Over The Edge into The Wild – When AI Met IoT

This spring marks a 5-year anniversary of putting into operation of a country wide LoRa based IoT network in Serbia (https://mts.rs/Poslovni/IoT). The handpicked team of devoted professionals from Telekom Serbia, Actility, TERI Engineering and Busit had joined their skills and forces to make it happen. If measured from the real inception – the first PoC – it took us nearly 2 years to get there. The ride was rough, but the motivation was strong, and the result was a reward by itself.

Today, while the network is perpetually growing and the “things” are being added for the benefit of various verticals, let’s have a closer look on that driving force that had generated such a great enthusiasm. Moreover, let’s see if it had delivered its promise.

Frankly speaking, 7 years ago even the mammoths of business research and consulting had an issue with distinguishing what the IoT really means hence their lavish projections for the global IoT market in years to come (meaning now) were taking into account smart phones, TV sets, and whatnot. Therefore, in average they have seen this year in a zone of well over 3 trillion USD of IoT endpoint spending.

These days we get to see the real-life statistic and it is clear that the numbers were not reached. Hence the more moderate forecasts for the years to come. Where once it was a x2 now we have something that appears to be much closer to sqrt(x). While for the most of us the actual monetization of the project was not the number one of the list of drives, all of us wanted to have our business leaders happy with our achievement thus giving us justification to invest more into taking hold of the new technologies and knowledge. OK, the trend is global and although it is not hitting us locally (yet) it does have scary implications on the macro scale. This year SAP, IBM and Google are shutting down their IoT platforms and while there are expectations that the enterprise customers will all migrate to Azure and AWS, all the sudden the infamous quadrants are peppered with so many niche players riding on open-source waves. And it does not have to do with the link layer – no matter if the last mile was WiFi, LoRa, or 5G or with the ability to process data and wrap it all up into the shell of a business app.

So, what went wrong? Was it because the “digital water” in those data lakes became stale, and they all turned into data swamps? Or maybe because it is really so easy to implement a shoot-and-forget principle using all the offered features of the said platforms and so hard to extract the real value from the date and bring it to its meaningful use?

IMHO the later offers a partial explanation. Speaking from experience, our own LoRa based solution had been built to provide all tools and means that will support the network expansion, an easy-to-do addition of endpoints and third party applications. There is a user friendly NMS/DMS, sleek application IDE, OSS an BSS as well. All functioning smooth and dependable. Nevertheless, we are frequently facing another type of a challenge – the end users want more from their IoT endpoints than just a mere data reading or remote actuation. They do want smarter “things”, that will help them achieve what they need without fishing in those damn lakes.

Sure thing, it appears to be much easier case for the Industrial IoT with all those edge computers offering some level of independence to IIoT devices for their data management and analysis. However, after more than 30 years in industrial automation I am confident that nothing beats smart “things”.?And for that opinion I have very powerful advocates. Intel, Qualcomm, ST and TI to start with but also some of the smaller players among the semiconductor manufacturers, have been steadily building up their portfolios of components, firmware libraries and toolkits supporting development of IoT devices that are either capable of independent learning or acting based on AI-like heuristic programs. As a result of that their revenues are steadily growing. Clearly the main reason for that is the demand from so many ODMs trying to push AI over the edge into the wild ecosystems of IoT endpoint devices.

While the trend is still very shy of going too far, the actual demand from the market keeps pushing. So, no wander that one our clients in Romania had asked us to come with smarter sensors that not only detects an event of opening the trashcan lid but also a detection of a bag overweight (anomaly). Confused - on the first thought. Back at the whiteboard we realized that the guy was not fantasizing. At least not more than those smart kids that came with wearable health monitors that report anomalies, not mere measurements. And there is no end to it. Intel and the likes are rubbing their hands.

So, there it is: AI met IoT. They got married. Their kid’s name is AIoT. It is growing fast and wants to play. You better join while the gate to the playground is still open. ?

For those automation folks gatekeeping the edges and punching the EN and ISO tickets – you will maybe prefer to wait for the kid to graduate and become a chairman of the Industry 5.0. No worries. It won’t take too long. It may as well happen while you are reading this last sentence.