 800 that may be used during transmission over a wireless sensor network, according to an embodiment of the present invention. The use of a fixed frame length slightly reduces the energy efficiency of the transceiver, compared to the alternative that allows the use of variable length frames depending on the number of actual bits that are transmitted. However, fixed frame lengths simplify memory management and header decoding and improve the robustness of the TDMA scheme.

[0078] The exemplary frame type selection defines 13 frame types which are mapped to three MAC frame formats. The network beacon frame type uses the format of the network beacon frame. Accordingly, the cluster beacon frame uses the format of the cluster beacon frame. All other types of frames (i.e., data, binding, box reservation, acknowledgment, connection box binding and reservation, combined box acknowledgment and reservation, combined data and binding, combined data and box reservation, combined data and acknowledgment, combined data and binding and reservation. the windows and the combined data and the acknowledgment and window reservation) use the data frame format. When summing data, association, acknowledgments and reservations to shared frames should always be used whenever possible, if only the transient energy required to start the transceiver is of the same order of magnitude as the energy required to ensure correct transmission. Summing also helps to minimize latency due to network topology changes, as data can be transferred immediately with an appropriate control message (such as a binding or window reservation message). Summation is also most advantageously used by head nodes which carry simple data messages (e.g. measurement results) from several subnodes towards the exit node such that the payload of data and the corresponding node identification data are combined into a common frame if possible.

[0079] All frames are 32 bytes long and consist of physical header 801, MAC header 802, header type 803, payload 804, padding 805 (if needed), and padding 806. Exemplary field lengths are 40 bits for physical header 081, 64 bits for MAC header 802, 136 bits for header type 803, payload 804 and padding 805, and 16 bits for padding 806.

The physical header 801 consists of an initial sequence 811 (e.g. 8 bits) for transceiver bit synchronization and a network address 812 (e.g. 32 bits) to distinguish real frames from background noise and to identify a single wireless network sensor. The MAC header 802 begins with a code type 821 (e.g., 4 bits) of the frame. TX power field 822 (e.g., 4 bits) specifies the transmit power used for the current frame. The address field includes the destination cluster address 823 (e.g., 16 bits), destination node address 824 (e.g., 8 bits), source cluster address 825 (e.g., 16 bits), and source node address 826 (e.g., 8 bits). A payload length field 827 (e.g., 8 bits) terminates the MAC header 802.

[0081] The formats of the type 803 header and payload 804 depend on the frame type. The unused charge is filled with padding 805 to ensure a constant frame length. Padding length 805 can be derived from MAC header payload length field 827. At the end of each frame, padding 806 is provided for error detection purposes.

[0082] The data frame type header begins with a priority field 831 (e.g., 1 bit) that describes the priority level of the payload. A node field type 832 (e.g., 1 bit) defines whether the transmitting node is a subnode or a head node. An acknowledgment for a given data frame is requested by request field 833 (e.g., 1 bit) of an acknowledgment. The status field 834 (e.g. 1 bit) of the acknowledgment serves to acknowledge the previous data frame. If the longer acknowledgment status field or acknowledgment status bit in field 834 is used, the receiving device tells the receiving device to look for an acknowledgment message in the payload portion of the frame, or the acknowledgment is performed by transmitting specific acknowledgment frames, the acknowledgment may also include a sequential number or other type of data frame identifier. which is confirmed. The payload frame includes user data 841 communicated between the upper protocol layers.

The type header for cluster network beacons begins with a frequency field 835 (e.g., 8 bits) that defines the cluster RF channel. An energy field 836 (e.g., 8 bits) defines the energy resource of the root node. The cluster load, resulting from the number of associated subnodes, data routing activity and data summation, is determined by the load field 837 (e.g. 8 bits). Based on the energy and load information, the new subnodes can select the most appropriate head node. Payload in beacon frames depends on type. The payload of the network beacon includes a field 824 (e.g. 32 bits) giving the time to the next cluster beacon, e.g. a resolution of 100 [micro] s. Thus, the other nodes may go back to sleep until the start of the next super frame. Additionally, the payload of the network beacon includes a field 843 (e.g., 48 bits) for full reference time distribution and a field 844 (e.g., 8 bits) for transmitting the time accuracy indicator. Cluster beacon payload defines length 845 (e.g. 32 bits) of access cycle with 100 ps resolution to ensure cluster synchronization, possible sleep cycles 846 (e.g. 8 bits), for temporary suspension of communication in the cluster, short reference time 847 (e.g. 16 bits), pointer 848 (e.g. 8 bits) time accuracy and a bookable 849 (e.g., 64 bits) time slot allocation table that identifies current window reservations.

MANAGEMENT PROCEDURES

[0084] Wireless sensor network management routines define mechanisms to support automatic network configuration and energy conservation. The mechanism includes cluster scanning, distance assessment, networking, cluster association, window reservation and sleep cycles. The first four are schematically shown as parts of Figures 9a and 9b.

[0085] The cluster scanning procedure is used when a node needs to discover neighboring clusters to bind or transmit data. In Fig. 9, the cluster scanning procedure is illustrated by steps 901 to 906. After scanning has started in step 901, the node listens to the network channel in a loop spanning steps 902 and 903, trying to receive network beacons from neighboring clusters. The power consumption for scanning is significant because the transceiver has been in idle mode and in RX mode for a long time. Depending on the configuration, the scan may complete after one network beacon period ("time mode) during which network beacons are received with high probability from all neighboring clusters. Thereby, the node has completed steps 902, 905, and 903 at least one time before terminating in step 904. Another alternative is to stop scanning immediately after the corresponding cluster has been found (i.e. an idle or active network beacon has been received) which in Fig. 9a corresponds to the transition from step 905 to step 906.

[0086] Fig. 9b illustrates a process for an association with a head node. In terms of strength, the sub-knot should tie into the closest main knot. After starting in step 911, the node that wishes to associate is listening to the frequency of the cluster from which it has read a previously received network beacon. In addition, a time to start the cluster association process has been determined based on reading the corresponding value from the previously received network beacon. Only if reception of the cluster beacon has failed, proceeds through steps 912 and 913 to a new cluster scan in step 914. Otherwise, the node stops receiving at step 915 immediately after successfully receiving the cluster beacon.

[0087] The RSSI measurement is typically not included by default in low power transceivers; it is also not recommended, because it would complicate the receiver and consume power. Accordingly, according to an embodiment of the invention, all the beacons in the suggested protocol are transmitted twice using different transmit power levels (e.g., 0 dBm and -20 dBm), obtaining a different transmission range. If both beacons are successfully received in step 916, the node estimates it is close to the root node and selects low transmit power for its own transmission according to step 917. Otherwise, the node will be away from the root node and communication will proceed. will be at the higher transmit power level, according to step 918. In the latter case, the subnode also starts recognizing as to whether the nearest root node is available according to step 919. If another root node is found, conceptually a hop back to step 916. However, it is highly likely that only "distant" nodes will be available. major, so the node will have to be content with connecting to one of them at a higher transmit power level. In such cases, it is recommended that a network scan be performed from time to time, at a rate which may be higher than when head nodes are available nearby, to find out if there is a head node to which the node could connect on the lower transmission power level.

It is possible to use more than two transmit power levels, or to transmit more than two beacon frames - each at a different power level - in each signal transmission.

Or using cyclic selection of beacon transmission levels so that the first beacon transmission includes two frames at power levels 1 and 2, respectively, the subsequent beacon transmission includes two beacon frames at power levels 3 and 4 (or 1 and 3) respectively followed by power levels 5 and 6 (or 1 and 4) respectively, and so on until all power levels have been used. The receiving nodes will estimate the distance and the required transmission power level according to the beacon transmission at the lowest level that can be successfully received.

[0089] When a node is activated for the first time it carries out the cluster scanning procedure. The FFD can establish its own cluster and become the head node or bind to an existing cluster and become a subnode. The decision depends on the load and energy state of the clusters found and the estimated distance to them. The decision is schematically shown in steps 902, 921 and 922 in Fig. 9b. The RFD can of course only bind to an existing cluster as it is not able to act as a master node. The implementing the cluster in step 921 includes selecting the cluster address and the cluster RF channel according to information obtained from the neighboring clusters. Once a cluster is completed, other nodes and existing clusters can bind to it and create global network connectivity.

[0090] A cluster association is required for data exchange with the cluster head node. A subnode can only bind to one cluster, while the master node needs to bind to several adjacent clusters to establish efficient multi-hop connectivity. The node connects after receiving the cluster beacons by sending a cluster association request in the ALOHA pane in step 923. If the head node receives and accepts the request, it will send an acknowledgment and the node address to the downlink part of the window. Receiving an acknowledgment on a bound node is shown in step 924 in Fig. 9b. As the network is prepared for high dynamics, separate network disconnection requests are not applied. A node is automatically disconnected from the cluster if the master node does not receive data from the node within several access cycles. This amount is a parameter set by the network designer and may be, for example, ten.

[0091] A Node may reserve a bookable time slot by transmitting a window reservation request in the ALOHA time window. If the root node receives the request, it typically sends an acknowledgment in the downlink part of the window (although the requesting node may accurately ask for no acknowledgment, as in any other case where the node wishes to save power as much as possible, without even turning on its receiver, to receive it). notifications). The correct allocation of a window, if allocated, will be in the next cluster beacon from that root node. Bookable windows are allocated according to the desired priority levels, preferring the root nodes. The allocation is specified in the cluster beacons and is updated every access cycle. If the root node is unable to receive data in the corresponding bookable window within four access cycles, or if the window is assigned to a node with a higher priority, the previous window reservation is automatically deleted. The master node is able to set the entire cluster to idle for one or more access cycles using the sleep cycle field in the cluster beacon. During sleep cycles, all communication in the cluster is temporarily suspended. Sleep cycles allow additional energy savings at the expense of increased transfer latency.

[0092] The network channel is typically predetermined. In conditions involving severe interference, it is possible to predetermine the number of network channels, one of which will be selected

-22i used if possible. The responsibility for deciding to select from among a plurality of predetermined network channels, and for determining when a network channel changeover time occurs, typically rests with a single valid node, which is also typically the escape node. When a new node joins the network, a network channel, which may be provided in any number of channels, scans the possible network channels until it finds one through which it can receive network beacon transmissions and continues to use (and scan) such a network channel until until it stops working.

[0093] The principle chosen for network scanning is very important from the energy consumption point of view. The wireless sensor network according to the invention uses short periods of activity and long sleep times. The clusters are synchronized with each other and operate on different frequencies, which forces the knowledge of transmission schedules and frequencies of neighboring nodes. Listening to their beacons on a network channel, as previously described, is much more energy efficient compared to prior art solutions in which a node must listen on all possible frequencies long enough to hear traffic or beacons.

[0094] In principle, it is possible to replace the frequency split between the clusters with a code split where all the clusters would share a common transmission frequency but with different cluster specific (and / or node specific) spreading codes to avoid intra-cluster collisions. However, since ODMA requires extensive synchronization and a more complex transceiver architecture, it is anticipated that it will be a feasible alternative only for specific applications.

DATA TRANSFER

[0095] Wireless sensor network communication protocols involving clusters must define user data exchange mechanisms during intra-cluster communication (i.e. between nodes belonging to the same cluster) and between clusters (between nodes belonging to different clusters). A subnode or source root node is associated with the sink or destination node, which is most often the root node. The data exchange is carried out during the target cluster superframe and the source cluster idle time using ALOHA windows and windows that can be reserved. In order to be able to share the network resources between nodes in a more sustainable manner, the subnodes may advantageously only use one time slot per access cycle, thus providing a throughput of 132 bits per access cycle. In order to ensure effective communication between clusters, the use of the master node windows is not limited. Thus, the maximum throughput for the eight bookable data windows is 1056 bits per access cycle.

[0096] Communication between clusters from a subnode via root 1 to root 2, along with an exemplary window reservation procedure, is shown in Fig. 10. The dashed segment shows transmitting and white receiving. The lowest row of segments represents the radio activity of head node 1, the highest row of segments represents the radio activity of head node 2, and the middle row of segments represents the radio activity of the subnode. In the figure, time runs from left to right. Master node 1 and master node 2 operate on cluster channels 9 and 5, respectively. Network signaling is performed on channel 1. Identifiers 1001, 1002, and 1003 represent access cycle, superframe period, and cluster idle periods of root node 1. Similarly, identifiers 1011, 1012 and 1013 represent the access cycle, super frame period, and cluster idle periods of root node 2. Both nodes

Principals transmit idle network beacons twice during the access cycle. The relative time lengths in Figure 10 are not to scale.

[0097] Subnode 1, previously associated with head node 1, receives transmissions 1021 of cluster beacons of root node 1 and transmits the data via ALOHA window 1022 and a clean type data frame. Head node 1 responds by sending an acknowledgment in the downlink portion 1023 of the same ALOHA window. Assuming Master Node 1 does not yet know the timing of the corresponding next hop or cluster node, communication between clusters begins by scanning the cluster 1031. Scanning completes when idle network beacons 1032 are received from root node 2, which is the selected destination (next hop). Then, after reading the appropriate time from the received idle network beacons 1032, root node 1 receives the cluster beacons 1041 on channel 5 and transmits a bonded data type frame and window association and reservation using ALOHA box 1041. Head node 2 responds with an acknowledgment frame 1043 which includes the node address it assigns to head node 1. Immediate data transmission in said linked frame improves overall energy efficiency and minimizes latency compared to the case where the binding node should perform first. binding, and then start uploading.

[0098] In the next superframe of the master node 1 cluster, the data frame 1051 is received from the subnode 1 and acknowledged at 1052. We assume that the subnode has not requested a bookable window, so the data frame 1051 is again provided in the ALOHA window. The new root node 1 is already an associated cluster node of root node 1, so it is only necessary to receive the cluster beacons from root node 2 at 1061 and transmit a data frame in the bookable time window 1062. Head node 2 responds with an acknowledgment 1063.

[0099] As shown in the figure, communication between clusters does not significantly increase the target master node's work cycle, which simplifies power management. In general, the active transceiver time at the master node is several percent of the access cycle length, while for the subnode it is several per mille. The source master node is loaded by cluster scanning. However, energy consumption drops due to idle network beacons. After cluster affiliation, communication between clusters is energy efficient. Communication inside the cluster is not disturbed because communication between clusters is carried out during the idle time of the cluster. In addition, the subnode's duty cycle is low, requiring only three receptions and one transmission per access cycle.

ROUTING

[0100] The selected routing protocol depends very little on the initial information available regarding the expected topological nature of the wireless sensor network. The following example is based on the assumption that the network has only a relatively few outlets compared to the total number of nodes, and messages from outlets to other nodes are very rarely specific to a node and typically affect the entire network or some parts of it. The reactive routing principle is preferred, whereby the routing protocol only handles the storage of routing information for specific destinations that are known to be necessary. Each node only needs to maintain a next-hop routing table, instead of remembering entire paths to destinations. In the case of a subnode, the next-hop routing table is particularly straightforward because the next next hop to any destination is a connection to the cluster root node. Beneficial

There are sequential frame numbers and multipath routing support such that it is possible to improve the routing reliability of important data by passing it along alternate paths.

[0101] According to a simple exemplary alternative, the exit node interested in temperature readings from an area within the coverage area of ​​the network sends a temperature request in a data frame, which will also include the identifier of the sending exit node. The request is sent via the network according to the flooding principle. To avoid loops and duplication, the frame contains a sequential number. Each head node forwards the request to all subsequent nodes it hears except the one from which it received the information. The hop count is updated in the frame that contains the request: when the head node starts forwarding the request it has received, it increments the hop count by one.

[0102] Each node that receives the request records the value of the hop count at the time of receiving the identifier of the originating node (sink) and the node from which the request came last. Thus, the routing table register includes a destination identifier, a next-hop identifier, and a hop count. If the head node later receives a copy of the same request with a higher hop count, it deletes the later request without forwarding it. Correspondingly, if the root node receives a copy of the same request with a lower hop count value, it deletes the previous record from the routing table and replaces it with the one containing the hop count and forwards the node identifier received from the next reception. When designing the routing protocol, it is also possible to define (at least whether the hop counts of the two copies received are closer together than the set limit) that the receiving head node creates a copy record based on both received copies of the request, which records can be used for multi-path routing .

[0103] Request network flooding will expire when all head nodes have received and forwarded them. If a node that knows that it has a temperature sensor receives the request, obtains a temperature reading and sends according to the information in the routing table - a frame containing its source identifier, temperature reading, possibly location (with reference to other nodes) and possibly a time stamp towards the exit node who generated the request. By observing the contents of the reverse direction transmission, the root nodes that are on the routing path towards the appropriate sink node can regenerate the records for the routing tables, these records will include introductions about the subsequent hops for routing to the node from which the temperature reading is taken. From time to time, the escape node may make a new request to update all routing information.

PROTOTYPE NODE DEVICE

[0104] Fig. 11 shows the architecture of an exemplary node device 101, according to an embodiment of the invention. Computing subsystem 1104 is adapted to implement the MAC protocol, upper protocols and application algorithms, depending on the available capabilities and memory. An exemplary practical implementation of the computing subsystem includes a Xemics XE88LC02 (MCU) 1141 microcontroller unit, which includes a CoolRisc 816 processor core, 1132K internal 16-bit ADC program memory, and a 1K data memory. External memory 1142, such as 8KB EEPROM (electrically erasable programmable read-only memory), provides non-volatile memory. The maximum execution speed of the example MCU instruction is 2 MHz.

[0105] The communications subsystem 1105 includes an RF transceiver 1151, an antenna 1152, and an MCU portion 1141 that implements the communications protocols. A practical, example implementation of the communication subsystem 1105 uses a 2.4 GHz NordicVLSI nRF2401 transceiver. An example of the transmission speed is 1 Mbps. The transceiver 1151 has an integrated pattern recognition capability

- 25 and 16 bit CRC and data buffer functions for transmission and reception. Pattern recognition is preferably used to detect the network address. The radio transceiver CRC significantly alleviates the MCU error control workload as only the acknowledgment and retransmission mechanisms need to be implemented in the program.

[0106] The detection subsystem 1103 preferably uses an internal ADC 1132 and an MCU 1141. The ADC typically has preamplification and error compensation steps that increase sampling accuracy. The typical maximum sampling rate using 16-bit resolution is approximately 2 kHz. MCU 1141 can be used to implement ADC driver and sample transmission tasks for application layer. Virtually any type of sensor can be used as the sensor 1131, depending on the application.

[0107] The power subsystem 1102 may be designed in various ways. The exemplary design in Fig. 11 shows an energy source 1121, which may be, for example, a piezoelectric array or photovoltaic array. The regulator 1122 is used to regulate the feed supplied to the rest of the node device. Although the switch mode controller may be more efficient, a linear controller, for example the TPS71525 linear controller, still may be of advantage due to lower quiescent current, lower noise, less electromagnetic interference, and a more compact structure. A battery or a supercapacitor can be used as a temporary energy storage and peak demand storage. In the prototype device, a 0.22 F capacitor was used.

[0108] The dimensions of the prototype were 31mm x 23mm x 5mm. The top side of the prototype printed circuit board contains the transceiver, antenna, EEPROM, sensor and connectors. MCU and controller are mounted on the bottom. The prototype's energy consumption measurements are shown in the table below.

<td>MCU mode</td><td>ADC mode</td><td>Sensor mode</td><td>Radio mode</td><td>Power (mW)</td>

<td rowspan="7">Active</td><td rowspan="6">Active</td><td rowspan="5">Active</td><td>Receiving data</td><td> 46,06</td>

<td>Data transmission (power: 0 dBm)</td><td> 31,76</td>

<td>Data transmission (power: 20 dBm)</td><td> 21,15</td>

<td>Loading data</td><td> 3,74</td>

<td rowspan="4">Sleep</td><td> 2,43</td>

<td rowspan="3">Off</td><td> 2,41</td>

<td rowspan="2">Off</td><td> 1,35</td>

<td>Sleep</td><td> 0,019</td>

[0109] The maximum power consumption is 19 pW when the system is in sleep mode and 46.06 pW when all components are active and the transceiver is in RX mode. The power consumption of the MCU is 1.33mW at a clock speed of 1.8MHz. The radio transceiver uses 33 times more energy

-26than MCU. Receiving data requires twice as much energy as transmitting. In a wireless sensor network, power consumption in sleep and RX mode is the most important.

[0110] Fig. 12 illustrates an exemplary software architecture for a node as shown above in Fig. 11. Basic services of the operating system 1201 include, but are not limited to, a general node state control machine 1202 whose operation is regulated by clock 1203. Control machine 1202 the node carries out the MAC function calls, indicated by the arrows. Frame assembly 1204, queue 1205, and radio controller 1206 (including TX frame 1207 and RX frame functions 1208) are associated with data flow and control. The management functions on the right side of Fig. 12 are performed on demand and include estimate 1209, cluster scan 1210, cluster binding and resolution 1211, window assignment 1212, and node power control 1213. The ADC controller 1214 controls the sensor sampling and is a specific part of the sensor application and routing 1215 which belong to the application layer in the protocol stack. Software can be developed using available tools such as RIDE (Raisonance Integrated Deve! Opment Environment). Sample software developed with this tool and the c816-gcc A compiler (ver. 2.8.0) with optimization level 1 led to the MAC master node implementation size of 14K, requiring 137B data memory. The software subnode version requires 13K program memory and 109B data memory. Data storage requirements do not include frame queues.

ENERGY CONSUMPTION AND PRODUCTIVITY

[0111] During the research work leading up to the invention, the power consumption was estimated in a five-node network with two subnodes (subnode 1, subnode 2) associated with head node 1. Head node 1 receives data frames from both subnodes. The received data frames are concentrated on one data frame, which is redirected to the head node 3 via the head node 2. The data frames are transmitted in bookable time windows. The estimation includes subnode 1, head node 1, and head node 2. Cluster association and window reservation procedures are disabled.

[0112] Since communication between clusters occurs during the cluster idle time of less than 1 s, access cycle lengths are not convenient. Maximum Access Cycle Time is set to 10 s. Network beacon period is set to 250 ms. During the access cycle, subnode 1 receives two cluster beacons and transmits a data frame and optionally receives an acknowledgment from root node 1. Head nodes transmit network and cluster beacons, receive ALOHA windows and allocated bookable time windows, and transmit data to the neighboring root node. Acknowledgments are used for each data frame.

[0113] For the estimated power consumption, subnode 1 has reached the lowest power consumption. Using the longest access cycles of 10 s, subnode 1 consumes 31 pW with acknowledgments and 27 pW without acknowledgments , while the power consumption of node 1 and node 2 was 171 pW and 166 pW, respectively. Using the shortest access cycle of 1 s, the power consumption of subnode 1 increases to 140 pW with acknowledgments and 103 pW without acknowledgments. At the same time, the power consumption of head node 1 and head node 2 is 574 pW and 526 pW, respectively.

[0114] In comparison, the actual power consumption measured on the subnode 1 prototype ranged from 36 pW to 134 pW with acknowledgments and from 22 pW to 97 pW without acknowledgments. The power consumption of Head Node 1 and Head Node 2 ranges from 312 pW to 850 pW and 283 pW to 855 pW respectively. The power consumption of the main nodes is higher than assumed. This is due to the greater activity of the MCU as well

-27 receiver than modeled due to rather complicated time management and inaccuracies in timing. It cannot be accurately modeled during estimates. However, the estimate of the energy consumption of subnode 1 is accurate. As the length of the access cycle increases, subnode 1 is rapidly approaching the power consumption in sleep mode.

[0115] The throughput between the two nodes was analyzed as a function of the access cycle length, since the number of bookable time windows used has increased from 1 to 8. The highest throughput is achieved using the shortest access cycle length of 1 sec. As each data frame carries 132 data bits. user, the maximum throughput when using the 8 bookable time windows is 1056 bps. Due to the downlink windows, the throughput is symmetrical in two directions, reaching a maximum total throughput of 2112 bps. The capacity from the subnode to the master node varies from 13.2 bps to 132 bps as the access cycle length drops from 10 s to 1 s and only one bookable time window is allowed. It should be noted that, contrary to some of the previous assumptions, low speed data transfer is not recommended if the goal is to save energy. Conversely, the transmission - primarily the transmissions of beacon signals to be received by a greater number of receiving nodes - should be performed at relatively high transmission speeds (e.g. Mbps) in such a way that the holding time of the receivers of the receiving nodes can be kept as short as possible.

[0116] The average protocol delay for the transmission of data frames between two nodes is approximately half the access cycle length when offered capacity is below the estimated MAC protocol capacity. In this case, the kolej protocol queues are emptied on every superframe.

[0117] The estimate can be expanded into a larger network with four clusters each with seven associated subnodes. Data is transferred between clusters to the sink using three bookable data windows per cluster. The five subnodes in each cluster are assigned to bookable time windows, while the other two subnodes in the cluster transmit frames in ALOHA time windows. Acknowledgments are used for each data frame. The network beacon period is 250 ms. For simplifications, collisions or framing errors are not included in the estimation. The energy consumption of subnodes and main nodes is estimated as a function of the achieved capacity.

[0118] One-way capacity is performed from the subnodes to the head node and between the head nodes. The maximum throughputs for the subnode and head node are 132 bps and 396 bps, consuming 140 pW and 977 pW respectively.

[0119] Assuming a node is sampling 16 bit at 1 Hz, 16 bps capacity is required between the subnode and head node. Based on the measured results on the prototype, an access cycle of 8.2 sec was obtained with a power consumption of 34 pW per subnode for the required capacity. The power consumption of the master node using the same access cycle length is 230 pW, resulting in a data transfer capacity of 48 bps. Thus, the average power consumption of the node is 59 pW. Due to the large number of subnodes, data summation should be performed on the main nodes. Since a 2s transmission delay during the hop is required, the access cycle length must be 4s. This leads to a maximum throughput of 33 bps and a subnode power consumption of 50 pW. The main node's capacity and power consumption are 99

-28bps and 340 pW. The average power consumption per node increases to 86 pW. This energy can be obtained from a 1 cm 3 sieve battery with a service life of more than 1 year.

[0120] The energy can also be obtained, for example, from a temperature difference of 5.5 ° C, using a thermoelectric generator with an area of ​​1 cm² or vibrations of small machines, e.g. the housing of a microwave oven, using a 1 cm3 piezoelectric generator. In total, the verified subnode power consumption is 27 pW with an access cycle length of 10 s without acknowledgment. This leads to a transfer delay of approximately 5 sec and 13.2 bps of throughput from the subnode to the master node, which meets the requirements of most WSN applications. The maximum total throughput in both directions using the bookable time windows is 264 bps between subnode and head node and 792 bps between head nodes.

CIRCULATION OF HEAD NODE RESPONSIBILITY

[0121] Running in Master Node mode will consume significantly more power than in Sub Node mode. On the one hand, the average number of subnodes per root node influences the optimal execution rate of the beacon transmission: and the higher the average number of subnodes, the more frequently beacon transmissions should be performed (see equation 8). Thus, even if the first time it seems appropriate to have only a minimum number of root nodes in the wireless sensor network according to an embodiment of the invention, increasing the average number of subnodes per root node may lead to a situation where a certain beacon rate is will no longer be optimal. Moreover, should one of the root nodes fail in such a network, the absence of redundant paths for communication between clusters will require a relatively large amount of network scans by multiple nodes before full connectivity is reestablished.

[0122] A lower limitation on the number of root nodes is obtained either when the distance between adjacent head nodes is too great, translating into uncertain communication between clusters, which for the above-mentioned 2.4 GHz NordicVLSl nRF2401 transceiver means around 10 meters inside buildings otherwise, the head nodes will not be able to provide the expected throughput from the subnodes to the mouths, even if they sum the data in transmission.

[0123] There are two basic approaches. Firstly, it is possible to realize that the beacon rate is cluster specific, so that each head node can dynamically make decisions about its own beacon rate based on the number of subnodes it is currently associated with. If this is the case, then - considering only the power consumption - it is even advisable to program each FFD not to act as the head node if it can avoid it. However, it is not recommended that any root node reduce its beacon rate below a rating that determines the rated beacon period, the length of which should be sufficient to receive at least one network beacon from a root node that is close enough to its beacon signals. beacons may have been picked up. Dense network sections (&gt; 1 knot / m²) may benefit from realizing higher beacon speeds than the described ratings since the overall savings in energy required for network scanning are much greater than the additional energy used for beacon transmissions. The upper limit on the rate of beacons then occurs when collisions between network beacons from different root nodes become too common.

[0124] A second possibility is to determine a predetermined rating of the subnodes each head node should have in its cluster, and to establish a constant beacon rate based on the rated number of subnodes. As a consequence of the second alternative, if the new FFD reads the load information from the transmission of the neighboring master nodes' beacons and reports that they all have the full rating of the subnodes, it starts to act as a new head node and will not squeeze as an additional subnode into one of the existing ones. clusters. Even if the head nodes cannot change the beacon rate on their own initiative, it is possible to define the beacon rate as a network parameter that the managing head node can change by sending a corresponding change message over the network.

[0125] The head node need not comply with this command if the power consumption of the head node operation threatens to exhaust energy resources below a certain threshold or if the energy consumption rate exceeds a suitable replenishment rate (e.g., draw) for a specified period of time. Such a node device may simply refuse to operate as the head node and announce its withdrawal, no longer transmitting the beacons. Other nodes around it will notify you that the previous cluster is down and start looking for replacement connections. Alternatively, one FFD, previously operating as a subnode, will notify that it can only receive beacon transmissions at the highest transmit power level, meaning that it is relatively far away from all existing nodes and decides to switch to being the master node (or whatever is more preferably, it has been programmed not to do so or to postpone it as far as possible if its energy reserves appear to be low). It announces its decision by transmitting the cluster beacon, after which it waits for other nodes around it to send their binding frames. The former subnode may also decide to become the root node if it notifies via cluster beacon transmission that the load on its former cluster has become excessively heavy.

[0126] If two FFDs so decide to become new root nodes simultaneously, their beacon transmissions will collide. Given that the network channel will only include network beacons that are very short (2 x 256 microseconds) in the suggested beacon transmission system, the likelihood of collisions is low. However, if there is a need for further reduction, the deterministic behavior of one node can be used: the decision to become the root node will be made after meeting certain criteria, which - taking into account the influence of external factors - will occur in a specific common instance of time. It can be defined that after the moment when said criteria are met, each previous sub-node waits for a delay time, the length of which is either randomly selected or calculated by an algorithm that takes into account the energy of the node (the more energy, the shorter the delay), the remaining main nodes are known to the node (the smaller the number of known head nodes, the shorter the delay time; however, at least one remaining root node is required to be heard to maintain connectivity) and / or node mobility (if a node knows that it is very mobile, it uses more delay). As a result, the new head node is most likely to be the optimal choice in terms of energy demand and network scaling and topology.

OVERLOAD AND SCALABILITY

[0127] Even if there are many different frequency channels to choose from, in large networks it can happen that a new cluster simply cannot find a free frequency.

In that case, it may occupy the frequency channel where it considers there is the least traffic. The long idle time in access cycles, together with the fact that the network beacons are transmitted on a separate frequency, means that a single frequency can successfully find itself in more than one cluster if the newer root node chooses its superframe period in such a way that not to overlap the frames of previous head nodes.

[0128] Contrary to general belief, using TDMA to spread the available communication capability does not exclude the possibility of network scalability. The network has a cluster topology, with TDMA being used inside clusters (intra-cluster communication). Intra-cluster scalability is not a problem as all cluster members (subnodes) are within the scope of the cluster node (root node). Furthermore, the master node controls the number of subnodes associated with it. Moreover, the ALOHA windows enable a large number of low-activity subnodes to communicate with the head node on demand.

[0129] For inter-cluster communication, the scalability problem is solved by spreading TDMA control over the network. Network TDMA schedules are avoided. A Master Node that wishes to communicate with others, a Master Node, ties with an adjacent cluster in a manner similar to the subnodes operating there.

MINIMIZING DATA DELAY IN ONE DIRECTION

[0130] If each master node starts an access cycle in its cluster at any point in time, the starting times of the access cycles (and thus, e.g., cluster beacon transmission times) will be uniformly spread over time. This has an advantage: for example, spreading the use of transmission time avoids collisions and compensates for radio interference generated by nearby electronic devices. However, this also leads to the average inter-cluster hop-to-hop delay being substantially half the access cycle length. If the wireless sensor network is large and the access cycle is long (say 10 seconds), the overall delay for generating data on the subnode for it to become available at the outlet may be too long. This delay, due to the fact that each node that transmits data has to wait for the next suitable transmission moment, is generally known as data delay.

[0131] A simple but effective method for minimizing data delay is shown in Figs. 13 and 14. Fig. 13 is a simplified diagram of a network topology where subnodes are omitted. We assume that a subnode in the cluster of the first master node 1301 is generating data that should be routed by the second and third master nodes 1302 and 1303 to the exit node 1304. Further, we assume that the routing protocol has been implemented, causing each of the first, second, and third root nodes 1301, 1302, and 1303 to write a corresponding sequential hop record for routing data towards the exit node 1304. Relatively few exit nodes in a typical wireless network sensor function ensures that at most head nodes, the routing table only has one entry for the next hop towards the mouth node.

Referring to Fig. 14, each Principal Node has timed its access cycle such that its superframe, shown as a white segment in Fig. 14, immediately occurs before the superframe of another root node, which is indicated in the routing table. another hop towards the exit node 1304. Transceiver activity, which does not belong to the root node's own superframe, but relates to communication with a neighboring node, is shown as a dashed segment. If the first head node 1301 has taken data from a subnode in the pane of the super frame that appeared in the first

14, may receive cluster beacon 1401 from second root node 1302 next and forward the data and receive acknowledgment in step 1402, still in the same superframe of second root node 1302. Same high speed beacon reception order cluster and the data transfer with confirmation is repeated in steps 1411 and 1412, and 1421 and 1422. A similar data transmission chain is repeated in steps 1403, 1404, 1413, 1414, 1423, and 1424.

[0133] The th hopping delay is fixed and is the length of the super frame. The overall subnode-to-exit data delay corresponds to the length of the super frame minus the number of hops required between the clusters, adding one incomplete super frame length that occurred before the first head node could start transmitting data. An exemplary super frame length of 260 milliseconds has been discussed above, which is a much more acceptable value for inter-cluster data delay than an average of half a 10 second access cycle. In the case of high-speed transceivers capable of realizing a 1 Mbps transmission, there is a high probability of shortening the window length to e.g. one millisecond, which will further reduce the data delay by a decade.

[0134] As a limitation, we must remember that the delay constraint scheme described above is only valid for unidirectional transmission. In the opposite direction, it maximizes data latency, with a constant inter-cluster latency of virtually the entire access cycle length. For multipath support, each transmitting node must either select the favorable next hop entry from the routing table to optimize the access cycle time based on, or choose the number of equally favorable consecutive hop entries and the time of its access cycle to send data to any suitable adjacent master nodes followed the least possible delay given the different access cycle schedules.

[0135] Setting the access cycle time is best supported when the root node wants to change the access cycle start time, it announces a shorter or longer access cycle length than normal in one cluster beacon, which is only valid for one (or more) cycle times. after which the time is corrected and can revert to the normal access cycle length. Extremely early or late start of the immediately next access cycle should also be considered when entering values ​​into the "time to next cluster beacon" field that affects network beacons.