Sensor network

from Wikipedia, the free encyclopedia

A sensor network (from English wireless sensor network ) is a computer network of sensor nodes , tiny ("speck of dust") to relatively large ("shoebox") computers that communicate wirelessly , either in an infrastructure-based (base station) or in a self-organizing one Ad-hoc networks work together to query their surroundings using sensors and to forward the information. The envisaged size of future sensor nodes made the idea known under the catchphrase Smart Dust ("intelligent dust").

Sensor networks were designed as a military early warning system for monitoring pipelines and national borders. However, modern research also sees them as a replacement for costly sensor arrangements in vehicle construction, goods managers in warehouses and monitors of natural areas for pollutants, forest fires and animal migrations ; the conceivable applications are just as diverse as the sensors available (see sensors according to measured variable ).

Sensor networks are always in the stage of further development; there are practical applications for test and demonstration purposes. There are common sensor networks for professional applications. The best-known sensor network is that of the weather stations of various providers, although the networking is carried out using conventional telecommunications networks. Comparable networks of actuators are not known, since the energies required for actuators and the protection against malfunctions place considerably higher demands on the network and the nodes.

The smallest existing sensor node has a diameter of one millimeter (status: 2007), the largest sensor network to date covered an open area of ​​1300 by 300 meters with around 1000 sensor nodes (status: Dec 2004).

history

The Sound Surveillance System (SOSUS) can be regarded as a forerunner of modern sensor network research , a network of underwater buoys installed by the United States during the Cold War that tracks submarines using sound sensors. Although SOSUS is not a computer network, it gave rise to the idea of ​​a comprehensive sensor arrangement.

Sensor network research began around 1980 with the Distributed Sensor Networks (DSN) and Sensor Information Technology (SensIT) projects of the United States Military Defense Advanced Research Projects Agency (DARPA). DARPA works with military and university research institutions to develop new militarily and economically important technologies. As a rule, your results are not subject to confidentiality, which also applied to sensor network research.

In the 1990s, sensor network research experienced an upswing, supported by increasingly smaller and more powerful computer hardware . Today sensor networks are examined by research institutes all over the world. Results have been presented since 2003 at the " ACM Conference on Embedded Networked Sensor Systems " (SenSys).

hardware

A sensor node of type TelosB (Crossbow Technology). There is a battery compartment (2xAA) on the bottom.

A sensor node basically consists of a processor and a data memory (usually flash memory ), just like a normal computer . There are also one or more sensors and a module for radio communication. All parts are powered by a battery . In newer models, all components are housed on a single computer chip , which significantly reduces the size compared to assembled individual components.

Some designs provide for network nodes without sensors for communication and management purposes only. If all nodes of a network have the same sensor equipment, one speaks of a homogeneous sensor network, otherwise of a heterogeneous one . Heterogeneous sensor networks are particularly useful when the sensors have different usage patterns, i.e. when they are very different in terms of frequency, duration and volume of data.

Radio technology is currently traded as a promising means of communication , but other communication media such as light or sound have also been considered. It is assumed that the communication module, like a radio device, knows two states ready to send and ready to receive , between which it is possible to switch over with a short time delay.

Sensor nodes do not receive any new energy reserves after deployment; if the reserves are used up, the life of the node is also exhausted. The battery of a sensor node must therefore be as efficient as possible, while all other parts must have the lowest possible power consumption. In order to further reduce power consumption, each sensor node can be put into a standby state in which all parts except for the processor-internal clock are switched off. When switching off and on, one speaks of “going to sleep” and “waking up”, which results in regular “daily routines”. The complete abandonment of batteries through the use of renewable energies such as photovoltaics would be desirable, but is not feasible according to the current state of the art.

The desire for ecological compatibility is also a long way off: When their energy reserves are exhausted, sensor nodes deployed in the wild should rot without leaving any residue, and animals should not be harmed by accidentally swallowing a sensor node.

The hardware costs should be so low that large-scale sensor networks - DARPA speaks of hundreds of thousands of sensor nodes - can also be financially viable. Sensor nodes with a unit price of 2 € would be justifiable in vehicle construction, large-scale applications for landscape monitoring with several thousand nodes are only worthwhile from a unit price in the lower € -cent range; Today's sensor nodes rarely reach the lower price range and usually cost over € 100 each.

Existing systems

In the past, some sensor nodes have been developed that are used to test specially designed software and communication schemes. The nodes differ greatly in size, features and price because they follow different objectives: While some developers try to make their sensor nodes as small and cheap as possible, others rely on versatility and ease of use for scientific test applications. Known systems are:

  • BTnode . Sensor node platform from BTnodes, which can be individually expanded to include sensors. The current model BTnode rev3 has the dimensions 58.15 x 32.5 mm.
  • eyesIFX
  • FireFly from FireFly
  • iDwaRF , iDwaRF-328 and iDwaRF-Box . Radio modules based on Atmel AVR for the simple construction of wireless multipoint-to-point (N: 1) networks in the 2.4 GHz ISM band.
  • Imote , Mica and Telos . Sensor node platforms from Crossbow Technology, which can be individually expanded with sensors or come with standard equipment. The current models measure 36 × 48 × 9 mm (Imote2), 58 × 32 × 7 mm (Mica2) and 65 × 31 × 6 mm (Telos B).
  • INGA (Inexpensive Node for General Applications). OpenHardware sensor node, developed by the Institute for Operating Systems and Computer Networks at the TU Braunschweig .
  • iNODE (intelligent Network Operating Device) sensor node platform from Forschungszentrum Jülich in Flex-PCB design. 20 × 20 × 5 mm (folded)
  • iSense is a modular sensor network platform from coalesenses. In addition to a basic module with processor and radio interface, there are various sensor modules (acceleration sensor and passive infrared, temperature and brightness, magnetic sensor), energy modules and a gateway module.
  • Particles . Sensor node of the TecO of the University of Karlsruhe with temperature, light and acceleration sensor . The current model has a size of less than 10 mm³.
  • Preon32 , a new kind of sensor node with a virtual machine
  • Rene
  • ScatterWeb
  • s-net . Extremely energy-saving, wireless sensor networks from the Fraunhofer Institute for Integrated Circuits .
  • SNoW5 . Expandable sensor node from the University of Würzburg .
  • Sun SPOT . Project Sun SPOT sensor node platform.
  • TinyNode 584 . Expandable sensor node with temperature sensor from Shockfish SA. The flat knot has the dimensions 30 × 40 mm.
  • Tmote Sky . Sensor node with temperature, light and humidity sensor from Moteiv. The flat knot measures 32 × 80 mm.
  • Waspmote . modular sensor node with the option of attaching various sensors as an extension.
  • WeC
  • WiseNet
  • Z1 . Sensor node with temperature and acceleration sensor from Zolertia.

communication

Multi-hop network

Sensor networks form ad-hoc networks , i.e. networks without a fixed infrastructure between end devices. Ad-hoc networks are meshed networks in which network nodes are connected to one or more neighbors. This results in multi-hop communication in which messages are passed on from node to node until they have reached their destination.

Such networks are characterized by unpredictable, dynamic behavior, because, unlike permanently installed computer networks, the network topology is insecure: the number and locations of network nodes as well as the "line quality" cannot be foreseen, nodes can be added during operation or fail without warning.

Tasks of the network protocols

The communication between the sensor nodes is a central area of ​​current research. The aim is to find network protocols that transmit data as efficiently as possible and at the same time conserve the energy reserves of the sensor nodes by enabling long sleep times and addressing energy-intensive components such as the radio unit as rarely as possible.

A complete network protocol defines the behavior of the nodes in four points:

  • The initialization is the phase in which the sensor nodes find each other after deployment and establish the network topology by finding their neighbors. A clean structure of the network topology is decisive for the subsequent success of routing.
  • The daily routine is understood to mean the alternation between waking and sleeping times of the nodes. Since sleeping times save energy, but make nodes inaccessible, it is important to find a reasonable middle ground here.
  • The communication scheme determines how a single data exchange between two sensor nodes takes place. It must be ensured that the data is transmitted quickly and without errors and that the nodes do not interfere with one another.
  • The routing ultimately determines how messages are passed through the sensor network. The shortest route is not always the best here, as this could lead to a one-sided network load and thus premature failure of important connection nodes. Previous research has mostly approached these challenges separately and left it to the operator of the sensor network to put together a suitable process from the individual parts.

Sensor networks are particularly susceptible to the classic communication problems in computer networks, on the one hand because a large number of end devices share a common communication medium, on the other hand because sensor nodes are more affected by wasted resources than devices with a power grid connection or rechargeable energy storage. For sensor networks, only protocols that effectively avoid these problems are suitable.

Special sensor network protocols

Already in the early investigations of the military it became clear that conventional network protocols are not suitable for sensor networks. Even today's standards for radio networks such as IEEE 802.11 or Carrier Sense Multiple Access are too wasteful with the energy supplies of the end devices or, like Bluetooth, cannot be transferred to networks with a large number of participants. In addition, sensor networks stand out from other mobile ad-hoc networks in one important respect: While many different applications or users usually compete for common resources in a network, there is only one network-wide application in a sensor network competes with itself, as it were. Since the overall goal of the application has priority over the equal treatment of individual nodes, the concept of fairness in computer networks must be reinterpreted here.

Protocol research therefore develops and examines network protocols that are specially tailored to the needs of sensor networks. In doing so, it goes in different directions without a uniform standard having emerged so far. Some researchers argue that the areas of application of the sensor networks are so different that there will never be a protocol for sensor networks, but always a selection of protocols that are differently well suited for different purposes. The most important sensor network protocols are presented below.

Media access logs

A large group of sensor network protocols is dedicated to the common use of the communication medium (air) in the role of media access control (MAC, "media access control"). The reduction of energy consumption plays a primary role. This is in contrast to traditional wireless networks (WLAN, GSM), where the aim is to utilize the available bandwidth of the medium as fully as possible and at the same time to distribute it fairly.

The radio module is often the component of the sensor node that consumes the most energy. The energy consumption is similarly high for the different operating modes of the radio module (waiting for messages, receiving, sending). To save energy, the radio module is largely switched off ( Duty Cycling ). The MAC protocol must therefore not only decide when data is to be sent, but also when the radio module is to be switched on or off. Two methods are used: random access with carrier check and time division multiplexing .

In the case of random access with carrier checking , different variants of so-called low-power listening (LPL) are used. The idea of ​​LPL is that the radio is regularly stopped for a short time to check whether the medium is occupied. If this is not the case, the radio module is immediately switched off again to save energy. If the medium is occupied, the radio remains activated to exchange messages. For the sender, this approach has the difficulty of knowing when to send in order to ensure that the recipient is listening. The simple approach is to send a preamble that is longer than the recipient's wake-up interval ( Berkeley Media Access Control (B-MAC)). Alternatively, a long stream of repetitive packets can be sent (X-MAC, SpeckMAC). To save energy (and bandwidth), the transmitter can learn the receiver's wake-up schedule (WiseMAC). As an alternative to the LPL, the opposing low-power probing (LPP) can also be used. A short carrier (beacon) is regularly sent, which indicates that the node is ready for a short time to receive a message (RI-MAC).

With time division multiple access (TDMA), a schedule is created for which nodes transmit and receive when. This allows an energetically favorable data exchange. However, creating and maintaining the schedule and the required synchronization cause additional work. Protocols in this class are Sensor Media Access Control (S-MAC), Timeout Media Access Control (T-MAC), Dozer, SCP-MAC, LMAC, DMAC, TRAMA. With Dozer and DMAC it should be noted that MAC and routing are combined in one protocol.

Hybrid protocols such as Crankshaft, Zebra Media Access Control (Z-MAC) or SRTST-MAC try to combine the advantages of random access with carrier testing and TDMA.

Routing protocols

Routing protocols are primarily dedicated to routing, i.e. the question of how messages can be routed to their destination as quickly as possible and with as little effort as possible. Network protocols that ignore the question of routing usually start from standard procedures based on routing tables (see routing ). In fact, many routing protocols can be transferred to sensor networks with little or no adjustments.

Geographic routing methods are of particular importance for sensor networks . In many application scenarios, the user is specifically interested in measurement data from a specific geographic area or point. On the one hand, there are queries of the type “Deliver me all the data for area xyz” , and secondly, nodes are addressed with information such as “To the node that is closest to position xy” . The network protocol must relieve the user of the task of locating the affected nodes and forwarding messages to them.

The Geo-Cast method searches for all nodes in a selected geographical area from a sensor network. By fitting and cutting geometric shapes on a map, the affected sensor nodes can be quickly identified. At the same time, the user receives an easy-to-use graphical user interface.

The sensor network protocol Greedy Perimeter Stateless Routing in Wireless Networks (GPSR) plays a central role in routing . It forwards messages to geographical coordinates rather than names. It alternates repeatedly between a greedy strategy , in which data packets are passed on in a straight line towards the destination, and a perimeter mode, in which the data packet circles the destination point. The perimeter mode is intended to ensure that packets do not get stuck in dead ends in unfavorable network topologies. Geographic Hash Tables extend GPSR by the possibility of distributing information to several neighboring nodes and thereby ensuring data security in the event of failure of some nodes.

The routing protocol for sensor networks that is being developed by the IETF is RPL .

Protocol stacks

There are currently several competing protocol stacks from various consortia and organizations. Depending on the stack, some or all layers of the OSI model are covered:

Positioning and location

Certain application scenarios and communication protocols require that a sensor node can determine its own location ( location determination ) or the original locations of measured signals ( location ). Since both questions have been dealt with in other areas such as navigation and astrophysics for centuries, there are now a large number of methods for the most varied of starting conditions. However, it must be checked which methods can be implemented with the limited technology of the sensor networks and how the work that arises is sensibly distributed to the nodes of the network. For example, the widespread satellite positioning using the Global Positioning System (GPS) is unsuitable for sensor nodes because the necessary technical components are too large, heavy and expensive.

If there are at least two sensor nodes in a sensor network that know their own position in absolute geographic coordinates, and it is possible to measure the distance between two sensor nodes, each node in the network can generally determine its position in the geographic coordinate system. The idea is that each sensor node initially sets up a personal coordinate system, with itself at the origin and two neighboring nodes as directional indicators for the x and y coordinate axes. Using methods such as triangulation , each node assigns all neighbors it has heard to in its personal system. Then the individual systems are merged into an overall coordinate system by rotating and shifting. If the network is sparsely occupied or arranged in an unfavorable way, the location determination remains imprecise.

If no two nodes can determine their absolute position or if distances cannot be estimated, the location determination remains incomplete or imprecise. For example, if no sensor node knows its absolute coordinates at all, the sensor network can, under certain circumstances, be mapped geographically correctly, but cannot be placed in the larger context of the world coordinates. If, on the other hand, the possibility of distance measurement is missing, triangulation and similar methods are omitted and the positions can only be guessed at as intersections of several radii. If both initial conditions are not met, the network topology can only be mapped abstractly as a graph or an equivalent representation (such as an adjacency matrix or incidence matrix ).

A comparatively recent approach is the location determination by fingerprints (English fingerprints ). By listening to the radio channel, a node creates a “fingerprint” profile of the background noise. The background noise is affected by the environment, e.g. B. nearby electrical lines or walls on which radio waves are reflected, and thus differs from place to place. By comparing it with a fingerprint database, a sensor node can estimate its own position. This approach requires prior knowledge of the area of ​​application or an additional auxiliary system.

In order to be able to clearly locate the origin of a measured signal, the signal must have been received by at least three sensor nodes. The signal origin can be determined exactly from the different signal transit times to the sensor nodes via hyperbolic location. If the signal is only received by two sensor nodes or fewer, unambiguous location is not possible and the origin can only be limited by intersecting the transmission radii or the transmission radius alone.

synchronization

Measurement data are often dependent on absolute times. In addition, some communication protocols, such as B. SMACS the most precise possible synchronization of the sensor nodes with each other.

As in other computer networks , sensor networks also have to struggle with the typical inaccuracies in synchronization. The factors that influence the synchronization are the transmission time , i.e. the time the sender needs to be ready to send, the access time, i.e. the time the sender needs to store the data on the communication medium, the speed of propagation , i.e. how fast the message is from the sender to the recipient, and the reception time , i.e. how long the recipient needs to pick up a message from the medium and make the information accessible to the application in question. Since the different communication methods , in addition to other factors, mainly influence the access time , it makes sense to make the decision for a synchronization method also dependent on the communication method.

Synchronization by calculating the round trip time

You can determine the difference between two clocks in a computer network by subtracting the times of two computers from each other and then subtracting the round trip time again , which is caused by the messages that the two computers exchange to tell each other about their times to inform. In practice, this is done by calculating the mean value of the time spans that were required to transmit the request and its response. With this method, all of the above-mentioned factors have an effect except the speed of propagation . This method is therefore preferable if there is a high variance in the speed of propagation.

Reference broadcast synchronization

With reference broadcast synchronization (RBS), a synchronization signal is sent to all nodes from a central point. A node that sends a message after this synchronization message informs the recipient at the same time with this message that it has received the synchronization. With the help of this information, the receiver can then decide whether his clock is wrong and synchronize it with the clock of the transmitter. This method is particularly well suited if the transmission time and the access time vary during communication , as the time on which the synchronization is based is only sent once for all recipients and different access or transmission times do not affect the synchronization .

Timing-Sync Protocol for Sensor Networks

The Timing-Sync Protocol for Sensor Networks (TPSN) describes the process of how the synchronization is distributed in a sensor network. Synchronization by calculating the round trip time can be used as the synchronization method .

initialization

  1. A root node is level 0.
  2. The root sends a level_discovery message by broadcast to all nodes in its coverage area.
  3. All nodes that receive a level_discovery that is lower than their own level accept the level of the message increased by 1, wait a random time and then begin with point 2.

Since collisions can still occur despite the random waiting time, there is also the possibility that nodes request a level with a level_request . These are then based on the lowest level they receive.

synchronization

  1. The root node uses a time_sync packet to request level 1 nodes to obtain information about the time from it.
  2. If a node receives a time_sync , it requests the current time with a synchronization_pulse . Level X nodes also become active when the synchronization_pulse is received and, in turn, request the time from their parent node.
  3. The parent node responds to the synchronization_pulse with an ACK that contains the current time.

Aggregation

Some application scenarios call for data from the entire sensor network to be collected and ultimately transmitted to a single recipient (“central sink”). Naive approaches to performing such an aggregation can lead to the formation of communication bottlenecks with large amounts of data, which unnecessarily reduce the performance of the system. Data bundling and data compression help to avoid such bottlenecks. If, for example, the maximum measured temperature is to be determined in a sensor network, the naive approach would be to transmit all measured temperatures to the central sink, which then picks out the maximum, while an advanced approach compares the data as they are passed on and ultimately only one transmitted single temperature value to the central sink.

Tiny aggregation (TAG)

Tiny Aggregation (TAG) treats the sensor network like a database from which data is queried using a database query language. Data requests are propagated from the central sink into the sensor network in a simple, SQL- like format. The nodes evaluate the request based on their own data and data received from neighbors and sort out redundant and superfluous data step by step in advance.

Empirical mutual coding

The empirical mutual coding only passes on information if it does not correspond to the normal value. The principle is already implemented in naive approaches, for example when a fire-fighting sensor node only transmits temperature measurements that exceed a specified value. The empirical mutual coding deepens this idea due to the observation that the measurements of sensor nodes that are spatially close to one another are always similar. A sensor node indicates its measurement relative to that of its neighbor and only passes it on if the measured value deviates significantly from that of the other. The strength of the approach is that the correlations of the measured values ​​are determined automatically.

Criticism of sensor networks

Michael Crichton drew a gloomy vision of the future in his novel Beute in 2002 , in which he combined the smart dust idea with collective intelligence and nanotechnology and had his fictional characters killed by swarms of malicious microparticles.

Far more realistic criticism of sensor networks and the smart dust idea are expressed by data protectionists . You see another monitoring method in sensor networks that can be misused to monitor citizens and analyze consumers without their knowledge or consent .

See also

References

  1. ^ W. Dargie and C. Poellabauer, "Fundamentals of wireless sensor networks: theory and practice", John Wiley and Sons, 2010 ISBN 978-0-470-99765-9
  2. ↑ By 2020 we should have chips with components only a few nanometers in size. K. Pister: Autonomous sensing and communication in a cubic millimeter
  3. ↑ Analyzing the Yield of ExScal, a Large-Scale Wireless Sensor Network Experiment last visited on August 21, 2014.
  4. ↑ iDwaRF: Freely programmable AVR radio modules and boards for wireless radio sensor networks. ( Memento of the original from April 5, 2015 in the Internet Archive ) Info: The archive link was inserted automatically and has not yet been checked. Please check the original and archive link according to the instructions and then remove this notice. @1@ 2Template: Webachiv / IABot / www.chip45.com
  5. ↑ Preon32: radio module with superior technology. Accessed on November 10, 2017
  6. ↑ s-net: Fraunhofer IIS technology for extremely energy-saving, wireless sensor networks. Accessed on October 15, 2013
  7. ↑ Waspmote: The sensor device for developers Retrieved 5 December 2010
  8. ↑ I. Demirkol, C. Ersoy, F. Alagöz: MAC Protocols for Wireless Sensor Networks: A Survey. (PDF; 244 kB) In: IEEE Communications. 44 (4), pp. 115-121. April 2006
  9. ↑ R. Mati Shek: Real-Time Communication MAC Protocols for Wireless Sensor Networks, 2012, ISBN 978-3830063490 , see "Soft real-time shared time slot" (SRTST) MAC Protocol, pp. 107-128

Web links

Research projects