In the future, smart homes will ensure more security and convenience with lower energy costs. But they will also satisfy the needs of the residents without being asked.
Analysts have been dreaming of an enormous boom in smart homes for years. According to the Association of the German Internet Industry (eco), in Germany the number of high-tech homes will increase fourfold from the current two million to about eight million by 2022. And Gartner predicts that by 2020 more than half (12.8 billion) of the 20.4 billion networked things will be performing their duties in households.
However, according to a Deloitte study it is still too soon to talk of a boom. Even though smart home devices such as speakers, lights, and thermostats have become more popular by 50 to 67 percent over the past three years.
But the “intelligent” switches, sockets, lights, and even the smart speakers from Amazon and Google aren’t really all that smart. After all, users have to get them “running” with commands or coding. Smart promises will not be fulfilled until apartments and houses guess the needs of the residents and ensure that these are satisfied autonomously.
Smart Home 2.0
The startup Nest already took the first steps in this direction in 2011 with thermostats that learned from manual operation and subsequently made the appropriate settings. In the meantime, Nest has been bought over by Google and continues to develop the Learning Thermostat into the control center for the smart home. Recently, the company extended its portfolio with the Nest Cam IQ Outdoor, a weatherproof camera with face recognition, also for the German market. The fact that the data end up in a Google cloud is a concern for many users.
The original aim of scientists from Carnegie Mellon University was to replace the unpopular surveillance cameras in existing smart home hubs. With very little effort, their sensor prototype converts all rooms in a house or apartment into a true smart home. The “synthetic sensors” record all the ambient data that is needed to bring “smartness” into almost every household appliance.
So that the module with its ten sensors can become the eyes and ears of the room, machine learning algorithms process sounds, humidity, electromagnetic interference, movement, and light. With the data from various sensors, the artificial intelligence is trained to recognize different signatures simultaneously. The system also learns continuously in real-life operation.
House with a brain
IBM and HUF are going one step further with the world’s first house that has self-learning capabilities. It is located in Hartenfels show house village and gets to know its residents via their interactions. To do this, the individual systems continuously record data about the daily habits of the residents with networked sensors.
IBM’s Watson AI acts as a “data understander”. Residents are not passive spectators in their own four walls. Rather, at the beginning they must check and confirm the settings in the system. Based on deep learning, Watson then learns their habits in IBM’s cloud-based IoT platform. This dynamic approach differs from the smart home programming currently available on the market, which carries out pure command chains and maps a static set of rules.
Open interfaces and a shared protocol standard with the building automation are needed so that the data about the preferred room temperature, the ideal light settings, and the time when the TV should switch on finds its way into the system. They ensure that the networked heating system, the motor of the window blind, or the IoT-capable TV can communicate with each other and that the recorded data come together on a shared platform. According to the EU security guidelines, personal data remains the property of its owner and is used only anonymously to train the AI system.
Association of the German Internet Industry (eco) and Arthur D. Little “The German Smart Home Market 2017–2022. Facts and figures”
A free management summary on request. Non-members can purchase the complete study for EUR 149.