lab personnel
private area
project description
proof of concept
work plan
broader impacts

proof of concept

The challenge of wirelessly networking a large number of sensors stems from two facts.

Field Tests
To date we have demonstrated reliable intersensor wireless communication in harsh indoor environments.

Related Efforts
A number of companies (Crossbow, Microstrain, Millenial Net) offer wireless network technology that could conceivably be used in the environmental sensing application.

We have identified a number of unique goals for a wireless sensor network for environmental and ecosystem monitoring (West et al. 2001). Scientific data acquisition leads to a unique set of requirements. Our solution meets the following specific objectives:

• Minimal invasiveness and minimum total deployment cost
• Active, aggressive management of energy consumption for long field life
• Accuracy sufficient for research-quality scientific data acquisition
• Support of broad spectrum of sensing technologies
• Networking design that supports scalability in network size and spatial density
• Support of connection to the internet via long-haul links (satellite or terrestrial).

Our sensor architecture is designed to exploit currently available low-cost hardware, and to evolve as more capable and lower-cost technology becomes available. Each sensing unit, which we call a WISARD (wireless sensing and relay device), employs a modular design that consists of three subsystems: a suite of sensors, a microcontroller, and a radio-frequency (RF) transmitter/receiver. The modularity allows us to integrate different types and numbers of sensors, including new technologies as they become available.

Hardware/Software - Our design employs a two commodity 8-bit microcontrollers that provide hardware support to maximize energy efficiency, including measurement of battery status and the ability to dynamically power-down selected sections of the entire system.

The WISARD hardware supports multichannel temperature sensing via a multiplexed thermocouple amplifier and a four-channel transconductance amplifier for interfacing with photodiode-based light sensors. A general purpose analog input is included that may be used to measure the output of a variety of sensors. The digital peripherals section contains additional system support functions, including a real-time clock and a single-wire bi-directional synchronous serial port that supports expansion to measure wind velocity and direction, rainfall intensity, and numerous other analog sensors.

Our software is driven by a simple executive---essentially a timer-driven dispatcher---rather than a down-sized embedded operating system (as in, e.g., Cerpa et al. 2001). We agree with (Hill et al. 2000) that current real-time operating systems are not suited for the wireless sensor network application. This again reflects the environmental monitoring application: we can use a model of periodically-scheduled single-threaded tasks rather than an event-driven multitasking model. The savings in energy consumption are considerable, since a context switch in a typical embedded operating system can take hundreds of instruction cycles.

Many sensor network (“distributed instrument”) deployments will be remote or otherwise require long-term unattended operation. To maximize battery life, our software-hardware design employs aggressive dynamic energy management: all subsystems can be powered down independently, and the microcontroller is placed in a sleep mode whenever possible, reducing power consumption to ~50µW. These measures, in concert with networking innovations (described below), imply battery lifetimes on the order of months, depending on the rate at which environmental variables are sampled.

Efficient wireless sensor networks require an integrated circuit, system, and network design (West et al. 2001). Our design addresses the aggregation problem in networks with dominant information flows: if all information must ultimately arrive at one (or a small number of) gateway nodes, then the nodes closer to the gateways must handle more traffic, thus causing depletion of their energy at higher rates (Singh et al. 1998, Toh 2001). However, the spatio-temporal correlation of environmental data within a study site suggests mitigation of the energy depletion using distributed source coding (data compression) algorithms (Flikkema et al. 2002, Pradhan and Ramchandran 2000).

The radio subsystem utilizes a single-chip RF transceiver operating in the 902-928 MHz ISM (Industrial, Scientific, and Medical) frequency band. Like the microcontroller chips, these radios are produced in high volume for many other applications, minimizing cost.
Northern Arizona University EnGGen Homepage