Progress has been made in recent years in constructing energy-saving buildings. Now researchers have been working on an approach designed to enable buildings not only to adapt heating and lighting settings, etc to both the internal and external environment but also take account of the personal needs and preferences of the occupants.
Is it feasible to create systems capable of taking flexible, real-time decisions on lighting and temperature inside buildings, taking into account a full range of parameters including both external conditions such as sunlight, and the internal environment, plus the known preferences of the people using the various rooms in the building? Swedish researchers at the KTH Royal Institute of Technology and the University of Kalskrona looked into this question and came up with the idea of a multi-agent system (MAS) – i.e. a system composed of multiple interacting intelligent agents – which is capable of adapting and ‘reasoning’ when faced with uncertainty*. This means that the system is able to take rational decisions in real time even where unforeseen circumstances arise. The three scientists report that their system enables a reduction of up to 40% in energy consumption in a given building, while at the same time “increasing customer satisfaction through value-added services.” The MAS consists mainly of sensors, which capture data on temperature and light in the rooms and on the outside walls of the building, coupled with a ‘smart badging’ system, which keeps track of the people circulating in the building, and ‘actuator devices’ that turn heating and lighting up or down.
Intelligent systems plus automated decision support
A large amount of data is entered into the system, such as average body temperature, which helps to work out the number of people in a room. In addition, user preferences – individuals’ desired temperature and preferred level of light intensity, etc. – are also included. This information is then aggregated and the ‘agent’ – the system controlling the parameters of a particular room – can set the controls according to the time of day, the number of people present and their known preferences. Everything is done automatically. The system can also make adjustments in real time. If someone leaves the room, the system registers this and recalibrates for the remaining participants. However, such functionality requires very precise calculations, and the room agents tested by the researchers are not equipped to take decisions ‘on the fly’. They decided to put the decision analysis functionality needed to make real-time adjustments into an automated support tool, external to the MAS agents, which they call a ‘pronouncer’.
Substantial savings, human element remains
Each agent acting autonomously sends its data to the external entity - the ‘pronouncer’ - which will make the best decision according to template models designed in advance. Using the pronouncer pays dividends as it cuts the calculation time substantially. The researchers have run 10,000 set/evaluate tests, with the slowest tool solving the decision problem in hand in less than ten milliseconds, a speed that would be impossible if each ‘agent’ had to take the decision alone using its own calculation module. The researchers reckon that, give the pronouncer’s “short response time, the agents could make extensive use of the decision support provided by a pronouncer without any noticeable degradation of the system.” It is also vital that even when a building has an automated system, “it must always be possible to over-rule the decisions of the MAS agents by physical interaction with the electrical equipment” – i.e. the human occupants can still switch things on and off at need.
*Artificial Decision Making Under Uncertainty in Intelligent Buildings