LNCS 4621 - Applications of Sensor Networks

1 Applications of Sensor Networks Hans-Joachim Hof

1.1 Introduction Networks of small sensor nodes, so-called sensor networks, allow to monitor and analyze complex phenomena over a large region and for a long period of time. Recent advances in sensor network research allow for small and cheap sensor nodes which can obtain a lot of data about physical values, e.g. temperature, humidity, lightning condition, pressure, noise level, carbon dioxide level, oxygen level, soil makeup, soil moisture, magnetic field, current characteristics of objects such as speed, direction, and size, the presence or absence of certain kinds of objects, and all kinds of values about machinery, e.g. mechanical stress level or movement. This huge choice of options allow to use sensor network applications in a number of scenarios, e.g. habitat and environment monitoring, health care, military surveillance, industrial machinery surveillance, home automation, als well as smart and interactive places. The application of a sensor network usually determines the design of the sensor nodes and the design of the network itself. No general architecture for sensor networks exists at the moment. This chapter gives an overview of existing sensor network applications, shows some currently available sensor network hardware platforms and gives an outlook on upcoming applications for sensor networks. A sensor network can be of great benefit when used in areas where dense monitoring and analysis of complex phenomena over a large region is required for a long period of time. The design criteria and requirements of sensor networks differ from application to application. Some typical requirements are: – –

Failure resistance: Sensor networks are prone to node failure. Thus, it is crucial that algorithms for sensor networks can deal with node failures in an efficient way. Scalability: As sensor nodes get cheaper and cheaper, it is highly likely that the sensor networks consist of a huge number of nodes. Algorithms

D. Wagner and R. Wattenhofer (Eds.): Algorithms for Sensor and Ad Hoc Networks, LNCS 4621, pp. 1–20, 2007. c Springer-Verlag Berlin Heidelberg 2007 ⃝

2

–

–

–

H.-J. Hof

for sensor networks must therefore scale well with thousands and tens of thousands of sensor nodes. Simplicity: Due to the design of sensor nodes, the available resources for algorithms of the application layer are severely restricted. Algorithms for sensor networks must be very efficient considering computation cycles and memory usage. Unattended operation: Sensor networks are usually deployed in unattended areas, or administration tasks should be kept at a minimum. It is therefore necessary for an algorithm for sensor networks to work unattended after the deployment of the sensor node. However, it is usually no problem to configure sensor nodes before or during deployment. Dynamic: Sensor nodes must adapt to changes of the environment, e.g. changed connectivity or changing environmental stimuli.

This paper is structured as follows: First, some applications of sensor networks are presented. Second, some common sensor network hardware platforms are discussed. The paper ends with an outlook on upcoming applications.

1.2 Applications of Sensor Networks In this chapter, current sensor network applications are shown. The chapter covers habitat monitoring, environment monitoring, health care, and industrial applications. 1.2.1 Habitat Monitoring The main objective of habitat monitoring is to track and observe wild life. In the past, habitat monitoring has been done by researchers hiding and observing the wild life, or cameras were used for observations. However, these techniques are intrusive and uncomfortable, and long term observations are difficult and expensive. Usually, live data is not available. Sensor networks offer a better way for habitat monitoring. The sensor network technology is less intrusive than any other technique. Thus, the wild life is less affected, resulting in better research results. Also, long-term observations are possible and sensor networks can be designed in a way that live data is available on the Internet. Human interaction is usually needed only for setup of the sensor network and for removal of the sensors after the end of the observation. Hence, sensor networks help to reduce the costs of habitat monitoring research projects. Two examples of habitat monitoring, the Great Duck Island project and ZebraNet, are presented in the following. Great Duck Island. The Great Duck Island project [273] was one of the first applications of sensor networks in habitat monitoring research. The main objective of the research project was to monitor the micro climates (e.g. temperature and humidity) in and around nesting burrows used by the Leach’s

1

Applications of Sensor Networks

3

Storm Petrel. The great advantage of this sensor network compared to standard habitat monitoring was that it is non-intrusive and non-disruptive. The project is named after a small island at the coast of Maine, Great Duck Island, where the research took place. At first, a network of 32 sensor nodes was deployed (see Figure 1.1). The sensor network platform consists of processor radio boards commonly referred to as motes. MICA Motes (see Chapter 1.3.1) were used as sensor nodes. They were manually placed in the nesting burrows by researchers. The sensor nodes periodically sent their sensor readings to a base station and got back to sleep mode. The base station used a satellite link to offer access to real-time data over the Internet. To get information about the micro climate in the nesting burrows, the sensor nodes collected data about temperature, humidity, barometric pressure and mid-range infrared. Between spring 2002 and November 2002, over 1 million of sensor readings were logged from the sensor network. In June 2003, a larger sensor network, consisting of 56 nodes, was deployed, and it was extended in July 2003 by 49 additional sensor nodes and again augmented by 60 more sensor nodes and 25 weather station nodes in August 2003. Hence, the network consisted of more than 100 sensor nodes at the end of 2003. The network used multi-hop routing from the nodes to the base station. The software of the sensor nodes was based on the sensor network operating system TinyOS [364]. The sensor network of the Great Duck Island project was preconfigured and did not self configure, e.g. each sensor node got assigned a unique network layer address during compilation of the code prior to deployment.

Fig. 1.1. Placement of sensor nodes in the Great Duck Island project: sensor nodes were placed in the nesting burrows of storm petrels (1) and outside of the burrow (2). Sensor readings are relayed to a base station (3), which transmits them to a laptop in the research station (4), that sends it via satellite (5) to a lab in California. Image source: http://www.wired.com/wired/archive/11.12/network.html

4

H.-J. Hof

ZebraNet. Another habitat monitoring project that uses a sensor network is ZebraNet [403]. The goal of the ZebraNet research project is to understand the migration patterns of zebras and how these patterns are affected by changes in environmental conditions like the weather or plant life. While the Great Duck Island project observed the nestings of birds and not the birds, in ZebraNet the sensor nodes are attached to the animals themselves. Those sensor nodes are integrated into a collar. For deployment of the sensor nodes, zebras are caught using a tranquilizer gun. The sensor collar is then mounted at the necks of the unconscious zebras. The design of the sensor nodes is limited by the maximum weight that a zebra can carry on its neck without being severely hindered. Figure 1.2 shows a zebra equipped with such a sensor collar. The design of the sensor nodes and its deployments make the ZebraNet project different from the Great Duck Island project: while in the first project the sensor nodes are mobile, they are fixed in the second project. This also has implications for the design of the base station. As zebras have a quite large territory and the sensor nodes have a limited communication range, it is not guaranteed that a sensor node has always a communication path to the base station. Therefore, the base station of ZebraNet is also mobile. It can be realized by a jeep which regularly drives through the zebras territory, or by a plane that regularly flies over the territory. The sensor nodes of ZebraNet collect biometric data of the zebras, e.g. heart rate, body temperature, and the frequency of feeding. The main focus of the research project lies on the migration pattern of zebras. Therefore, the sensors acquire a GPS position regularly. Latency is not an issue in ZebraNet,

Fig. 1.2. Zebra with mounted sensor collar. Image source: http://web.princeton. edu/sites/pumpala/pumpala/princeton%20research/princeton_research.htm

1

Applications of Sensor Networks

5

but reliable delivery of data is important. ZebraNet’s sensor nodes send data to peer sensor nodes, and each sensor node stores its own sensor data and those data of other sensors that were in range in the past. They upload data to a mobile base station whenever it is available. However, as data is swapped from node to node, it is not necessary to encounter all animals to get full data coverage. Sensor nodes use two communication modules: one for short-range transmission and one for long-range transmission. Of course, the short-range transmitter is much more energy-efficient than the long-range transmitter. The sensor collars are equipped with a solar panel to recharge the battery of the sensor nodes. However, the system is designed in such a way that a sensor node can run without any recharge for 5 days. The sensor nodes of ZebraNet use a complex duty cycle to allow a network lifetime of one year: – – – –

–

30 GPS position samples are taken per hour for 24 hours a day. A detailed zebra activity log is taken for 3 minutes every hour. The sensor nodes search for other sensor nodes in their direct neighborhood for a total of 6 hours per day. In this time, data is transfered between neighbor nodes using low-power short-range radio. The sensor nodes search for the mobile base station for a total of 3 hours a day. However, the time interval a sensor node searches for the base station overlaps with the time interval the sensor node searches for other sensor nodes in the direct neighborhood. To search for the base station, long-range radio communication is used. During a 5 day period, a total of 640 kilobytes of data is sent to the mobile base station using long-range radio communication.

Data storage is limited, therefore there is a prioritisation when to delete data: First, a node deletes data obtained from other nodes. Then, it deletes its own data, starting with the data with the oldest timestamps. If data is transmitted to the base station, it is registered in a delete list. The delete list is also communicated to peer sensor nodes so they know which data they can erase. All incoming data is compared to the delete list to avoid storage of data which is already known to the base station. 1.2.2 Environment Monitoring While habitat monitoring deals with observation of animals and their surroundings, the task of environmental monitoring is to sense the state of the environment. Sensor networks in environment monitoring are extremely helpful in biology research and they allow exploration of areas that were not accessible up to date. In the past, a small number of wired sensors was used in biology. With a sensor network consisting of a large number of nodes, it is possible to explore the macrocosmos in a much better way because it is easier to obtain data of natural phenomena at many different places. Hence, better prediction models can be build. Other applications of environmental monitoring include structural monitoring, which ensures the integrity of bridges,

6

H.-J. Hof

Fig. 1.3. Berkeley motes in a redwood tree. Image source: http://www.cs. berkeley.edu/˜culler/talks/mobihoc.ppt

buildings, and other man-made structures. In the following, four environment monitoring research projects are presented. Redwood Ecophysiology. A lot of research has been done to research the micro climates of forests. However, past research work obtained data either at the ground or above the canopy of the trees. The goal of the Redwood Ecophysiology project [92] is to learn about the ecophysiology (ecophysiology studies the interaction of physiological traits with the abiotic environment) of redwood trees, especially about the micro climate of the redwood trees between the ground and the canopy. They chose to use a sensor network for this research project. The sensor network was deployed by manually placing sensor nodes on different elevations of a number of redwood trees. Each tree was equipped with 40 to 50 sensor nodes. Figure 1.3 shows some sensor nodes in a redwood tree. The sensors measured sapflow, temperature, and humidity. The sensor nodes used a simple sleep schedule: each single sensor reported its sensor data every five minutes to a data logger and it was inactive the rest of the time. The experiment lasted for 25 days. Figure 1.4 shows a schematic view of the node placement. Meteorology and Hydrology in Yosemite National Park. In [269], a sensor network was deployed to monitor the water system across and within the Sierra Nevada, especially regarding natural climate fluctuation, global warming, and the growing needs of water consumers. Research of the water system in the Sierra Nevada is difficult, because the monitoring takes place in severe terrain with limited access. Therefore, the sensor network was designed

1

Applications of Sensor Networks

7

Fig. 1.4. Schematic view of node placement in the Redwood ecophysiology project. Image source: http://www.cs.berkeley.edu/~culler/talks/mobihoc.ppt)

to operate with little or no human interaction. A prototype network of meteorological and hydrological sensors was deployed in Yosemite National Park in different elevation zones ranging from 1200 to 3700 meters. The network used different communication technologies to fit in the area: radio, cell-phone, landline, and satellite transmissions. Twenty sensors were installed in basins. They hourly reported water level and temperature of the water. Also, 25 meteorological stations were installed, which measured air temperature and relative humidity. It turned out that communication was the major problem in areas like the Yosemite National Park because the park has a high relief. For the project’s success it was important to monitor the environment conditions in river valleys where the communication conditions are especially challenging. Also, battery life and long-term data logger backup were major design issues. Low power, low cost data loggers were important parts of the sensor network. These loggers were equipped with 32 MB of memory. Energy was the limiting factor, and solar panels were out of scope, because their large, reflecting surface conflicts with wildlife protection. GlacsWeb. Glaciers are very important in climate research because climate changes can be identified by observing size and movement of the glaciers. GlacsWeb [280] is a sensor network for monitoring glaciers. A sample network was installed at Briksdalsbreen, Norway, in 2004. The aim of GlacsWeb is to understand glacier dynamics in response to climate change. Sensor nodes (see

8

H.-J. Hof

Fig. 1.5. Basestation of GlacsWeb. Image source: http://envisense.org/ glacsweb/norway04/index.html

Figure 1.6) were deployed into the ice and on the till. A base station (see Figure 1.5) was installed on the surface of the glacier. The sensor network was designed to acquire data about weather conditions and the GPS position at the base station from the surface of the glacier. The sensor nodes used a simple duty cycle. Every sensor node sampled every 4 hours temperature, strain (due to stress from the ice), pressure (if immersed in water), orientation (in 3 dimensions), resistivity (to determine if the sensor is sitting in sediment till, water, or ice) and battery voltage. Thus, there were 6 sets of readings for each sensor node every day. The sensor nodes directly communicated with the base station once a day at a fixed time. All sensor nodes were queried during a 5 minutes interval each day. The base station recorded its location once a week using GPS. Obtaining the GPS position took 10 minutes. During those 10 minutes, remote administration of the base station was also possible. After this, the base station sent all its readings to a reference station PC via long range radio. The radio range of the sensor nodes was significantly reduced in ice to less than 40 m while it is 500 meters in air. The sensor nodes had a clock drift of 2 seconds a day. The base station synchronized them daily. The base station was equipped with a solar panel which failed when the panels

1

Applications of Sensor Networks

9

Fig. 1.6. A sensor node of GlacsWeb. Image source: http://radio.weblogs. com/0105910/2004/05/31.html

were covered with snow. Future versions of the base station will also have a wind turbine. The Four Seasons Project. Structure health monitoring of human-made structures is especially important in areas with frequent earthquakes. California is such an area. One of the buildings damaged during the 1994 earthquake in California is the Four Seasons building in Sherman Oaks (see Figure 1.7). This building had been used in the Four Seasons project [394] for structural health monitoring research. A sensor network had been deployed in the building to find out what the reasons for the observed damages were. The building is scheduled for demolition and was used for a series of force-vibration tests. MICA Motes (see Chapter 1.3.1) were used for the sensor nodes. They acquired vibration data at different locations in the building. Figure 1.8 shows one of the vibration sensors. The sensor nodes organized themselves in a large sensor network and sent their data to a base station. Special effort was made to have reliable data transport. Another problem was data compression. As the Four Seasons project is a data acquisition system which needs high accuracy, a lossy compression scheme could not be used. Thus, an algorithm was used that encodes sequences of samples whose values lie within a small range (a so-called silence period) in "run-length code", meaning that one sample value and the number of samples is given. This compression scheme is easy to compute and in the cases where there is a lot of activity, data is given totally accurate. Another problem was synchronization of vibration data collected by different nodes. Achieving global time synchronization would have involved too much effort. Thus, the total time of the packages in the network is calculated. The algorithm for time synchronization takes as granted that the time a node needs to handle a package is multiple attitudes higher than the time delay of transmission. Hence, each node added the time it needed for proceeding a package to a header field.

10

H.-J. Hof

Fig. 1.7. Four Seasons Building, which was equipped with sensor nodes for structural health monitoring. Image source: http://research.cens.ucla.edu/pls/ portal/docs/1/12805.JPG

Fig. 1.8. One of the vibration sensors used in the Four Seasons Building. Image source: http://research.cens.ucla.edu/pls/portal/docs/1/12821.JPG

1.2.3 Health Care In health care, especially in hospitals, a lot of sensors are used today to monitor the vital signs and states of patients 24 hours a day. Most of those sensors are wired, so patients are often severely restricted in their mobility. Hence, the use of wireless sensors would result in a great gain in life quality for the patients and it even would give them more security, because an automated system may react fast on emergency situations. Another issue in health care is the seamless transfer of patients between first aid personal, emergency rescuers and doctors at a hospital, so that a seamless monitoring and data transfer can take place. Sensor networks can solve a lot of problems in health care scenarios. The most important design criteria for applications in health care are security and reliability. CodeBlue. CodeBlue [263] is a system to enhance emergency medical care with the goal to have a seamless patient transfer between a disaster area and

1

Applications of Sensor Networks

11

a hospital. Wearable vital sign sensors (body sensors) are used to track a patient’s status and location. They also operated as active tags, which enhance first responders ability to assess patients on the scene, ensure seamless transfer of data among emergency personal, and enable an efficient allocation of hospital resources. Current body sensors are able to obtain heart rate, oxygen saturation, end-tidal CO2 , and serum chemistries measurements. The body sensors of the CodeBlue project are based on MICA2 Motes (see Chapter 1.3.1). Figure 1.9 shows a typical body sensor node. The network of CodeBlue does not rely on any infrastructure but is formed ad hoc. It is intended to span a large disaster area or a whole building, e.g. a hospital. The system is designed to scale with a very dense network consisting of lots of nodes. The nodes of the network are heterogeneous and range from simple body sensors and PDAs to PCs. CodeBlue addresses data discovery and data naming, robust routing, prioritisation of critical data, security, and tracking of device locations. For data discovery, a publish/subscribe-model is used. Body sensors publish its data and devices used by nurses or doctors, e.g. a PDA, can subscribe to that data stream. CodeBlue also uses data filtering and data aggregation. For example, a doctor may be interested in the full data stream of one patient, and wants only to be notified if some other patients are in an unusual or critical state. The CodeBlue project uses a best-effort security model: if an external authority can be reached, strong guarantees are needed to access data. However, if no external authorities are available because of poor connectivity or infrastructure loss, weaker guarantees may be used. This situation may arise for example in a disaster area where it is not possible to spend time setting up an infrastructure, keying in passwords or exchanging keys.

Fig. 1.9. A body sensor node of the CodeBlue Project which is build into a wristband. Image source: http://www.eecs.harvard.edu/~mdw/proj/codeblue/

12

H.-J. Hof

1.2.4 Military Application Sensor networks in military applications are often used for surveillance missions. The focus of surveillance missions is to collect or verify as much information as possible about the enemy’s capabilites and about the positions of hostile targets. Sensor networks are used to replace soldiers because surveillance missions often involve high risks and require a high degree of stealthiness. Hence, the ability to deploy unmanned surveillance missions, by using wireless sensor networks, is of great practical importance for the military. EnviroTrack. EnviroTrack is a middleware for tracking of objects. In [179] the design and implementation of a system for energy-efficient surveillance based on EnviroTrack is described. The main objective of the system is to track the positions of moving vehicles in an energy-efficient and stealthy manner. For the prototype, a network consisting of 70 sensor nodes was used. The sensor nodes are based on MICA2 Motes (see Chapter 1.3.1). They are equipped with dual-axis magnetometers. Issues in the design of the network were long network lifetime, adjustable sensitivity, stealthiness, and effectiveness. To prolong the network lifetime, it is important to use as little energy as possible. The system allows for a trade-off between energy consumption of the sensor nodes and the accuracy of the tracking. To save energy, only a subset of nodes, the so-called sentries, are active at a given time. The sentries monitor events. When an event occurs, the sentry nodes awake the other sensor nodes of the network and form groups for collaborate tracking. Hence, in the absence of an event, most sensor nodes are in a sleep mode, using only very little energy. The sensor nodes can adjust the sensitivity of their sensors to adapt to different terrains and security requirements. To achieve stealthiness, zero communication exposure is desired in the absence of significant events. The sensor network can only be used in challenging military scenarios if the tracking is effective. Especially, the estimated location of a moving object must be precise enough and the event must be communicated as fast and reliable as possible to the base station. As the collaborative detection and tracking process relies on the spatio-temporal correlation between the tracking reports sent by multiple sensor motes, it is important that the sensor nodes are timesynchronized and that the positions of the sensor nodes are known. Counter Sniper System. Snipers are a danger to military operations, especially because they are usually difficult to locate. The aim of the counter sniper system PinPtr [349] is to detect and accurately locate shooters based on acoustic signals. PinPtr allows to locate a sniper with one meter accuracy in average and the tracking of a shot needs 2 seconds. The system consists of a large number of sensor nodes, which communicate through an ad hoc wireless network. The large number of sensors is needed to overcome multi-path effects of the acoustic signals. All sensor nodes report to a base station, which is also

1

Applications of Sensor Networks

13

used by the user to access the system. The sensors measure acoustic signals, in detail muzzle blasts and shock waves of guns. The sensor nodes of PinPtr are based on MICA2 Motes (see Figure 1.10). A special sensor board was implemented and connected to the MICA2 Motes. On this sensor board, an FPGA is responsible for fast signal processing. Precise time synchronization is crucial for the application because the application uses the fact that an acoustic signal travels with a certain velocity. PinPtr uses the Flooding Time Synchronization Protocol [277] for time synchronization. Communication problems do not arise from multi-hop communication but from the situation that in the case of an event (a shot), a lot of sensors want to send their sensor data at the same time. This problem is solved by data aggregation at the sensor nodes. To allow fast response of the system after an event, the base station must receive as many sensor data as possible in the first second after the event. PinPtr uses a fast, gradient-based "best-effort" converge-cast protocol with built-in data aggregation. Data needs to include the position of a sensor node so that it can be used to calculate the position of a sniper. PinPtr uses a self-localization algorithm based on acoustic range estimation between pairs of sensor nodes. This method can be used to locate sensors in a range of 9 meters and the accuracy is 10 centimeters.

Fig. 1.10. A sensor node of PinPtr. Image source: http://www.isis.vanderbilt. edu/projects/nest/applications.html

1.2.5 Industrial Application Industrial production relies on more and more complex machinery. Breakdown of a machine can severely interfere the production, resulting in a loss of money. Thus, a close observation of industrial machinery is of advantage.

14

H.-J. Hof

Condition Based Monitoring. Maintenance of machinery usually prevents unplanned breakdowns. There are different ways for machinery maintenance: Run-to-fail maintenance does not include any maintenance at all. A machine is repaired or changed when it fails. Scheduled maintenance has a schedule when to do what kind of maintenance. The schedule may for example state that the oil of a machinery must be changed every 3 month. While prolonging the lifetime of a machine, scheduled maintenance is based on expert knowledge and not on the actual state of the machine. Thus, it is possible that a machine is maintained even if it does not need maintenance, resulting in unnecessary costs and outage time. Condition-based maintenance deals with this disadvantage. When using condition-based maintenance, a machine is maintained regarding its condition. However, prior to condition-based maintenance, it is necessary to obtain the current condition of the machine. Sensor networks may be used for this. In [82] the company Rockwell Scientific ported existing diagnostic routines and expert systems to wireless sensor network hardware and modified the software for autonomous data collection and analysis. The main task in developing diagnostic software for a sensor network was to abstract and design the algorithms independent from the actual machine so the sensors did not need to know on which machine they were running. 1.2.6 Home Automation and Smart and Interactive Places The goal of home automation and smart and interactive places is to offer people context-aware assistance with the tasks at their hands. Easy installation and wireless communication are important to allow a user to augment his space in a convenient way. Smart Kindergarten. In the Smart Kindergarten project [351], a classroom was equipped with a sensor network to learn about behaviour patterns of children in a group setting. The goal of the research project was to learn from the Smart Kindergarten classroom how to create more interactive learning environments and how to improve the quality of teaching. The sensor nodes were integrated in tags which the children wore (see Figure 1.11), and in everyday objects like textbooks or tabletops. The sensors obtained three data sets: The physical location, the orientation, and the speech of a child. All sensor nodes sent their data to a central database which was used later for evaluation of data.

1.3 Current Hardware Platforms Sensor network design is highly application specific. To achieve the goal of the application, it is necessary to choose the right sensor network hardware platform, because the limitations of the hardware greatly influence the design of the sensor network, its protocols, and its algorithms. Some popular platforms

1

Applications of Sensor Networks

15

Fig. 1.11. A child equipped with a Smart Kindergarten Badge. Image source: http://nesl.ee.ucla.edu/projects/smartkg/photos.htm

will be introduced in this chapter to give the reader a basic understanding of hardware limitations. In the following, the MICA hardware platform (MICA2, MICAz, MICA2DOT), the Intel hardware platform (MOTE2), the Telos platform (TelosB), the BTnode platform and the ScatterWeb platform (Electronic Sensor Board) will be described in brief. 1.3.1 The MICA Sensor Network Hardware Platform The MICA sensor network hardware platform was one of the first available hardware platforms, hence it is widely used. The platform was developed by the University of Berkeley and is now maintained by Crossbow Technology. Nodes of the MICA platform, the so called MICA Motes, use an Atmel ATmega128L micro controller [21]. The ATmega128L has an 8-bit RISC Harvard architecture. Hence, it uses different address spaces for code and data. A 128 KByte Flash memory is used to store code. 4 KBytes of SRAM are available for data. In addition, the Atmel ATmega128L has 4 KBytes of EEPROM for nonvolatile storage of data. The clock frequency of the micro controller can be regulated in a range from 0 to 8 MHz by software. The Atmel ATmega128L has two 8-bit timers, two 16-bit timers, and a 10 bit A/D converter. The AT-

16

H.-J. Hof

mega has a JTAG interface, an SPI bus interface, and UARTs. The CPU can be in active mode, or in one of six energy saving modes. Different types of MICA Motes are available: MICA2 (see Figure 1.12) [285], MICAz [287], and MICA2DOT (see Figure 1.13)[286]. The motes differ for example in the communication technology: the MICA2 and the MICA2DOT Motes use the Chipcon CC1000 low power RF module, which can be used at frequencies of 916 MHZ, 868 MHZ, 433 MHz, and 315 MHz. The MICA2 Motes and MICA2DOT Motes use proprietary communication protocols. In contrast, the MICAz Motes implement the IEEE 802.15.4 standard and use ZigBee for communication. The motes also differ in their size and shape: while the MICA2 Motes, and the MICAz Motes look quite alike, the MICA2DOT Motes are much smaller and have the size of a US quarter.

Fig. 1.12. MICA2 Mote. Image source: http://computer.howstuffworks.com/ mote.htm

Fig. 1.13. MICADOT Mote. Image source: http://sensorweb.geoict.net/news. htm

1

Applications of Sensor Networks

17

1.3.2 The Intel Sensor Network Hardware Platform Mote2 The Intel sensor network platform Intel Mote2 [195] (see Figure 1.14)is based on the Intel PXA27x XScale CPU [399]. The PXA27x is a 32 bit CPU with a clock frequency of 104 to 624 MHz. However, only three clock frequencies are used for the Mote2 platform: 320 MHz, 416 MHz, and 520 MHz. The PXA27x uses 265 KByte of SRAM and has 32 KByte instruction cache and 32 KByte data cache. In addition to the internal CPU memory, the Mote2 platform uses 32 MB of external Flash and 32MB of external SDRAM. In summary, the PXA27x is one of the most powerful CPUs used in sensor networks. In addition to a relatively large amount of memory, the PXA27x also has a wide array of interfaces, including an FIR/SIR infrared interface, an AC97 interface, an I 2 C bus interface, an USB host controller, an UART interfaces, a JTAG interface and an LCD driver. The CPU can be in active mode or in one of four energy saving modes. The Intel Mote2 platform uses the Chipcon CC2420 Chip for 802.15.4/ZigBee communication. The Intel Mote2 can be battery powered or powered by the USB-Port.

Fig. 1.14. Intel Mote2. Image source: http://www.ixbt.com/short/images/ news121206_007.jpg

1.3.3 TelosB Sensor Network Hardware Platform The TelosB sensor network platform [359] is the current research version of the MICA Motes. Different from the MICA Motes, the TelosB motes use the

18

H.-J. Hof

TI MSP430 micro processor [299], which was designed especially for ultra low power operation. The MSP430 is a 16-bit RISC micro controller with a von Neuman architecture. It has a clock speed of 8 MHz. The MSP430 of the TelosB motes has 10 KBytes of RAM and up to 1 MB of flash ROM. It has a 16 KBytes EEPROM for non-volatile data storage, e.g. to hold a configuration. The MSP430 has two 16 bit timers, two 12 bit D/A converter, one 12 bit A/D converter and up to 2 UART interfaces, an SPI interface, an I 2 C interface, and 48 I/O pins. The MSP430 can be in active mode, or in one of 5 energy saving modes. Changing from one of the sleep modes to active mode is very fast. The MSP430 is very energy efficient. The TelosB motes use the Chipcon CC2420 chip for 802.15.4/ZigBee communication. It achieves a data rate of up to 250 kbps. Figure 1.15 shows a TelosB mote.

Fig. 1.15. TelosB mote. Image source: http://www.xbow.jp/tpr2400.jpg

1.3.4 Electronic Sensor Board The Electronic Sensor Board hardware platform was developed in the ScatterWeb project [342] at Freie Universitaet Berlin. The Electronic Sensor Board uses the TI MSP430 Micro controller (see Chapter 1.3.3). For communication, the TR1001 low power RF module is used. It operates on 868 MHz. The Electronic Sensor Board platform uses a proprietary communication protocol. Figure 1.16 shows an Electronic Sensor Board. 1.3.5 BTnode The BTnode rev3 sensor network hardware platform [53] was developed at ETH Zürich. The BTnode rev3 uses the Atmel ATmega128L micro processor (see chapter 1.3.1). Two different communication chips are used: the Zeevo ZV4002 communication chip is used for Bluetooth communication and the Chipcon CC1000 chip is used for proprietary communication protocols. The BTnode ref3 has 244 KByte of external RAM. Figure 1.17 shows a BTnode.

1

Applications of Sensor Networks

19

Fig. 1.16. Electronic Sensor Board. Image source: http://www.inf.fu-berlin.de/ inst/ag-tech/scatterweb_net/hardware/esb/overview.htm

1.4 Upcoming Applications Sensor networks allow applications that were not possible in the time before sensor networks. This is especially true for applications that require dense monitoring and analysis of complex phenomena covering a large region and lasting for a long time. Meteorologic research is one of the research fields that deals with complex phenomena. Even today, we still do not know a lot about the climate of earth. Hence, more sensor network applications in meteorology research can be expected for the near future. Sensor networks are also good to advance into areas that have limited accessibility. The oceans are one example. A better knowledge of the oceans will result in a better understanding of the climate because the oceans are an important factor, e.g. for the emergence of hurricanes. However, to explore the oceans with sensor networks, research into underwater communication and sensor design for underwater missions is necessary. In the long run, sensor networks may be of good use in space-related research. For example instead of sending one single sensor system like the Mars Rover to a distant planet, it would perhaps be possible in the future to deploy a sensor network consisting of thousands of nodes. As the system

20

H.-J. Hof

Fig. 1.17. BTnode rev3. Image source: http://www.btnode.ethz.ch/

would consist of a lot of nodes, a total failure of the overall system would be not very likely. Thus, the risk of a failure of a mission would be smaller.

1.5 Chapter Notes There are few textbooks about applications of sensor networks yet. However, the paper [335] and the sensor network reading list [227] give a good overview of current applications.