A "Hitchhiker's Guide" to Netted Sensors

By Garry Jacyna and L. Danny Tromp

A furious hurricane rips up telephone poles and pulls down relay towers, cutting communications to a devastated community. A barren stretch of the U.S. border, too remote to be effectively patrolled, invites easy entry into the country by a cadre of terrorists. A harried mother, her call to dinner unheeded again, wonders where everyone in the house is hiding.

These are just a few examples of problems whose solutions may emerge from MITRE's Netted Sensors Initiative. To understand how netted sensors might close communication gaps, monitor hard-to-reach places, or even locate kids listening to loud music in the basement, we first must understand what netted sensors are, how they evolved, and how they work.

Once Upon a Time

Many years ago, folks started to use large stand-off (that is, long-distance) sensors for surveillance applications. Information from them was culled in command centers and fused using human operators. This activity slowly evolved to support the notion of sensor networks connected through hub-and-spoke systems to a processing facility where some autonomous fusion operations took place. The sensors were still fairly large and the operations still labor intensive.

A few years ago, it dawned on people that with advances in technology, small sensors could be built with some onboard processing and communications capabilities to support a network of distributed nodes. The applications seemed limitless. Low-cost sensors could reconnect disrupted communications back to a working telephone exchange or a microwave relay tower, tirelessly monitor the most remote stretches of borderland, and keep tabs on all the mundane activities of a busy household. It's fairly obvious this technology can have important ramifications for the way the military and other government agencies address their sensing needs.

What Are Netted Sensors?

The definition of netted sensors depends on who you talk to. Basically they are collections of sensors that can communicate with each other in order to form a network that "collaborates" to solve a problem. The problem could be as simple as turning on a light or as sublime as making sure your meal isn't overcooked while you're stuck in traffic. Netted sensing requires groups of sensors to provide overlapping coverage so they can collaborate to provide the best integrated picture of the environment. Through the power of the network, netted sensors can expedite the development of actionable knowledge—whether it's locating your kid in the basement or keeping track of military operations on a battlefield. The sensor nodes can range from large stand-off sensors to micro-sensor platforms called "motes" or "smart dust."

Let's dig down a little deeper: What are motes? Motes are small devices that support an array of sensing modalities: acoustic, magnetic, seismic, and light sensors. Motes also have on-board processor and wireless capabilities that allow them to connect together into a distributed network, all powered by nothing more than two AA batteries or a single watch battery. These devices can be pretty small—anywhere from the size of a match box to the size of a quarter. There are a number of companies, such as Crossbow, Ember, and Smart Dust, that manufacture these devices. In addition, MITRE has developed a prototype device referred to as the MITRE Mote.

Flat vs. Hierarchical

OK, so how do you wire these devices up? There are two main schools of thought here. One group of people advocates a flat network topology—all sensor nodes are on an equal footing and diffuse information across the grid. Another group, ourselves included, advocates the use of hierarchical layering. In this paradigm, low-cost sensor nodes perform operations suitable to their processing capability while sending information to more capable nodes that can handle the intensive processing operations such as array processing, multi-target tracking, or all-source fusion. The cheap sensors are "up close and personal" to the targets, while the more capable nodes support costly but precise stand-off sensors with a broader view of the target domain. These nodes contain micro-PCs or comparable devices that operate from larger batteries.

This approach, incidentally, has an added advantage that's often overlooked—the approach is backward compatible with legacy systems. If the military wants to enhance the capability of an existing system but does not want to pay the development costs associated with an upgrade, then netted sensors may provide a cost-effective alternative. The legacy system would take the place of the more capable nodes in the network, relegating basic operations to the low-cost sensor nodes.

What would each of these hierarchical layers look like? The lowest layer, referred to as Tier 1, includes sensor nodes like the Mica motes that usually make the first contacts with a target of interest. The nodes in this layer establish an ad hoc network; perform some basic operations such as distributed detection, classification, and tracking; and route their information to higher-tier nodes for additional processing and information fusion. There can be further differentiation within a Tier 1 layer depending on the processing capabilities of the nodes, i.e., Tier 1B nodes are more capable than Tier 1A nodes but reside within the same layer.

The next layer up, referred to as Tier 2, includes nodes comparable in computing power to micro-PCs. This layer is where the more intensive signal and information processing tasks are performed. There may also be higher-level layers that further refine and consolidate information within the network, but most people (including us), focus primarily on the Tier 1 and Tier 2 layers.

Advances in technology have led to the design of tiny sensors that require little power. By tying these unobtrusive sensors together into a netted system, you can monitor a wide area.

Separating Reality from Hype

As with most new technologies, a lot of hype surrounds netted sensors. For example, some folks talk about networking thousands of small sensors together to form a pervasive surveillance grid. It's important to realize that most applications today are quite simple—measuring moisture content in a vineyard, monitoring earthquake damage in a building, or keeping track of where your cattle are going.

What's keeping netted sensor technology from undertaking more complicated tasks, such as in real-world military operations? Very little work in the field focuses on what we believe are the important problems—signal and information processing, resource management, communications and networking, and information management. In order for netted sensor technology to be operationally relevant to the military, these problems need to be addressed. That is where MITRE comes in.

Value and Longevity

MITRE is addressing two questions that lie at the crux of netted sensors: What can you do that has any practical value with these devices, given the limited processing capabilities of a Tier 1 node? And how can they serve effectively over extended periods of time powered only by cheap batteries?

MITRE's pioneering work has proven that Tier 1 nodes can provide a surprising amount of sophisticated processing. We have developed detection and classification algorithms that can be used to process target information, including operations that are used to track people and vehicles, directly at the sensors. We are developing a sensor localization algorithm to help determine where the sensors are at any given time. We are also developing tracking algorithms that can follow single targets at Tier 1 or multiple targets at Tier 2, and we are involved in questions related to information management at Tier 1, such as how to query nodes for information.

All of this raises an important question: Why bother with all this processing at a node when we should be using the power of the network to distribute computing tasks across the nodes? Fortunately, there is a reasonable answer. The power cost associated with communicating is much greater (between 200 to 2000 times) than the power cost associated with on-board computation. This necessarily forces the processing down to the node level. At least for the foreseeable future, until more efficient methods are developed, this will be the case.

Doing Their Duty

The other issue concerns the longevity of sensors in the field. How much battery life can you get out of the sensors? For the sensors to be operationally useful to the military, you probably need a week or more. For special operations applications, this may extend to a month or a year—a tall order for nodes powered by AA or watch batteries!

Getting the most from such small power supplies requires careful resource management. You have to be clever in how you stage your processing operations. It all comes down to the "duty cycle"—the fraction of time the node is turned on. For extended operational deployments, we have determined a one to two percent duty cycle is all that can be tolerated. This means the node is on for only one or two percent of the time—it's usually asleep. Over the past two years, MITRE has had a strong focus on resource management issues for this very reason. We have been able to deploy nodes for a month or more without a battery change. This is possible only by simultaneously dealing with all aspects of the problem, including signal and information processing, communications and networking, and resource management.

Now that you have a better idea of exactly what netted sensors are, read on to discover some of the interesting work being conducted by MITRE's Netted Sensors Initiative.

For more information, please contact Garry Jacyna or L. Danny Tromp using the employee directory.

Solutions That Make a Difference.®
Copyright © 1997-2013, The MITRE Corporation. All rights reserved.
MITRE is a registered trademark of The MITRE Corporation.
Material on this site may be copied and distributed with permission only.

Privacy Policy | Contact Us