iMinds makes the Internet of Things intelligent

Claiming that the IoT raises great expectations is an understatement: it promises to fundamentally disrupt the way we live and work - provided that it has the intelligence to cope with the data tsunami that it will generate.

IoT systems should have the flexibility to migrate intelligence to the best spot — closer to the sensor in some cases or toward the cloud backend in others, in a truly distributed and ever-shifting fog.

Little Sister: an intelligent monitoring framework that uses low-resolution visual sensors

iMinds’ Little Sister applied research project (January 2013 till December 2014) aimed to relieve the pressure from nursing homes’ limited capacity by pursuing the development of an affordable, round-the-clock monitoring solution that can be used in assisted living facilities. The intelligent proof-of-concept that resulted from the project empowers older people to live (semi-)autonomously for a longer period of time, as it alerts the right caregivers when assistance is required. Little Sister uses low-cost, low-resolution visual sensors to do so, monitoring people 24/7 and gathering long-term, objective information about their physical and mental wellbeing. Continue reading

But where should all that intelligence reside?

At one extreme, some suggest all underlying processing and intelligence should live in the cloud. This has the advantage that communication is made relatively easy: everything talks to the cloud. Yet, this strategy also comes with a number of concerns: delays could be introduced, massive network congestion could occur, and data that is transported to and from the cloud can more easily be compromised.

At the other extreme is the suggestion that intelligence should be embedded in each individual device. Yet, many of the devices in the Internet of Things will be significantly resource-constrained. And even as processors continue to miniaturize and advance, and battery lives continue to extend, there is only so much computing a small, single sensor can support. As a result, functions like authentication cannot typically be performed at the device level, because they require too much computing power.

In summary, there is no single solution for determining exactly where intelligence should sit in the Internet of Things, but as a general rule the answer will lie somewhere in the middle of the two extremes — a hybrid scenario in which certain ‘thinking’ happens at the device level while other functions, such as the generation of encrypted keys for secure transactions, will occur in the cloud.

This is the approach advanced by iMinds in its research, with systems having the flexibility to migrate intelligence to the best spot — closer to the sensor in some cases or toward the cloud backend in others, in a truly distributed and ever-shifting fog.

The iMinds approach to making the IoT intelligent

Making the IoT intelligent is a complex task. A good example of the underlying complexities can be found in manufacturing settings – where robots use cameras for wayfinding. These cameras are statically integrated: if a new camera is installed in the factory ceiling, the existing robots won’t use it because they don’t know it’s there and can’t communicate with it. However, if some intelligence is shifted into the factory Internet so that the robots can be alerted once a new camera has been deployed — and then find it, search for the protocols they need to communicate with it, download those protocols and integrate them — the system gains an evolutionary capability that will allow it to continue to make optimal use of deployed resources on a real-time, ongoing basis.

This kind of capability is in fact going to be essential when IoT deployments reach scales at which manually controlling each independent machine becomes prohibitively complex. iMinds researchers have been building expertise in smart manufacturing and robotics control to address precisely these kinds of requirements.

Intelligent weaving machines powered by the Internet of Things

iMinds researchers worked with Picanol, a pioneering manufacturer of weaving machines, to evolve Picanol’s traditional weaving machines into smart, connected devices. “With intelligence embedded in our networked weaving machines, you can monitor metrics like oil temperature, for example. You collect the data, run an algorithm on it once a day, know if one or more machines is exceeding a given threshold and so might need to be serviced. And going a step further, machines could even be shut down remotely if a threshold is exceeded by a certain amount, preventing further damage.” – Matthias Marescaux, Picanol.

Continue reading

FallRisk: an intelligent framework that uses sensor data to assess the fall risk of older adults

Accurate fall detection is a hot topic. The partners in the iMinds FallRisk project therefore investigated how to improve existing fall detection systems and develop an intelligent framework to assess the fall risk of older adults.

The project led to the development of an intelligent backbone that captures and interprets all sensor data, and that intelligently distributes information to formal and informal caregivers. In case of an event, messages are forwarded to that specific caregiver who is best placed to respond – taking into account event context (e.g. day or night; bathroom or kitchen; event classification) and older adults’ social networks (proximity and availability of formal and informal caregivers, circle of trust, etc.).

Continue reading

Big Data or Small Data?

Big Data and the Internet of Things are increasingly becoming two sides of the same coin: the millions and billions of connected devices that make up the IoT generate a tsunami of data – and traditional data processing applications can no longer keep up.

When it comes to Big Data, the challenge consists of picking and analyzing the right data to take decisions, and make sure that the different kinds of data formats are interoperable. But the IoT is not just about Big Data: sometimes you are dealing with just a few readings (Small Data) from a sensor somewhere. In that case you have to be absolutely sure of the quality of your data and use it as wisely as you can. Having the means to deal adequately both with Big and Small Data will be crucial to making the Internet of Things a success.