[ Page 4 of 7 ]

By Barbara Knecht

Getting smarter

Oatfield Estates Extended Family Residences in the Portland, Oregon, suburb of Milwaukie is a retirement home proving the use of RFID technology. The system is not as intelligent now as it will be, but it already allows people with cognitive disabilities to live much more independently. Bill Reed and Lydia Lundberg are the founders of Elite-Care residential care facilities (www.elite-care.com) and the owners and developers of Oatfield, which has just opened its sixth 15-person building.

Reed explains some of the system functioning: “Each ‘house’ is hardwired into the central computer, which is a programmable logic controller (PLC), a microprocessor used to automate machinery. We don’t have smart gadgets, we have a smart controller. For example, let’s say ‘Joe’ goes into the kitchen. The location-tracking technology is constantly relaying where Joe is, and the motion sensors provide the computer with the information that he is there by himself. The computer already knows that if Joe is alone in the kitchen, it needs to turn off the power to the microwave because Joe is unable to remember that cans should not be put into the microwave.” The central controller must be programmed with the facts about Joe’s propensity to put cans in the microwave. However, as the system becomes more robust, it will be able to learn resident habits.

 

The control systems in Michael Mozer’s residence (above and below) are based on neural-network reinforcement learning and prediction techniques. Seventy-five sensors provide information about the environmental conditions being monitored—temperature, ambient light levels, sound, motion, door and window openings—and actuators control the furnace, space heaters, water heater, lighting units, and ceiling fans.. Photography and diagram: Courtesy Michael Mozery

Michael Mozer, professor of Computer Science and Cognitive Science at the University of Colorado in Boulder, outfitted his own home, a 19th-century schoolhouse, with 75 sensors to detect motion, light, sound, temperature, and window and door openings, plus actuators to control heating, lighting, fans, and water heating. The system he and his research students designed controls basic residential comfort systems: HVAC, lighting, and water heater. The research question he set out to answer in the Adaptive House (www.cs.colorado.edu/~mozer/house/) was whether there are significant regularities in a person’s behavior that a system can exploit to make the occupant’s life more comfortable.

“In most research settings, there is little concern about the invasive nature of the devices. But because I live here, this issue had high priority. On the other hand, we had to do this with readily available materials that were much cruder just six or seven years ago,” explained Mozer. “You could do an even less obtrusive installation now, and have much more sensitive and sophisticated sensors.”

What makes his system intelligent is that it can learn and adjust actions accordingly. “Over time, the system learns more about you and can pick up subtle patterns in your behavior. The more variables you give the system, the smarter it becomes. It learned that when I come home late at night, I usually stay at home longer the next morning, so on those days, it sets the thermostat to keep the house warm longer in the morning. It learned that I sit at a particular table in the great room to read. After a few times of manually turning the light up when I sat down there, it was trained to do it automatically when the sensors picked up that I was at the table.”