Reading the large amount of news on connected technologies, I often wonder what the writer actually means by IoT. I do not want to open the Pandora vase holding the discussion on the definition of IoT; there is already a huge corpus for this and any reader animated by a scientific approach — even a slight one — should look into it and find the definition that suits him best. It is actually interesting that while I was writing this piece, on the IoT Agenda site there is yet another definition. As in the standards world, where too many standards mean no standard, here too many definitions are heading toward no definition. I have the feeling, though, that there is a strong interest in leaving a certain amount of fuzziness on the topic. As it’s possible to play on the vagueness of the concept, everything that has a sensor can be labeled “IoT.” I’ve heard of standard HVAC systems that suddenly became IoT; even better, a switch and a lightbulb are now IoT as they are connected.
“Smart” is likely the most used word in this era, trespassing the boundaries of technology and entering the real life. We live in smart cities, where we can find a smart parking using smartphones connected to smartwatches. Everything now that has a nanogram of silicon is smart. In the past, I’ve often made the remark that smart is a synonym to intelligent. Intelligent comes from the Latin “inter legere,” which means to read between the lines, to go beyond the surface. This definition implies a certain amount of nondeterminism as there is an interpretation of reality, and not a mere action-reaction, we enter the world of subjectivity. It’s very easy to see it when two very intelligent people debate on a topic, having opposite opinions. Therefore, literally at least, a smart device is a nondeterministic device. Now that there is huge excitement on AI, maybe we’ll be entering the realm of really smart objects.
But my focus here is a bit different. What I’m interested in debating is what the components of an IoT system are, and when a system can be labeled IoT. In the internet, the discussion is very easy. Internet is a global network of computers that use the IP protocol — and more in general, the TCP/IP protocol suite. If we add small objects to this definition, using a lightweight approach, we could just be extending network connectivity to objects that don’t use the IP protocol. My Bluetooth mouse would then belong to the IoT universe. Back in 2008, I defined IoT as “a worldwide network of interconnected objects uniquely addressable, based on standard communication protocols” — simply extending the definition of internet. I’m not sure, 11 years later, if this would be the best definition of the phenomenon.
IoT in most cases is perceived in a different way. IoT is when atoms and bits meet. It implies that there is a real word and a digital representation of it, often called the digital twin. The transmission, or network, mean is irrelevant in this vision, and if IPSO had its way, there the IP protocol will be everywhere. This enlightens two facts: First, that the I in IoT just means connectivity; second, given then that there is a necessity to provide an interaction between the digital and the real world, the presence of sensors and actuators is a necessity. So far, I am discovering the hot water, as this is basically the definition from ITU-T dating back to 2005. But do all sensors belong to the IoT domain? In other words, I read recently about something labeled “IoT sensor” which implies the fact that there are sensors that are not IoT. Does it make sense? Not really in my opinion. Every sensor that is instrumental to create a digital twin is belonging to the IoT domain, no matter how simple or complex. My understanding is that an IoT sensor is used as a synonym to smart sensor. This implies that there is a clear class distinction between dummy sensors and smart ones; supposedly, the second class is operating some sort of processing before sending the data. I guess a temperature sensor that transmits only when the temperature changes, so it has a byte of memory to store the information of the previous temperature and a circuitry to see if the data corresponds to the new measurement, may not be smart as it does not process the information. But how much processing is needed to classify a sensor as smart — forgetting for a minute that smart means nondeterministic? If a sensor is able to send the temperature in Celsius or Fahrenheit, and to switch between the two upon reception of an input, would that be enough?
Closing this digression, and saying that it seems logical that any sensor can be part of an IoT system, we get to the point if an IoT system itself must be smart, or the processing capability is something that is common but not necessary. If it’s not necessary, then connecting sensors and actuators would be enough to make an IoT system. A light switch and a bulb would then be indeed an IoT system — maybe not a traditional one, but we just need to connect the bulb and the switch with a wireless bridge so that the switch uniquely identifies the bulb and sends a command using a communication protocol. Or a wireless temperature sensor with a screen. In my opinion, this sounds indeed too little to be considered an IoT system; we need some processing in order to separate purely electromechanical appliances from the IoT world.
Summarizing, I came to the conclusion that that an IoT system is a system that:
- Instantiates a digital representation of a physical entity — a digital twin;
- Uses of sensors of any kind and is able to modify the environment using actuators; and
- Is able to perform — at least some — information processing.
All IoT Agenda network contributors are responsible for the content and accuracy of their posts. Opinions are of the writers and do not necessarily convey the thoughts of IoT Agenda.