Rural Remote Control: Small, Distributed Computer Devices Are Being Developed to Act as Embedded, Self-Learning Environmental 'Agents' on Rural Properties, Remotely Reporting and Managing Complex Agricultural Processes, Monitoring the Behaviour, Health and Productivity of Stock, and Optimising Environmental Conditions. Farming, as We Know It, Could Be Revolutionised

Article excerpt

Making a living from natural resources in the driest continent on earth is difficult. Doing it in a socially, economically and environmentally sustainable way is even harder. Landowners, natural resource managers and communities have to juggle many conflicting demands and constraints while working hard to run successful operations. Information overload is a real problem--many agricultural and environmental systems are just too complex for humans to understand holistically, let alone manage. Even so, an intelligent solution may be just around the corner. In fact, it's already in early, promising, development.

CSIRO scientists and engineers from what's called the Smart Spaces initiative are suggesting that a new breed of computers may be u p to the task. Ultimately, it's a project that seeks to understand how to combine diverse information on livestock physiology and movements, local climate and landscape function, to optimise rural district productivity and environmental quality.

Smart Spaces is a collaborative initiative involving five CSIRO divisions, the Australian National University (ANU), private landowners and multinational companies. It brings together researchers with interests ranging from information systems architecture to signal measurement and telemetry, with a unique focus on an environmental demonstration project, called SmartLands.

'Rural and environmental systems are complicated in terms of the vast numbers of interacting processes and volumes of information tied up in them, and complex in terms of the unexpected and emergent interdependencies in the systems,' says CSIRO Land and "Water scientist and SmartLands Project Leader, Mike Trefry. Conventional computer-aided resource management systems can only use relatively sparse field data in decision making for planning and automatic control. This works well for aspects of the rural environment that are essentially predictable, but the failures can be spectacular where decisions are based on inadequate information.

As Geoff James, CSIRO Smart Spaces lead scientist, puts it 'The weakness is in the architecture of the approach--trying to specify beforehand all the information required to make a complex decision. Humans are limited in their choices of data sources, and limited in their understandings of data correlations.' Instead, Geoff and his team are trying to shift the burden of analysis from a relatively small number of humans to a vast array of new, cheap, and ultimately disposable, computers, called 'agents' that collectively learn the best management responses.

Intelligent agents

On the lowest level, a computational agent is an artificial device that is able to sense, think (i.e. perform logic), act and communicate. There is, therefore, the possibility of both physical (hardware) and virtual (software) agents. The Smart Spaces vision involves vast numbers of cheap agents deployed throughout a space, sensing variables, analysing data, performing actions and communicating with each other.

The vision has been solidified with the increasing availability of tiny but powerful sensing and communicating technologies, ranging from the new, cheap millimetre-sized computers called 'Motes' developed by the University of California, to mass market handhelds. These physical agents incorporate hardware with publicly available standards for computation and wireless communication, and have the capacity to self-organise into communicating networks.

Rural and environmental data sources

SmartLands aims to test these new agent technologies on an operating livestock and pasture farm near Binalong, New South Wales, over the next two to three years. As a first step in developing the site network, fixed agents will be deployed to measure climatic variables, surface and ground water, and soil moisture. Mobile agents will be mounted on individual animals to measure location and direct physiological data such as body temperature, feeding characteristics or rumen properties. …