Here, an intelligent home is considered to be a home (machine) which can make intelligent and autonomous decisions about controlling the environment within a home to the satisfaction of the occupants of that home. This work is not focused on home automation where a user has direct local control over many devices in the home. The intelligent home will however have control over many such devices but the decisions made about how to control these devices (eg when to turn a light on) are taken by the intelligent home itself.
An earlier research project which was carried out between 1985 and 1990 is described here in the Intelligence section. This work was designed to investigate the potential benefits of Machine Intelligence in this environment. Three papers were also produced during the original research, see our papers here.
This resource is being made available to discuss and describe some new and on going work on the intelligent home and also to provide some reference to the many years of experience gained in designing, building and using the previous system.
The New Intelligent Home
When the original research was completed and the results reported, the High Level Processor (HLP) was disconnected. The Autonomic Microprocessor Controller (AMC) was redesigned and extended to include a few of the capabilities from the HLP and then left fully connected to the premises. At the time of writing, that version of the AMC provided an uninterrupted monitoring and control service between 1995 and 2013. The current AMC is therefore the starting place for the new work. As of 2013, a new AMC has been developed and installed and an HLP capable system has been developed and installed which currently provides user interface to the AMC..
The original research focus was on how well the system could deduce how many people were in each location of the premises using only simple movement sensors. The original work showed that learning context related sensor patterns helped to improve the mainly deductive processes used for occupant location.
In this new work, the focus is more likely to be on Occupancy Probability Level (OPL) which will be a value representing the likelihood of any location in the premises being occupied. This is all that the controller needs to know in order to provide effective control. The number of people at each location was a useful research tool but it is very difficult to achieve with simple sensors and it is not actually necessary.
What to Learn
The most obvious thing for the system (the Intelligent Home System (IHS) to learn is its environment. Events are things that happen within the environment and must also be things that the IHS can detect. However, it is not very likely that the same sensors will repeatedly fire in the same order at the same times of day since movement sensors detect movement and not presence. The average sensor activity over a meaningful period of time is more likely to reveal useful information on which to base control decisions. Whether the information which the IHS deduces is a useful thing to remember is not clear at present but the control decisions taken by the IHS should also be remembered.
More recently, sequences of movement sensor activity are being seriously considered as the basis for long term memory. Sequences of activity are likely to repeat within the environment and some may be more common than others within a specific environmental context. Sequences are also potentially very useful to aid prediction of movement patterns and in turn, these predicted patterns could help the IHS to control the home more effectively.
As of 2013 two new systems have been installed which replace the older and very reliable AMC. The new AMC is fully focused on the environment but also handles communication with the higher level controller. The new HLP is simply a user interface to the AMC at present but it contains the capability through software development, to support the HLP functionality described.