A sensor network typically consists of a number of sensor nodes which each acquire signals from a sensor or multiple sensors and a system to transmit and process the data acquired from sensors. The sensor nodes can communicate the data within one another or to a centralized location over a wired or wireless network.
The sensor network can thus be classified as a data acquisition network and data distribution network. The data acquisition system typically consists of sensors and circuitry to handle the real-world information available and the data distribution network involves the communication protocols, network topology, and methodology to transmit and handle the data. The basic network topologies used are star, ring, bus, and mesh, as shown in figure 1.
Figure 1: Various network topologies.
The choice of the sensor network topology depends upon the application and the kind of processing and data handling required. The need for improving connectivity from PCs to the real world is gaining momentum. There are a many sensors and actuators in use, and interconnecting them by integrating the data available is becoming a necessity.
The numbers of nodes in a sensor network continues to increase and wired connectivity is often not an option since sensors must be placed in remote locations. The cost per node is also decreasing, enabling wider reach of sensor nodes. There are also many improvements in low power radio technologies which can be used to design more efficient systems.
Wireless networks also offer better scalability compared to wired networks and deploying a new node in a wireless network is easier. Sensor networks need to balance performance versus the lifetime of the sensor node. Wireless nodes can be configured dynamically to balance this tradeoff, as well as operate autonomously to permit local control of operation and power management. A number of wireless protocols can be considered for sensor networks namely Zigbee, Bluetooth, GSM, Wi-Fi, etc. The choice of wireless protocol depends upon the application needs for the sensor network.
Low power capability
Wireless sensor nodes require very little maintenance and must run for days and sometimes months using the same battery. Thus, low power design is critical for the design of real-world wireless sensor networks, and it is a primary requirement that sensor nodes process and transmit sensor data while consuming very little power.
As sensors in a sensor node typically measure slow varying analog quantities, nodes need only be active for a short duration to transmit data before they go back to sleep. This means that sensor nodes have to have excellent standby current capabilities. Also, most of the data transmission occurs between the sensor nodes to the base station.
Network architecture and communication protocols must exploit this asymmetry of sensor communication from sensor node to base station. Design of a low power sensor is critical. Micro Electro Mechanical Systems (MEMS) based sensors with low power capabilities are also critical. Sensor nodes may operate in an environment of densely distributed nodes from different sources. Sensor nodes may also need to transmit using very low power in noisy environments.
Aggregation of data from sensor nodes
The data from a sensor network must be aggregated and processed in a centralized location. Data handling in a sensor network can be split into data dissemination and data gathering.
Data dissemination is the process by which information is routed in the sensor network. This information could be data acquired from the sensor or requests for data from other sensors. A number of algorithms are available for disseminating data across a sensor network.
Data gathering algorithms maximize the number of communications that happen with a sensor node before the node dies. The trade off in this case is between delay and power consumption. In case of a direct transmission, every node sends collected data directly to a centralized network as in the case of nodes with GSM capabilities. Nodes of a wireless sensor network would have an operating system ported onto it. This enables an easy expansion through the addition of more wireless sensors.
The operating systems for sensor networks resemble embedded operating systems since they are developed keeping an application in mind and are not generic. Also, since the system is built with low-power and low-cost capabilities, most general purpose operating systems must be eliminated. Given that most sensor networks do not require real-time capabilities, a smaller operating system such as TinyOS that has been specifically designed for sensor nodes may be used.
Figure 2 shows a typical implementation of a sensor network using a GSM (Global system for mobile communication) modem. Here all of the sensors communicate their data to a centralized server. The server has control over individual sensor nodes; however, individual sensor nodes cannot communicate between themselves. The server has to be involved for any communication between any two sensor nodes.
Figure 2: Typical implementation of a sensor network.