In December 2004, the
OSU DARPA-NEST team
headed by Anish Arora
completed the
first demonstration and experiments of ExScal. This demonstration covered an area 1.3km by 300m with about 1000 sensor nodes and around 200 backbone nodes making it the largest wireless sensor network assembled to date. ExScal's demonstration is also the largest ad hoc 802.11 network thus far created.
It is widely believed that someday there will be sensor network deployments of hundreds of thousands of nodes. The challenges in scaling to networks of this size are quite different than the ones encountered in fielding much smaller networks of dozens or hundreds of nodes. The former subsumes the latter and add a host of new problems. The motivation for the DARPA Extreme Scaling project, code-named "ExScal," was to investigate the challenges in scaling to a network of 10,000 sensor nodes. While 10,000 nodes is still a fraction of "hundreds of thousands," we have encountered many of the basic challenges of extreme scaling that we believe will be encountered with larger numbers of nodes. Consequently, the ExScal project has provided us with a rich set of experiences and has given us a visceral understanding of the "real" problems that are posed by networks of extreme scale.
Figure 1.1: Inside an XSM
Figure 1.2: ExScal Demonstration Topology
Equipment
The Sensor-Actuator Nodes
The sensing and actuating nodes used in the project, called XSMs (Extreme Scale Mote), were designed specifically for this project by The Ohio State University and CrossBow Technology, and manufactured by
CrossBow Technology.
They feature a variety of sensors and actuators including a magnetometer, a microphone, four passive infrared receivers, a photocell, a sounder, and feedback LEDs. The XSMs were installed with a trusted software base, also known as the "factory image". This default software includes the Deluge Network Reprogramming Service and the Sensor Network Management System both of which are provided by the University of California at Berkeley (UCB). They also have simple power management capability. The ExScal application consists of the software components deployed during the on-field operation of the XSMs. The ExScal software provides the motes with reliable communication, routing, localization, fault tolerance, and applications to utilize the onboard sensors. The XSMs comprise Tier 1 of the network;
more information on their design can be found here.
Figure 2.1: A deployed XSM in its usual casing
Figure 2.2: A Stargate circuit board
The Backbone Network Node
The XSMs were organized under a second tier of devices called Extreme Scale Stargates (XSS) running the Intel Stargate platform. ExScal customized the stargates by adding an 802.11b Wireless Networking card with requisite software, an external antenna, a housing for the device, and a battery pack. The stargates were placed strategically in the topology (seen above) such that most motes were able to communicate with a stargate. These,
Tier 2,
nodes ran a controller application that served to orchestrate the localization and reprogramming services at Tier 1. They also facilitated retrieving data from the motes to be analyzed on PC's (Tier 3).
Application
There are many applications for an large scale sensor network. The application ExScal is developing for is the detection & classification of multiple intruder types over an extended perimeter. This would be ideal for protecting an area that is too vast to be patrolled by human guards such as an
oil pipeline or a national border.
To this end, the data retrieved from Tier 1 is used to visualize and track intruders in the sensor network. The visualization system is able to display the motes that detected a target, the target using an icon, the path any target has taken through the network, the topology of the sensor network, and the overall health of the network. This is done using group bassed classification. Teir 3 can identify how many XSMs detect a given object and use this "influence feild" to determine what type of object is being sensed. Below is an example of how different types of intruders appear to a single XSM. The images are frequency spectrums of data collected from the passive infrared receivers. The car's energy is in a higher frequency band than the human's energy.
