OVERVIEW

CENS Systems and Habitat staff, students, and faculty have collaborated on the design, implementation and ongoing deployment of the Extensible Sensing System (ESS) for microclimate monitoring in support of a wide range of ecophysiology studies.  The ESS offers degrees of scalability and flexibility that push the frontier of currently deployed sensor networks.  The ESS is one of the very first wireless sensor networks to actually be deployed in the scale of hundreds of nodes.

This system has been designed using sensor network staples and commodity components; such as an implementation of Directed Diffusion: called Tiny Diffusion, Mica Motes, Compaq iPAQs, and Sensor Interface Boards.  In addition to the sensor deployments, the ESS incorporates an extensible middleware tier that supports a wide range of industrial data analysis tools; such as a functioning deployment of Oracle 9i, bindings for Matlab, and bindings for LabView. The middleware system is called the Subject Servers, and they are an Internet-based publish-and-subscribe bus.

APPROACHes

The ESS addresses the challenges of microclimate habitat monitoring by providing an extensible end-to-end solution. The ESS is comprised of 2 tiers. The sensing array is composed of Mica2 motes and a data acquisition system. Sensing comprises the first tier of the system. The second tier of the system is a publish/subscribe bus and a set of standard clients and extensible bindings that are designed to streamline data-analysis.

At the first tier, sensor nodes are Mica2 Motes ( http://www.xbow.com ) running TinyOS and equipped with CENS Sensor Interface Boards ( http://www.cens.ucla.edu/~mhr/daq/ ) and microservers that provide resource-rich access points into the sensor network. The sensor nodes form an ad-hoc network using a package called Tiny Diffusion (http://www.cens.ucla.edu/~eoster/tinydiff/). Tiny Diffusion is an implementation of Directed Diffusion (Heideman et al). In the Tiny Diffusion network, sensor nodes act as data sources and provide their data to clients in the network. Microservers act in a few capacities. Microservers participate with sensor nodes by delegating queries and acting as data sinks.

One of the primary facilities provided by the ESS is its flexible query structure. Queries in the ESS offer users the ability to query sets of channels, create conditional queries that relate channels, detect rising edge on sensors, create time-series queries, create aggregate operations, and more. The query system in the ESS is called the Data Service Engine (DSE). (http://lecs.cs.ucla.edu/~tschoell/services/) Conditional queries and event-based queries allow data to be filtered within the network, reducing radio communication and energy costs.

The second tier of the ESS is centered around a publish/subscribe middleware package called the Subject Servers (http://cvs.cens.ucla.edu/sbjsrv/). The duty of the second tier is to provide data analysis tools an easy means for operating on sensor data. The second tier has bindings for Oracle, MySQL, LabView, and MatLab, and it is generally extensible to others.

SYSTEMS / EXPERIMENTS

The ESS deployment at the James Reserve is being done in a systematic fashion. The following is a site-map (see Figure 1) for the ESS deployment.

Figure 1

Within this map clusters will dynamically form themselves, according to the ESS algorithm (see Figure 2), using cluster heads positioned at the specified locations.

Figure 2

Database:
The data from this system is stored in a local Oracle 9i instance with a schema (see Figure 3) that has been designed and normalized so as to provide both researchers with data that aids in their system development, and application scientists with abstractions that represent the higher level semantics they need for their research (in the form of views and processing scripts).

Figure 3

The Subject Servers:
The interconnection system between deployments, users, and the database is a local implementation of a high-performance publish-and-subscribe bus called the Subject Servers (see Figure 4). This system provides a general mechanism for interconnectivity between all of the heterogeneous components of this system. Components subscribe to subjects and publish data to subjects. This is a useful utility for this system as the cardinality between publishers and subscribers tends to require some form of multi-cast logic.

Figure 4

The SensorNet Instrumentation:
Sensor nodes without the ability to attach sensors are useful in computer science research, but are not necessarily as useful for sensor network research at the application level. Therefore, attaching such nodes to sensors plays a significant role in the research conducted by CENS in different areas, specifically in habitat monitoring research. The instrumentation board is meant to attach to a Berkeley Mote. This combination acts as a complete wireless instrumentation system. It has analog input channels, high gain analog input channels, interrupt driven general purpose inputs/outputs, and counter channels. We have used the system with many environmental sensors including wind speed, gust wind, temperature, humidity, pressure, leaf wetness, rain fall, motion, and soil moisture sensors.

Figure 5

In addition to the main objectives of this project we also created a clean way of attaching the sensor network to traditional Data Acquisition Systems. Data Acquisition systems are originally meant to be connected to hardware that is attached directly to them, in order to monitor different analog or digital channels. These systems are usually used in industrial Automation Systems or for Process Monitoring or Control Applications. We have created interfaces that enable a Data Acquisition System be used with a combination of sensor network applications to monitor network information. This enables researchers to use a sensor network exactly the same way they use a traditional measurement systems attached to their local computer.

ACCOMPLISHMENTS This Year

The accomplishments this year have been the roll-out of successive versions of the, up to ESS 1.2.  This system met its targets and now provides a flexible re-taskable user interface and an end-to-end solution for sensor data.

FUTURE DIRECTIONS

Future directions for this system include further deployments in related fields, and increased scale of deployments over time.  In-network processing is a current ability of the ESS, and with the moderately complex queries available in the current ESS, the future versions of the system are expected to evolve higher level collaboration.

Over the coming year the ESS will be used to support long term data collection of microclimate data, to support triggered imaging for the MossCam project, and to support triggered image, infrared and acoustic observations of nesting birds. ESS will serve as our testbed for many forms of nested queries, in network aggregation, and region characterization. We plan to use it to generate a wide range of datasets for the community, as well as a target for our early work in macroprogramming.

Based on our successes and lessons we will be able to serve the very large community world wide with the information needed to replicate our hardware base and use our software and supported algorithms to conduct similar work around the globe.

PEOPLE

Faculty:

Prof. Deborah Estrin

Graduate Students:

Tom Schoellhammer
Mike Wimbrow

Staff:

Eric Osterweil
Mohammad Rahimi