|A sensor web can be described as a combination of individual sensors which are networked together, collectively taking on the shape of a web. The sensors may be of varying types (ie. temperature, level recorders, snow gauges, weight sensors, visual monitors – almost anything that generates a signal). Those that are land based, are often referred to as terrestrial sensors. However, satellite and airborne imaging sensors can also form part of a network and are simply called airborne sensors.
Most if not all sensors today will be digital devices. Most will be able to provide instantaneous measurements for parameters for which they have been designed and configured for use. Some will be able to store information onboard for later interrogation and download, while others may transmit data instantly through hard-wired or wireless networks. Yet others may not be able to store data results, nor communicate results; in these cases the sensor is monitored immediately.
Sensors are directly related to decision support systems with two primary considerations:
Type of sensor for intended application
Data frequency and timing needs
Sensor type selection is directly related to the phenomenon intended to be measured. Depending on its characteristics and nature, only specific sensor types (i.e. sensitivity) may be useful. Data frequency and timing is a function of telecommunication ability and input storage capabilities. If you need real-time data for your decision support system, then a sensor with telcom capabilities, such as a wireless sensor, may be needed.
Sensor applications take planning. Say you put a GPS collar on a moose or elk in the wild, and a sensor is collecting temperature information on the animal, then you are not likely going to find it immediately – though its position and sensor data are being recorded. How long can you wait, how long do you want to wait? This raises the point of sensing living beings or animals. For long-term studies involving animals, a percentage of them may not live over time, for various reasons. Thus, beginning studies and research on a population must consider the likelihood of having a representative number of subjects in the population at the end of the study.
In the oil and gas industry, Supervisory Control & Data Acquisition System (SCADA) are used to monitor the processes within a total operating environment. For example, sensors are connected throughout an oil refinery or pipeline network, each providing a continuous flow of information and data about the operating processes. But, in this case the system may trigger other events, for example, if a temperature or pressure rises, then a valve will open or close. There are many kinds of SCADA applications, but most of them inter-connect to not only measure and monitor, but to provide automated decision support of ongoing processes.
NASA JPL Volcano Sensorweb defines a sensor web as a networked set of instruments in which information from one or more sensors is automatically used to reconfigure the remainder of the sensors . In this scenario, NASA states, “ Specifically, in our application, we use low resolution, high coverage sensors to trigger observations by high resolution instruments.” For example, Moderate Resolution Imaging Spectrometer (MODIS) flying on Terra and Aqua provides a coarse resolution of 250m. But it might trigger higher resolution satellite sensors, depending upon what it discovers - the same principle as SCADA being applied.
When people use the term ’sensor web’ in terms of connectivity alone, as in connecting a series of sensors, then nothing very innovative is happening. After all, electrical power grids, pipelines and weather networks have been doing this for quite a while now.
We might think of sensor webs as those that are loosely coupled as compared to those that are tightly coupled. Loosely coupled sensor networks simply attach one sensor to another. Each monitors something, but a triggering or decision making logic is not embedded into the process. On the other hand, tightly coupled sensor webs will not only be measuring values, but those measurements will be used for decision making purposes to control other processes.
The tightest couplings will often involve higher levels of logic, requiring higher levels of understanding, support and cost. Earlier we talked about intelligent modeling. Sensors bring not only more data, but more complex types of information into a network, thus sensor webs are fuel to intelligent modeling.
As models become more intelligent using sensor derived data, then the need for more enhanced computing power to analyze and process the information, particularly in decision support systems, where the information is triggering other sensor events, becomes a greater need. This is the connection of sensor web to GRID computing or cyberinfrastructure.
I want to touch on GIS in terms of a sensor web. In my view, a GIS is the integral component (for geospatial people) that enables higher levels of intelligent modeling and can perform the needed spatial calculations that embody a distributed sensor web. A GIS takes simple readings from monitored or connected places and integrates them into a framework for decision making. For this reason, the most adaptable, useful and valuable sensor webs will find a GIS connected to the processing streams and monitoring workflows. It is the decision support system in the framework.
Finally, having worked with sensor networks across many disciplines, my observation is that projects often place more emphasis on connecting the network and using the data (whatever its accuracy), rather than selecting the best sensors and calibrating them and maintaining them properly, so they provide accurate and representative data in the first place. Today, connection is sexy – but when working with sensors for decision making purposes, the knowledgeable will ask, “how did you decide which sensors to use in the first place, and how do you maintain them?”
||An often-quoted Business Week article from 1999 stated that, “In the next century, planet Earth will don an electric skin…” The electric skin that this feature refers to is the concept of the Sensor Web. The sensor web was pioneered by NASA’s Jet Propulsion Laboratory and relates to a network of sensors (pods) distributed throughout the environment that each communicate with one another wirelessly.
Communication is the key element of the sensor web. Through the wireless connection, the individual sensors can be programmed remotely and communicated with individually or collectively. Sensors continually share readings with each other, and a grouping or web of sensors for one location can receive communication from a web of sensors in another location.
The web of sensors becomes an intelligent and adaptive network as data from individuals and groups become fused on the fly. Readings from individual pods detecting anomalous events can trigger adaptive behavior in nearby individual pods or the entire sensor network.
Realizations in Real Time
Sensor webs add real-time data into decision support systems. The fact that the sensors share and fuse information amongst themselves provides pre-validation for data so that it can be trusted and acted upon immediately.
The adaptive and distributed nature of the sensor network ensures that the network will continue to collect and communicate information about a changing environment, regardless of sensor failures at individual locations.
What Types of Sensors?
There’s a broad range of different sensors that can be deployed in a sensor web for various observations and purposes—from environmental monitoring, to hazard detection, to security observation, etc.
Sensors might include temperature, moisture, wind, noise, video, infrared, radio frequency, seismic activity, air quality, chemical and biological, etc. Individual sensor web pods might have specific sensors or a cluster of sensors that each inform one another.
Sensor pods don’t have a determined size, shape or function. An individual sensor pod can be towered clusters of sensors such as border patrol monitoring, networks of unmanned aerial vehicles, and sensor clusters at uniformly distributed monitoring sites. There’s also the idea of biodegradable nano sensors that can be seeded on the wind and distributed widely.
Tried and Tested Sensor Networks
There are a number of sensor web projects that have been deployed successfully.
NASA’s Jet Propulsion Lab has designed and deployed a Volcano Sensor Web to monitor scientific readings and hazard levels at 50 of the Earth’s most active volcanoes.
SensorWare Systems, a spin-off of the NASA’s sensor web project, has deployed a number of networks in a variety of environments. Projects have included agricultural sensors to test irrigation methods, a network in Antarctica that tested the performance of their system in the ultimate of harsh conditions, and a flood monitoring network in the Tucson desert.
Microsoft Research has deployed a SenseWeb project and SensorMap application with designs to map the world in realtime.
The U.S. Army is also heavily interested in sensor webs for their Future Combat System and Objective Force for the Warrior Program. They’ve tested a network to coordinate ground sensors, robotic vehicles and unmanned aerial vehicles for battlefield operations.
Earth Monitoring Advancements
The area that most intrigues me is Earth observation. There’s a research movement afoot to connect earth observing satellite sensors with ground-based sensors for optimal use of resources. The combination of these sensors with adaptive earth system models should provide considerable added insight into our planet’s complex systems.
Sensor networks will continue to advance to provide a significant increase in our knowledge and understanding of our planet’s systems and those system’s interactions.