"For a home to be smart, it must be sensitive"
A smart home, what exactly is it? Julien Cumin belongs to those who are working to address this question. The Orange Labs doctoral student explains, “the smart home is able to offer personalised and adapted services to its occupants“. And to this aim it has to build a more refined knowledge of its occupants, whose everyday activities are a key element, and of their environment, thanks to what is known as “contextual” information.
Information that “strongly characterises the services the occupants of the home will need“, underlines Julien Cumin, whose thesis deals with activity recognition within the home. The objective: to understand the actions taking place there thanks to environmental sensors (electricity consumption, humidity, motion, etc. sensors) that provide very varied raw data.
However, “in order for a home to be smart, it must be sensitive“, Julien Cumin states. Take automatic lighting for example: in the smart home as we imagined it in the past, lighting switches on and off automatically when a person enters and leaves a room. In the sensitive home of tomorrow, if a person leaves a room momentarily to go and get something, the system is capable of understanding this information. Capable of understanding that this occupant intends to come back because they have not finished their activity, thus the system will not switch off the light.
From smart home to sensitive home
So where the smart home of yesterday aimed to offer services based on the detection of events within the home, the smart and sensitive home that Julien Cumin calls for, goes even further. It contextualises these events, translates them into activities, thus using information that bears more meaning for each occupant.
Everyday activities are recognised by machine learning. The team within which Julien Cumin works is teaching a computer to establish a link between the sensors’ events and the associated activities by feeding it with examples: if I am cooking, taking a shower, or watching TV, here is the sensor information that is fed back. “Eventually, the doctoral student explains, we hope that the algorithm will be “smart” enough to recognise the activity taking place when it is confronted with a new event.”
As happy as a doctoral student at Orange
Julien Cumin joined Orange Labs in Meylan (Isère) after an end-of-studies internship in the company. He commends the importance given here to the works of doctoral students. “The worry I had when starting my thesis, was that I would be isolated, he says. I was to interact with my thesis director and my supervisors, and publish papers, but my work would remain quite confidential. In the end, I observe that this is not the case: I take part in many events during which I have the opportunity to present my work and to exchange with lots of people.”
“At Orange, he adds, doctoral students really have the possibility to disseminate their work within the Group, to the scientific community, but also to professionals and companies, which constitutes an extremely interesting opportunity.”
In fact this shows the importance given to research as a whole, on which the young man looks positively: “Firstly it is vary varied and widespread, with a great wealth of research topics and researcher profiles. Secondly, it is strategic for the Group, both because of the large budget invested in it, and because it strongly values it.“