- Business areas
- Application areas
- Work with us
- Right now
- About us
All of these technologies underpin tools and methods that are expanding their area of usage from pure computer science towards industry, construction and the human domain giving rise to terms such as Industry 4.0, Building 4.0 and Internet of Interactive Things.
At the Alexandra Institute (AI) we have been working with these technologies for many years and have observed this development from the inside. Being a GTS institute that develops and facilitates the newest advances in IT, we have gained insights that we try to communicate far and wide.
Working with the newest technology and trying to facilitate it towards usage means we are all technology enthusiasts, though we are not naïve about the baggage that comes along when using it, which I try to formulate below, starting with IoT:
IoT can be, and has been, defined in many ways and forms, but for this, we will use this definition by the IoT Agenda:
The Internet of Things, or IoT, is a system of interrelated computing devices, mechanical and digital machines, objects, animals or people that are provided with unique identifiers (UIDs) and the ability to transfer data over a network without requiring human-to-human or human-to-computer interaction.
IoT is not just a black box sitting near a machine to take measurements of the workflow, but it can also be any kind of animal or human that have been fitted with a device that transfers data over a network. If you are thinking about your smartphone in your pocket right now, you are not wrong. We live and die by the information we get from that device, effectively also making us an IoT device. It will always be possible to argue free will and other meta-physical concepts, but the fact is that we are all controlled by it; being in a new city looking at the map or review apps, or just by being notified all the time about events happening elsewhere.
We try to look at IoT a bit differently here at AI where we try to push IoT Prototyping in Production as a way of moving away from black box devices, towards something that people can inspect and understand. By utilizing off-the-shelf components such as Raspberry Pi’s and Arduinos, combined with cheap sensors and open source software, we try to make the IoT concept more accessible and understandable.
We have used this concept in varying setups, ranging from 5 days of tracking of people flows to 5 months deployments and particle/gas measurements. There are, of course, tradeoffs; using this approach is not as energy efficient as custom-made hardware, the sensors might not be as precise and one should not expect the deployment to last for years on end. Coming from the software world though, we see this as a corollary to the agile development process, where we quickly and cheaply produce results and iterate on that. This is not a catch-all approach, but it can greatly improve the process of implementing a data-driven organization, by lowering the initial costs while evaluating interim results.
A raspberry pi – an example of Prototyping in Production – used to count the visitors in the city center of Ringe. The assumption beforehand was that Tuesday was a slow day, but the data showed that is was the same as the other days. Photo: Alexandra Instituttet
Machine Learning is another central tenet in working towards a data-driven culture and has been defined as such by Tech Emergence:
Machine Learning is the science of getting computers to learn and act like humans do, and improve their learning over time in autonomous fashion, by feeding them data and information in the form of observations and real-world interactions.
ML is, at its core, about nothing more than calculating the probability of an event, albeit using a very advanced formula. It is central to many of the new solutions we see today, such as image recognition and understanding sentiment in texts. If you are now thinking, “well, I always loved the Terminator movies and here we are”, I can assure you that we are not there…yet.
“If you are now thinking, ‘well, I always loved the Terminator movies and here we are’, I can assure you that we are not there…yet.”
ML algorithms are generally very good at tasks that humans are not, either because they are tedious or slow, such as when multiplying large numbers. In the same way that computers in general are much faster at doing arithmetic than humans, so ML is faster at recognizing objects in large quantities of images. But there is tarpits when it comes to the usage of ML, two of the most serious being; ‘adversarial attacks’, where the algorithm is being tricked to produce a result that are not visible to the human eye; and ‘specification gaming’, where the algorithm is shortcutting the process to achieve a certain goal. Look them up, it is very interesting and sometimes funny reading.
We do a lot of work in ML and while some of it can be a bit esoteric, there are three interesting cases that I would like to emphasize; the first is in collaboration with FieldSense, a Danish company helping farmers with their logistics, where we have designed an algorithm that can identify boundaries on fields using satellite images. A great example of something that can be done by a human, but is very tedious work and prone to errors as we lose focus. By exploiting the publicly available satellite images it is possible to train an algorithm to detect and mark fields automatically.
Left is image recognition used to define field borders from satellite photos. The pictures on the right side show ML used to detect pumpkins in a field. Photo: Alexandra Instituttet
The second case is in collaboration with Sign2Me, where we have worked on recognizing sign language for better communication between the users of that language and those who do not understand it. By recognizing the signs using a smartphone and translating it to speech, it furthers the communication possibilities between these two groups of people.
It is a fine example of using the technology in a way that can have great impact on everyday people and does not require great investments on the part of users.
The third and last case, comes from our laboratory, where we are currently working on developing methods for better understanding the training of ML algorithms. If this sounds a bit hairy, remember the definition: ML works by feeding an algorithm data and looking at the results. This has great potential for better understanding what goes on and to prevent some of the tarpits mentioned above.
The last technology to be mentioned here is drones. Most people know drones as these white boxy things that fly around recording film for The Danish Broadcasting’s History of Denmark or as those responsible for shutting down Heathrow Airport for a day. That is not all there is to them, and let’s have a definition from researchers at SDU:
A drone is an IoT device that moves in multiple dimensions.
It is a bit shorter than the previous definitions, but it stands on the definition of IoT. As with all the other technologies mentioned here, drones can be a power for both good and evil. But the fact is that most drones today are used for recording video or taking photographs, which can be annoying and have ethical implications, but are hardly evil in the purest sense of the devil. Their limitations also come from the fact that drones have to be controlled by a pilot. Yes, I acknowledge that you can have them fly from point A to point B and so forth, but they do not know that they are not allowed to fly over closely populated areas. Similar they have a limited flying time of around 20 minutes and limited payload capabilities, at least when talking about the best-known quad copter drones.
“Their limitations also come from the fact that drones have to be controlled by a pilot. Yes, I acknowledge that you can have them fly from point A to point B and so forth, but they do not know that they are not allowed to fly over closely populated areas.”
We work with these flying machines as a tool for completing tasks on a personal level where our human form is an inherent limitation. From that outset we have worked and are still working on a number of cases utilizing drones as moving IoT devices: In collaboration with Force Technology we have developed a platform for gathering and visualizing data about gas leakage from landfills and autonomous landing of drones based on an understanding of their surroundings.
Particle measurement with a drone at Greve airfield. Photo: Alexandra Instituttet
If the last part sounds like nothing new, remember that most drones do not understand where they are flying, but are relying on a pilot to act properly, while our work, lets the drone understand where the roof is and find the best place to land – e.g. by finding the most level area of the roof.
Right now, our most exciting work in this area revolves around developing more autonomy into the drones themselves, thereby making it possible to use drone swarms for complex tasks: inspecting bridges and tunnels, and working in a collaborative way by having drones do different tasks, such as shining light on a pillar while another drone records the reflections. We are also very much aware of the potential dangers of drones flying around in our cities and are therefore also working on algorithms to have them avoid humans, while still performing a given task. As an example, take the recording of film at festivals: the operator wants the best images possible, but also has to avoid flying over large crowds gathering in front of a stage.
The above-mentioned technologies form the ground stone for the 4.0 version number for both the general industry as well as construction. This version number shows the growing emphasis on transforming to a data-driven organization that relies on IoT and ML to gather new insight in how to steer the process. The technologies have been used widely, and implementing them does not have to result in large investments and cultural change in an organization, but can be implemented gradually to nudge the culture in the expected direction. Similarly, the same technologies have changed how we view ourselves and our interaction with our surroundings.
This has led to the new understanding that it is imperative to create ways for humans to interact with technology in a meaningful way, giving rise to concepts such as Internet of Interactive Things. This concept focuses on utilizing the technology to give value back to the humans (in)voluntary interact with these technologies. A trivial example would be a real-time explanation of pollution in the local area, so as to give citizens a more informed choice about where to go for a run.
If you would like to know more about these concepts and trends, do not hesitate to contact me. I would also like to emphasize the InfinIT innovation network as good place to get in contact with researchers and other companies with similar questions.
“We try to look at IoT at bit differently here at AI (Alexandra Instituttet red.) where we try to push IoT Prototyping in Production as a way of moving away from black box devices, towards something that people can inspect and understand.”