Wireless Sensor Networks

Recent advances in the microsensor and semiconductor technology have opened a new field for computer science. The electronic miniaturization and the advances in the semiconductor manufacturing process enable for low-power and low-cost hardware. Small and smart devices equipped with a processing unit, storage capacity, and small radios for a wireless communication provide new application opportunities. Augmented with different kinds of sensors, e.g., for temperature, pressure, and humidity measurements, noise and movement detection, or lighting conditions, physical phenomena can be observed by deploying such sensor devices close to them.

The dense deployment of hundreds or even thousands of sensor nodes offers a wide range of new applications. Current and future application areas include habitat and environment monitoring, disaster control and operation, military and intelligence application, object tracking, video surveillance, traffic control, industrial surveillance and automation, as well as health care and home automation. Thus in the future, environment sensing will become more and more ubiquitous and part of our life. For example, modern vehicles already include several sensors connected to electronic systems in order to improve comfort and safety. Although those sensors are likely be wired within a vehicle, they can also form a wireless sensor network (WSN) if each vehicle is considered as a sensor on a larger scale.

Realizing such different kinds of WSNs is a great challenge, from an engineering as well as a researching perspective. Since the WSN as a whole serves a common application task, an individual node becomes less important. Thus even if some nodes fail, the application task must still be satisfiable. Another key challenge a WSN has to deal with is energy efficiency, as most sensor nodes may be battery-powered. In most cases, it may not be possible to change or recharge batteries, either due to the low-cost hardware being used or due to an inaccessible area the nodes are deployed in. Even if solar cells are used, the available energy will be limited. To prolong the overall network operational lifetime, the energy consumption of a sensor node should thus be minimized as far as possible. Most of the node's components will therefore be turned off most of the time and only be used if they are required. For example, the processing unit can be put into a low-power sleep mode while it is idle. Moreover, turning off the wireless communication radio will save the most energy since transmitting 1 bit consumes as much energy as about 1,000 processing instruction. Communicating in a WSN is thus the main energy consumer and deserves particular consideration.

In order to fulfill these requirements, the used protocols and algorithms should be energy-efficient. Otherwise, an early failure of nodes due to energy reasons might require a reconfiguration or even cause significant malfunctioning within the network layer. Researchers are therefore currently focusing on power-aware protocols concerning different network. At the physical layer, information intended to be sent to other nodes must be modulated onto the radio's frequency efficiently. The influence of disruptions on the channel can be minimized, if forward error correction (FEC) codes are used. Bit errors within a received packet can then be corrected, avoiding a retransmission of the entire packet by the sender. However, concerning energy-efficiency, there is a trade-off between FEC and automatic repeat request (ARQ) since error correction requires redundancy which must be added to a packet. In case the content contains no error after the transmission, the redundancy included is useless but causes additional energy consumption at the sending as well as the receiving side. The data link layer is responsible for an energy-efficient communication between two nodes, which is mainly performed by the MAC. Due to the shared medium, packets from different senders may collide if they are transmitted at the same time. The MAC layer tries to minimize such collisions and schedules data transmission periods and sleeping periods. During the sleeping period, the communication radio of a node is turned off completely, which saves a significant amount of energy. Data can thus only be sent during the transmission period. The network layer then takes care of routing the data through the network. Therefore, forwarding path must be established. By exchanging information, the nodes in the network are able to discover other nodes in their neighborhood and learn about already established forwarding paths. In general, the communication in a WSN will take place along a tree rooted at a sink node since this will be the likeliest communication scenario; either sensed data is sent towards the sink node or the sink node queries sensor nodes for information. Finally, the transport layer takes care of the end-to-end data flow and may provide a reliable transmission service for the application layer. It may also take care of network congestion.

Any protocol used in a WSN should therefore be power-aware and be designed in respect of energy efficiency. In contrast to traditional networks, the forwarding of data packet is not address-based but data-centric. That is, sensor nodes are not addressed by a globally unique identifier but rather based on data attributes. For example, instead of requesting the temperature value of an individual node, an application may be more interested if there are any nodes which detects a temperature above a given threshold or what is the average temperature in a specific region. This paradigm shift also implies that a WSN will likely be tailored for the sensing and application task. Furthermore, application information can be exploited during the forwarding process of data. Application-specific data forwarding may reduce the number of information that need to be transmitted significantly. By in-network processing and data aggregation, redundant and useless data can already be filtered out along the forwarding path. For example, consider the scenario where a sink node is interested in the average temperature each node measures over a certain period of time. Rather than forwarding the data readings individually, they can be aggregated by intermediate nodes along the path without any loss of information by only transmitting the sum and the number of readings.

In conclusion, the key requirements of a WSN can be summarized as follows:

  1. Scalability: A WSN can consists of thousands of sensor nodes, densely deployed in an regional area. Protocols must thus scale well with the number of nodes. This is often achieved by using distributed and localized algorithms, where sensor nodes only communicate with nodes in their neighborhood. Centralized approaches are not applicable, especially because of the single point of failure problem.
  2. Adaptiveness: All protocols should be able to adapt to changes in the environment, e.g., changes concerning the connectivity or changes concerning the sensing of physical phenomena.
  3. Resistance to failures: Due to the low-cost hardware or outside influences, sensor nodes are prone to failure. However, achieving the common application task should not be affected. In case of node failures, the network must be able to re-organize itself and if needed change assigned application tasks.
  4. Self-organization: Since a WSN is usually deployed in an unattended area, the network must operate without the need of manual configuration. For example, communication paths throughout the network should be established automatically. Also the cooperation between nodes must be organized in an unattended manner in order to achieve the global application task.
  5. Energy efficiency: As most sensor nodes will be restricted concerning their energy capacity, all protocols and algorithms must be energy-efficient and save as much energy as possible. Since the most energy is consumed during the wireless communication, the radio must be turned off most of the time. But also the transmission of data should be energy-efficient in order to minimize the number of sent and received packets.
  6. Simplicity: Besides their energy capacity, sensor nodes are also limited in their processing and storage capabilities. Thus, algorithms should be as simple as possible in order to minimize their computational complexity and memory usage.