SLIP Individual Report

21st January 2005

Introduction

In the System Level Integration Practical students are given the opportunity to implement an application of their choice on an embedded hardware platform. For this year the target platform was the ProSpeckz II [1] device, designed at Edinburgh for the Speckled Computing project.

The ProSpeckz II is intended as a prototyping and demonstration platform for "specknet" applications, running on networks of small devices ("specks") equipped with sensors, wireless communications and limited processing and storage facilities. At its core is a Cypress CY8C27643 [2] programmable system-on-chip (PSoC) device which includes an 8-bit CPU, 256 bytes RAM, 16K flash ROM and a number of programmable analogue and digital peripheral blocks which can perform a wide range of functions. Radio communications are provided by a Chipcon CC2420 [3] 2.4GHz 250kbps transceiver, compatible with the IEEE 802.15.4 [6] standard.

Each team was initially provided with three ProSpeckz devices, with more becoming available later in the project. The intention was to produce applications which would potentially scale to - and take advantage of - large numbers of Specks.

Choice of application

We realised that the key qualities of specks lay in the ease with which they could be attached to existing objects due to their small size, the potential for mobility afforded by wireless communications, and the additional capabilities of proximity sensing using the radios. With these factors in mind we set out to consider what moving objects we could usefully fit specks to.

Inanimate objects which are frequently moved around (e.g. items of equipment within a building) were one option - though for the tracking purposes these typically require, dumb RFID tags would suffice. Vehicles were perhaps another possibility, but these already have plentiful electronics and no shortage of space or power for additional equipment.

People were the next obvious choice; a speck can easily be attached to or carried by a person, presenting numerous opportunities for useful behaviour. This was the direction explored by the other two groups in the Disaster Area Network and Building Services applications. We felt however that most of the possibilities in this area might be better served by existing hardware such as mobile phones or PDAs, and the necessity to add a user interface of some form negates the benefit of a speck's tiny size.

The idea of attaching specks to animals seemed a far more interesting prospect. Animals are mobile and frequently come into contact with each other, and the patterns of this behaviour provide valuable information in the fields of biology, comparative psychology and ecology. Additionally the specks could be equipped with sensors to record data such as temperature.

Animal behaviour monitoring

We selected as our application a system for monitoring the behaviour of animals, using specks attached to individuals and distributed around a habitat. We realised that by transferring data collaboratively between specks to a small number of fixed listening stations we could collect information from a large and/or widely scattered animal population with far greater ease and less habitat disturbance than traditional tagging techniques.

This approach had been pioneered in the ZebraNet [8] project at Princeton University, in which GPS-equipped collar devices to track herds of wild zebras. The ZebraNet hardware [9] however is comparatively large and heavy at 500g, mainly due to batteries and solar cells to support its power requirements. Although an extremely efficient device (it was a contest winner in the 2003 International Symposium on Low Power Electronics and Design) the use of GPS does incur significant power usage and as such this design could not be used on much smaller animals.

There are other ways to gain some of the insights provided by GPS data, however. In particular, for analysis of social behaviour absolute location information is not necessarily required - proximity information gathered by radio rangefinding can equally show which animals spend time close to each other, revealing social groupings and other patterns, and we exploit this in our design.

Another benefit of GPS data is the ability to observe when animals move - this too though can be determined by simpler means. By fitting an accelerometer and counting oscillations caused by walking and other motions, we can monitor periods of both movement and of other activity such as mating or fighting. To provide further clues to behaviour we also include sensors for monitoring body temperature of the animals.

Even information about movements between specific areas can be gathered without GPS if base stations are deployed at key locations such as feeding/watering grounds, nesting areas etc. Using commonly visited areas for this purpose will also increase the efficiency of data gathering.

We set out to demonstrate that by using these techniques, highly resource limited devices could be used effectively to monitor animal behaviour.