Cutting edge: Time to sharpen up those bricks-and-mortar brains

Essex researchers are determined to make buildings more intelligent - but watch out, they could make your computer disappear. Sue Sharples explains

Intelligent buildings" have existed for years - haven't they? While the integration of systems such as heating, ventilation and lighting has undoubtedly increased energy-efficiency and security within buildings, the promise of human-like intelligence has not yet been fulfiled. Even today, the majority of systems involve only simple pre-programmed control of devices, giving minimal flexibility and little actual intelligence. At Essex University, we have set out to change that.

Our research is centred on the vision of a world composed of billions of networked devices, ranging from those that are electronic (such as household appliances or mobile phones) to those with no obvious electronics (such as clothing or paint pigment). This vision, which involves the "disappearance" of the computer into the fabric of our everyday life, is the subject of massive investment by governments and industry; it incorporates many areas of computer science, ranging from hardware and device-design to programming and communications.

Intelligent buildings involve a vast infrastructure of networking, sensors and devices. Consequently a wealth of data is available - such as temperature levels, rooms occupied or the status of applications. This is a rich source from which one can learn about the occupants of the building - their behaviour, habits, lifestyle and preferences. Rather than having humans interpret and analyse this, a more interesting solution is to use artificial intelligence (AI) - computers that are programmed to "think" like us, and are capable of reasoning, learning and adapting to new circumstances.

At Essex, in the intelligent inhabited environments group, we are developing a learning algorithm specifically suited to this: evidential learning. It is a solution that unites three fields of AI.

The first is classical case-based learning, in which scenarios are presented to the computer and it gradually builds up a database of required responses.

The second is fuzzy logic - a means of describing the world in imprecise, human-like expressions, such as "the room is warm" rather than "the room is at 20C".

The third is behaviour-based control, taken from the field of mobile robotics: this is an area in which software is required to give quick, online responses as the robots continually react to an unpredictable and rapidly changing environment.

By combining aspects of these three fields, we have a system that can react quickly to immediate circumstances but that is also continually learning from the building's occupants and adapts its own behaviour and responses over time as necessary.

Behaviour-based design generally relies on intelligence being distributed around the system rather than contained in one core central computer. With the increasing miniaturisation of hardware, tiny processors called agents, each containing some embedded intelligence, can be incorporated easily and inexpensively within the physical infrastructure of a networked environment. The system as a whole continually "monitors" the behaviour of inhabitants, through sensors and devices on the network, and over time can build up patterns, or evidence, of behaviour that is repeated or occurs regularly. Over time, it can then perform these actions itself.

The system contains some fixed behaviours for safety and efficiency, such as how to respond to a fire; but the majority of its behaviour is learnt from the occupants. It is unique in that it will continue to adapt to suit the occupants even if they or their preferences change - without any need for the system to be reprogrammed.

We have already implemented several prototype environments - the most successful being the iDorm, which is intended as an example of student accommodation. We have been particularly keen to expand on the use of this technology to help vulnerable members of society, such as those with physical disabilities or learning difficulties.

Systems such as this can provide simple applications, from a carer or relative being able to "log in" to the home of an ill person to see if they need help through to a complex, fully supportive habitat for an elderly or severely disabled person.

In time, the technology could possibly replace the need for some 24-hour care, or at least remove the burden of mundane tasks from carers, allowing them to spend more quality time with the recipients, while giving the recipients greater independence.

The benefits are not limited to the care sector. Many environments, from domestic to industrial offices, factories, hotels and leisure centres, could be enhanced. And it is not just applicable to buildings, either: this type of system might well be used in transport (we already see cars with on-board computers controlling environmental settings). In fact, it seems possible that almost every area of our lives may one day benefit from this revolutionary technology.

Sue Sharples is a researcher at the department of computer science at Essex University.