Advances in Wireless Sensor Networks Antonio G. Ruzzelli* School of Computer Science & Informatics, University College Dublin, Belfield, Dublin 4, Ireland Voice: +353-1-7192488 Fax: +353-1-2697262 Email:
[email protected]
Richard Tynan School of Computer Science & Informatics, University College Dublin, Belfield, Dublin 4, Ireland Voice: +353-1-7162930 Fax: +353-1-2697262 Email:
[email protected]
Michael J. O'Grady School of Computer Science & Informatics, University College Dublin, Belfield, Dublin 4, Ireland Voice: +353-1-7192922 Fax: +353-1-2697262 Email:
[email protected]
Gregory M.P. O'Hare School of Computer Science & Informatics, University College Dublin, Belfield, Dublin 4, Ireland Voice: +353-1-7162472 Fax: +353-1-2697262 Email:
[email protected]
(* Corresponding author)
1
Advances in Wireless Sensor Networks Antonio G. Ruzzelli, University College Dublin, Ireland Richard Tynan, University College Dublin, Ireland
Michael O'Grady, University College Dublin, Ireland Gregory O'Hare, University College Dublin, Ireland
INTRODUCTION The origins of networks of sensors can be traced back to the 1980s when DARPA initiated the Distributed Sensor Networks program. However, recent advances in microprocessor fabrication have led to a dramatic reduction in both the physical size and power consumption of such devices. Battery and sensing technology as well as communications hardware have also followed a similar miniaturization trend. The aggregation of these advances has led to the development of networked, millimeter-scale, sensing devices capable of complex processing tasks. Collectively these form a Wireless Sensor Network (WSN), thus heralding a new era of ubiquitous sensing technology and applications. Large-scale deployments of these networks have been used in many diverse fields such as wildlife habitat monitoring (Mainwaring, 2003,) traffic monitoring (Coleri, 2004) and lighting control (Sandhu, 2004). A number of commercial WSN platforms have been launched in recent years. Examples include the Mica family (Hill, 2003)), Smart-Mesh(Dust Inc. website), Ember (Ember Corporation, 2005), iBeans (Rhee, 2003), Soapbox from VTT (Soapbox website), Smart-Its (SmartIts, 2005), and the Cube sensor platform (O'Flynn, 2005). As the miniaturization of the constituent components of a WSN continues unabated, power consumption likewise diminishes, thus the current generation of sensors can function perfectly for years using standard AA
2
batteries (Polastre, 2004). Alternative solutions may not require any batteries; for example iBeans (Rhee, 2003) coupled with energy harvester can operate by scavenging energy from tiny vibrations that occur naturally. Miniaturized solar panels are another possible solution for outdoor operation. Production costs of single nodes are estimated to be less than a dollar, a significant cost reduction over the price of older sensor models, thus paving the way for largescale WSN deployments, possibly consisting of a number of nodes several orders of magnitude greater than that in ad-hoc networks (Akyildiz, 2002).
BACKGROUND The main components of a WSN are: gateways and sensor nodes. The sensor nodes can relay their sensed data either directly to the gateway or through each other depending on the scale of the network. In turn the gateway can send commands down to the nodes to, for example, increase their sampling frequency. In some networks, when the gateway is tethered to an adequate power supply, a greater transmission range can be achieved. This gives rise to an asymmetry in the data acquisition and control protocols, where control commands are sent directly to the node but the data sent from the node to the gateway is multi-hopped. Of course multi hopping of the control commands from the gateway can be used also. Multi hopping, while useful in extending the reach or scale of a WSN and reducing the overall transmission cost with respect to the direct communication, does have its limitations. The cost of transmitting a packet can be greatly increased depending on the distance a node is from its gateway. Secondly, since nodes nearest the base station, that is, one hop away, will not only have to send their data but also that of all other nodes greater than a single hop, there will be a greater demand placed on the power supply of these nodes. This means that, in general, a node lifespan is inversely proportional to the number of hops it is away from the base station. To
3
alleviate this problem, multiple gateways can be used, with the nodes only transmitting data to their local station. A second solution creates a hierarchy of nodes with varying power and transmission capabilities. Higher power nodes can act as gateways to the gateways for the lower powered sensors.
DEPLOYMENT CONSIDERATIONS There are a number of issues to consider when deploying a WSN. The first, and perhaps most significant, is the number of nodes required in the deployment. The quantity of nodes required will primarily be governed by the size of the area to be monitored and the frequency of the sampling required. In general, the more nodes there are in a given WSN, the better the quality of data; however there is a corresponding increase in the time required for processing. Given that the choice on node density has been made, another factor still remains – the node sampling frequency. The choice of this value will depend on what aspect of the environment is being monitored and the power resources available on each node. If the sampling frequency is too high, more power will be consumed than is necessary. However, if it is too low, important events could be missed. A useful strategy might be to alter this value opportunistically so as to deliver optimum performance. Getting the sensed data from the node requires the wireless transmission of a data packet, a process which can consume a significant portion of the available power resources. One transmission can occur per sensed value when a real-time picture of the environment is required. Or multiple readings can be bundled into a single message. Alternatively the node may intelligently decide not to transmit a packet if, for example, no change in the sensed value had occurred. This can dramatically increase the longevity of the node and, if this is a universal policy adopted by all nodes in the WSN, the lifespan of the WSN can be extended.
4
Due to the battery operation of the nodes, power management is critically important to the health of a WSN. Another approach to performing power conservation is to enable redundant sensors to hibernate. The rationale behind this is that a sensor consumes little or no power while asleep and so the more nodes that are hibernating, the less power being consumed collectively by the network. A hibernating node cannot forward any sensed data, effectively reducing the spatial sampling frequency. Therefore, caution must be exercised when selecting which nodes to hibernate. Two broad approaches exist for selecting nodes for hibernation. The first is based on defining a sensing radius for each individual sensor. An area is covered if all points with the sensed area lie within the sensing range of at least one sensor. When there are points covered by more than one sensor, it may be possible hibernate redundant sensors without breaking the coverage constraint. An alternative approach uses the data being received by nodes. It is based on interpolation and assumes that the required node density exists. Given a collection of nodes, it is possible for them to interpolate the sensed medium at a required point, assuming an interpolation function exists. By interpolating at the point of an individual node, an interpolated value can be obtained. By comparing this to the actual sensed value, an interpolation error can be derived. If this error is less than a particular threshold then this node is deemed redundant and will hibernate for a predefined period of time before rechecking its redundancy. Another fundamental issue for practical WSNs is that of sensor calibration (Whitehouse, 2003). When two nodes observe different values in their sensed data, is it because they are seeing different events or because one or both of the sensors has malfunctioned? Of course calibration can be done prior to deployment, but if the malfunction causes its accuracy to degrade over time
5
then a recalibration must occur on the fly after deployment. This is a significant problem, since the environment in which the nodes are sensing usually cannot be controlled for the calibration to occur. When sampling the environment, it may be a requirement for all the sensors to sample at the same point in time. This requires a clock synchronization technique that will work over the entire network and this is quite a difficult task to perform on such computationally challenged devices. Multi-hop routing can introduce a considerable lag between the time a message is sent from the node and the time it is received at its destination. When the destination is a gateway, it will in turn send control commands to the network based on the data it receives. These control commands may also be multi-hopped to their destination. The aggregate delay can be unacceptable and is usually symptomatic of an overburdened gateway. The introduction of an additional gateway which efficiently partitions the network would alleviate this issue. A final issue with a practical deployment of a WSN is that of programming and debugging the nodes themselves. A node considered to have one of the richest user interfaces consists of three LEDS, allowing eight program execution states to be displayed at any given point in time. To alleviate this, a methodology has been developed to allow the development of applications off the nodes so that accuracy of the approach can be verified (Tynan, 2005).
APPLICATION CONSIDERATIONS The autonomous nature of the wireless sensor networks makes the technology versatile with respect to applications. Sensors can be effectively deployed for monitoring and detecting of malfunctioning industrial machinery during normal production activity. Moreover, as no infrastructure is needed, their deployment is immediate and highly adaptive to the environment
6
in which they operate. By means of distributed sampling, sensor networks are able to provide a more accurate and in depth evaluation of the state of the environment at any moment in time. Sensors can also be programmed to take decisions at local stage or through a centralized approach. In the first case, nodes are organized in clusters in which a sensor head is elected. Collected data is sent to the sensor head, where it is evaluated before an appropriate decision is made. For instance it can actuate supplementary machinery in cases where production overloading is occurring. In cases where a sensor is not endowed with a decision making capability, or in particular circumstances where it does not have enough information to make an immediate decision, collected data is sent to the closest gateway for further processing, and the resulting decision is returned to the sensor. In an energy-saving manoeuvre, data is sent to neighbouring nodes that forwards it, via a routing procedure, towards the gateway. Such a scenario is an illustration of a centralized approach of decision making. For more effective management of the production activity, sensors can interact, possibly using either fixed networked or wireless technology, with Personal Digital Assistants (PDA) or mobile phones. In this way, wireless sensor networks can improve the supervision of an activity and communicate with a user in a more effective and efficient manner. For example, in the case of industrial machine monitoring, should the sensor head sense that a particular machine has excessive vibration or that a temperature exceeds a certain threshold, it may decide to raise an alarm and call a technician for assistance. However, the approach it takes to this may vary. Flagging an alarm in a centralised control system can be implemented quite easily. However, technicians and maintenance staff are generally mobile, as their duties call them to various places in the factory floor. Thus the sensor head must be able to route a message to them. As
7
technicians are likely to be equipped with PDAs or mobile phones, Instant Messaging and SMS are two obvious methods of contacting them. Applications of m-commerce through the use of collaborative wireless sensor networks and PDAs or mobile phones are numerous. In the case of sensors deployed in a shopping centre, a shopper with a PDA or smart-phone can request a particular product with certain requirements, for example model, price and so on. By means of sensor collaboration, products in the shopping centre that match the user requirements can be identified. The shopper can then decide whether to buy remotely or go and view the items in question before buying it. Alternatively, a reverse auction could be initiated with the shops in the mall all vying for the shopper's business. While such a scenario illustrates the potential synergic interaction of WSNs and standard mobile devices, a number of issues must be resolved before such a vision can become a reality.
FUTURE TRENDS Though significant research in WSNs and mobile computing continues, issues concerning the enablement of seamless and transparent interaction between each domain need to be resolved. A number of issues are now identified. Communication protocol issues: In order for a PDA to communicate with a sensor network, it is necessary that both PDAs and WSNs use the same communication protocol. At present, off the shelf PDAs have the Bluetooth protocol for short range communication provided. Unfortunately, studies of the Bluetooth architecture (Leopold, 2003) showed the unsuitability of such a protocol for wireless sensor networks. On the other hand, although recent advances propose a vast number of protocols tailored to WSNs, the communication compatibility between the two technologies is still an open issue.
8
Ontology issues: Such kinds of issues arise after PDAs and sensors agree which communication protocol to use. In the context of knowledge sharing between PDAs and sensors at the application layer, they should agree with the specification of a conceptualization, also known as an ontology. Although some research propose the study of semantic techniques for wireless sensor networks (Whitehouse, 2006), a comprehensive methodology of PDA/sensor interaction is still an open issue to be addressed. Trust management issues: Requests of m-commerce-related information from sensors to PDAs and vice versa raises issues of trust management. In fact, sensors should trust the quality of service offered by the PDA protocol. On the other side, PDAs should trust sensors when, for example, product availability or machinery condition are sent to a PDA. While the latter case can be considered as an instance of internet trust management, the former case needs to consider the issue of memory capability constraints of sensors. Procedures for realizing trust management on individual sensors, for example, through intelligent agent technologies, need further research. The big “umbrella” of trust management also includes more specific issues of security. In fact, the multi-hop routing of WSNs together with the relatively simple architecture of sensors pose an inherent risk, as an attacker may only need to compromise one device to compromise the security of the entire network. This concern is amplified in applications like m-commerce where private credentials must be fully safely encoded.
CONCLUSION Significant advances have been made in WSNs over the last decade. Nevertheless, power consumption remains a significant barrier to their widespread deployment. However, significant
9
strides are being made in this area. Though a mature research discipline in its own right, the issues of interaction and interoperability with conventional computing systems, and mobile computing devices in particular, is becoming increasingly important. In extending the reach of computing technologies into what are frequently remote and hostile environments, WSNs will enable an array of new and innovative applications and services for mobile users.
REFERENCES
Akyildiz, I.F., & Su, W., & Sankarasubramaniam, Y., & Cayirci, E., (2002). “Wireless Sensor Networks: A Survey”, Computer Networks Journal, 38 (4), 393-422. Bychkovskiy, V. & Megerian, S. & Estrin, D. & Potkonjak. M. (2003). “A Collaborative Approach to In-Place Sensor Calibration”, proceeding of 2nd International Workshop on Information Processing in Sensor Networks (IPSN’03), .................... Coleri, S. &. Cheung, Y & Varaiya, P. (2004). “Sensor Networks for Monitoring Traffic”, proceeding of Allerton Conference on Communication, Control and Computing, ........... Dust Inc. Website. http://www.dust-inc.com. Ember Corporation Website. http://www.ember.com. Mainwaring, A. & Polastre, J. & Szewczyk, R. & Culler, D. & Anderson. J., (2002). “Wireless Sensor Networks for Habitat Monitoring”, proceeding of International Workshop on Wireless Sensor Networks and Applications, ........................... O'Flynn, B., & Barroso, A., & Bellis, S., & Benson, J., & Roedig, U., & Delaney, K., & Barton, J., & Sreenan, C., & O'Mathuna C. (2005). “The Development of a Novel Miniaturized Modular Platform for Wireless Sensor Networks”, proceedings of the IPSN Track on
10
Sensor Platform, Tools and Design Methods for Networked Embedded Systems (IPSN2005/SPOTS2005), Los Angeles, USA. Leopold, M., & Dydensborg, M. B., & Bonnet, P. (2003). "Bluetooth and sensor networks: a reality check", proceedings of the First International Conference on Embedded Networked Sensor Systems, Los Angeles, California, USA. Polastre, J., & Hill, J., & Culler, D. (2004). “Versatile low power media access for wireless sensor networks”, proceedings of the 2nd international Conference on Embedded Networked Sensor Systems (SenSys '04), New York, USA, 95-107. Rhee, S. & Seetharam, D. & Liu, S. & and Wang. N. (2003) “iBean Network: An Ultralow PowerWireless Sensor Network”, proceedings of UbiComp, the Fifth International Conference on Ubiquitous Computing, ................. Sandhu, J. & Agogino, A. & and Agogino, A. (2004). “Wireless Sensor Networks for Commercial Lighting Control: Decision Making with Multi-agent Systems”, proceedings of the AAAI-04 Workshop on Sensor Networks, San Jose, CA, USA. Hill, J. (2003). “System Architecture for Wireless Sensor Networks” PhD thesis, UC Berkeley. SmartIts - http://www.smart-its.org. Soapbox - http://www.vtt.fi/ele/research/tel/projects/soapbox.html. Tynan, R. & Ruzzelli, A.G. & O'Hare, G.M.P., (2005). “A Methodology for the Development of Multi-Agent Systems on Wireless Sensor Networks.”, proceedings of 17th International Conference on Software Engineering and Knowledge Engineering (SEKE'05), Taiwan. Whitehouse, K. & and Culler, D. (2003). “Macro-calibration in Sensor/Actuator Networks.” Mobile Networks and Applications Journal (MONET), Special Issue on Wireless Sensor Networks.
11
Whitehouse K., & Liu, J., & Zhao, F. (2006). "Semantic Streams: a Framework for Composable Inference over Sensor Data", proceedings of the European Workshop on Wireless Sensor Networks (EWSN), Zurich, Switzerland.
TERMS AND DEFINITIONS WSNs: Wireless Sensor Networks PDAs: Personal Digital Assistants also known as palmtops. MAS: Multi agent system Gateway: In general, it is considered a more powerful node that is used to collect information from the networks and for some architecture to synchronize them. A WSN can have few gateways deployed in the network. Sometimes, they are assumed to be interconnected through alternative wireless technology like WLAN, WiMAX. Gateways are often referred as sinks or inappropriately called base stations. Node lifespan: Also known as node lifetime, it represents the operational life expectancy of a sensor. Usually, it is calculated for certain network parameters (e.g. network topology, network density) and certain node parameters (e.g. node data rate, duty cycle of a sensor equipped with two AA standard batteries).
12
