When it comes to technology-based “smart homes,” sometimes humans just won’t cooperate.

“People don’t readily accept having cameras in their homes or being asked to carry tracking devices,” said Aaron Crandall, a doctoral student in electrical engineering and computer science. “In real life situations — especially care facilities — people often forget their device or don’t want to carry it.”
Such resistance has become common enough in smart-home development that Crandall — under the direction of professor and adviser Diane Cook — has turned his efforts toward a system based entirely on simple sensors embedded in the home itself.
The result is the CASAS project — a comprehensive smart-home research testbed located at WSU Pullman. This unique project combines computer learning algorithms with off-the-shelf sensors and controllers to perform smart-home interventions for those with cognitive or physical limitations. (CASAS = Collaborative Advanced Studies on Adaptive Systems.)
In other words, CASAS uses sensors to detect the activity and behavior of individuals, which are then analyzed by artificial-intelligence software. Over time, the data provides predictive patterns that can be used to prompt elderly or disabled residents in their day-to-day tasks.
The system also may control devices in the environment — such as turning off a stove that is left on too long or delivering a reminder to take daily medications. If changes in behavior patterns are detected, the system could alert family members or primary-care providers.

World populations are aging
Such in-home monitoring will become more important as worldwide
populations continue to age. It is predicted that by the year 2030, nearly 20 percent of the American population will be 65 and over.
In Washington state, more than 800,000 individuals live in nursing homes. Cook — the Huie-Rogers chair and professor in electrical engineering and computer science — speculates that if one percent of these people could remain in their own homes for an additional two years, the result would be a savings of $9.4 million per day or $3.4 billion per year for the state.
Crandall calls this “aging in place.”
“The goal is to keep people in a familiar, safe environment while still providing them cost-effective and high-quality medical care,” he said. His first steps toward that reality began in spring 2007, when he met with Cook to discuss building a sensor-based smart home. Having previous experience in this area, she talked about what sensors and hardware they might need — and then basically “told us to go build it,” said Crandall.
With the help of artificial intelligence lab administrator Jim Kusznir, undergraduate student Brian Thomas and housing services, Crandall began outfitting a campus apartment with an array of sensors purchased from a local hardware store. Hundreds of the small devices were wired to a central computer where input can be collected and analyzed.
Today, motion detectors on the ceiling follow residents’ movements from room to room. Sensors gather data on hot and cold water use. Contact switches record when soup or pills in a cupboard are moved. All lights as well as refrigerator doors and stove burners are monitored for use.

                                                                        First to monitor multiples

The initial phase of Crandall’s project was concluded in July, when he demonstrated that the system can successfully monitor a single individual and predict behavior. The next phase will entail the much more difficult job of monitoring multiple people in the test home. Not only must the software be programmed to recognize and separate each person’s activities, but it eventually must monitor their health status at the same time.
Not an easy task. In fact, Crandall is believed to be the first person in the world to successfully run such studies using only simple sensor input and no tracking devices. He also is doing it on a more comprehensive scale than most other research efforts around the globe.
“Having a sensor-filled testbed to experiment with algorithms for health monitoring and intervention is very rare,” said Cook. “Not only does it require expertise in sensors, networking and middleware, but also the artificial-intelligence capacity to develop algorithms for resident profiling and activity monitoring.”
The WSU CASAS apartment is one of only 10 comprehensive smart-home testbeds in the world.

Commercialization soon
Clearly excited about the research, Crandall and Cook’s other graduate students have made significant progress on the development of smart-home environments. In fact, parts of the technology are ready for commercialization, Cook said.
“I think we will see a lot more of it in people’s homes within the next five to 10 years,” she said.