Author: Luka Mustafa, IRNAS’ CEO
The Things Conference 2021
At IRNAS, we are creating special hardware solutions and IoT devices, end-to-end for our clients. Over the past few years, we have been talking at The Things Conferences about some very interesting aspects of our work. We’ve taken the audience up in the air with a drone mapping LoRa signals, we’ve shown them the things under water and how LoRa behaves in those difficult cases, and last year, we’ve taken participants to high voltage, with real Tesla coils on the stage.
This year, unfortunately, the conference could only be done virtual. But they did an amazing job and we were happy to talk to all participants about how IoT has grown up, and how we can implement it into a lot of industrial actions.
To ease our thinking, we have split IoT devices into 2 categories:
Advanced IoT Edge solutions
How we build robust solutions
1. Single-purpose solutions
We define single-purpose solutions as IoT devices which have one purpose only, and that one purpose is very well defined.
One example is Sensedge’s Senstick, which is built to monitor micro-climate environments. It is a well defined solution which senses a known number of sensors in very well defined use-cases and environments. Because of that, it is very well built and over the years, it has been substantially improved. It is in a large production volume now, and since a lot of background work has been required to build a robust and reliable solution it is today, let us bring up the 75 % : 25 % rule here.
75 % innovation : 25 % homework
At the moment, building IoT solutions is mostly about homework with a ratio of 75 % homework demand for 25 % innovation on a solution. By homework we mean:
- dealing with the network stack,
- making sure the device runs properly, and
- building the whole system allowing us to take some sensor values and send them to the network server.
We would like to flip this around. We are working on an approach which will enable us to spend most of the time on the innovative side of the actual solution, getting it smarter, better, having more budget and effort for those aspects of the IoT solution, and not spending as much on the homework.
The Generic Node
It is a nice end-to-end example of hardware and software integration for a single-purpose LoRaWAN device. This allows us to build things end-to-end with a relatively simple process, on a basis, which is well debugged, well tested, and also used by a lot of other people, enabling this collaborative effort that we have seen grow with the community network of The Things Network. Our role in this process is building things on top of the design, putting our expertise of building very robust devices, very challenging sensing applications, and creating novel solutions, focusing mostly on innovation and much less on hardware. And if we all join our forces over this, this enables us to grow the LoRaWAN ecosystem much better and much more rapidly.
Our IoT devices are robust
Every year, we like to put something interesting on The Things Conference stage, and though we were virtual this year, we kept up with the tradition and showed the audience the following experiments with our IoT devices:
- We threw one rugged IoT device off a bridge, let it hit the ground, but also send a message just before it hit the ground, so we knew what’s about to happen, and also getting a report it survived.
- Some of our devices are taken deep under water (by tracked animals for example), and while we do not have very deep water in Maribor, we do have a test chamber which brings us almost 500 m deep, allowing us to validate the devices.
- Because it’s winter, we need to make sure devices survive at very cold temperatures. Batteries are always a special in those conditions. So we’ve built devices, put them out to very harsh conditions and validated them.
- Lastly, we drove over one of the devices with a car, to show they are really well-built and robust.
How we achieve robust IoT devices: Minimizing overhead with software defined solutions
Focus on creating better IoT solutions is minimizing the overhead and software defined solutions enable us to do this much better. Over the past year we’ve been sitting at home and created this whole Technology Roadmap, and gained the experience with all the relevant components, to build robust, long-lasting, and very easily upgradable solutions, which are mostly defining the software.
This has also allowed us to approach these projects with no-stack methodology. Particularly LR1110 with the Modem-E approach, integrated it with The Things Industries stack, allows us to create IoT devices without worrying about the stack and LoRaWAN too much, and focusing really on the added value solutions of this. Furthermore, by building the full integration with Nordic Semiconductor, Zephyr real time operating system, machine learning from Edge Impulse, and other components, we can create better, faster solutions very efficiently.
Hereby, we would like to appeal to all of you, to look at your technology roadmap, define it well, and then stick to it because this makes you efficient, and this creates much better products than jumping from solution to solution.
2. Advanced IoT Edge Solutions
IoT has grown up, and Advanced IoT Edge Solutions is second category we have announced in the introduction.
We define these solutions as IoT devices which are multi-purpose, doing multiple things, maybe having a primary role. But most importantly, doing much more than just collecting a sensor value and reporting it to the cloud. They interact with other devices and users alike, they particularly evolve over their lifetime. While we may have a piece of hardware in the field, and actually we want to put it out into the field as soon as possible, we keep evolving the logic and the brain of that device over the years to come over their lifetime, so that it becomes smarter, better and more efficient at what it does. A key unlocking part of this is obviously machine learning. Because that allows us to make complex decisions on a lot of sensor data, and just send a small answer out, which means we conserve a lot of battery power, and that allows us to build more long-lasting solutions.
How much added value do you create per mWh of battery consumption in your IoT device?
The key question we would like to pose is: How much useful things do you create, what is the added value per mWh of an IoT device’s power consumption?
This is a good way we can think about how do we make sure most of the actions are actually useful. Let’s break this down in a few cases.
Instant alerting is a concept where we can look at the edge IoT device being dormant, sleeping most of the time, conserving power, energy and spectrum, but at the same time having the ability to instantly say, hey, this important thing happened, you should really know about it! Best case to explain this is a project from our client, Izoelektro, RAM-1 for electrical grid monitoring. The basic function of the device is arrestor monitoring, making sure the component in the electrical grid is fine. As soon as something bad happens, we want to instantly know about it, so, obviously, this will let us know about that particular component. But because we are monitoring the whole electrical grid at that particular point in space, as well as we are seeing all the other effects, we can be so much smarter about how to make use of this information.
With the use of machine learning and some very nice mathematics on the devices, we can classify events and say: these locations are important, this is happening, you should be aware of it. And this gives really the extra added value from the same piece of hardware we had initially for the basic measurement function.
The second important concept is long term reporting. We have sensors, we have robots, we even have people, animals, everything that is moving in various environments, so we want to collect data over time. But also that data tends to be quite similar in a lot of the cases. We want to communicate only the most important parts, or really get the part, which is important, right away instead of waiting for a long time before we report it.
So that ties into the instant reporting, but looking at the case of DrainBot, an autonomous tunnel drainage maintenance system, we see we can control robots, we can put them in very difficult spaces, where they will slowly traverse underground in tunnels and so forth, but immediately when they come up with something interesting, they let us know through LoRa network, which works very nicely in such constraint and also underground environments. Instead of waiting another long while before the unit ends up somewhere with mobile data coverage.
ElephantEdge is a project we are working on actively with SmartParks for the past months. This allows us to create a meaningful solution full of latest and greatest technologies. But more importantly, it also showcases the interaction part between different trackers and IoT devices, and the potential this unlocks.
We have a tracker where we track the position, quite a standard solution. However, the interaction part allows us to even more: we can have an upgrade-node somewhere by the water hole, and when the elephant comes close by, we can upgrade it through BLE very efficiently. We can make it exchange data with other trackers or positions, we can use beacons in the filed or other things from that to determine proximity and location. We can even look at social behavior with such applications.
This is a very good example of how we can collect data from the environment, or have very basic low power and efficient communication either on LoRa on BLE with other devices around us, and with that create a much better solution than just standalone points, reporting to the cloud. As they become interactive, we gain much intelligence about our environment.
So obviously, we have elephants in very remote areas, and while we would like to have perfect network coverage there, that is not always the case.
Let’s make this possible from anywhere around the globe. So let’s take things to space.
We’ve been working with Lacuna Space company to evolve the relay idea. We are doing the implementation side for a real product, because we see very great need for this with many of our clients and projects in the field. Not all devices always have a direct line of sight to the sky, not all devices can necessarily have the right antenna to be able to talk to the satellite. But also, we get the additional opportunity of things aggregating data at a small, low-powered local node, which is not a gateway.
The not-a-gateway approach is something we would like to keep clear in this discussion, because we can do so much more if we are very low power, if we are smart about making decisions on the device side. The relay we have shown here below – we can treat it as an edge IoT device, which we can use to make some decisions, to aggregate information from a few local nodes, and make decisions what to send immediately, what is reported as statistics over time, and also seeing what is really critical so it should be alerted right away. The solution we are showing here combines the network from Lacuna Space, a really nice design by Fabien Ferrero from RF Things company, and our code, running on the device, creating this relay solution which is not only a backup channel for critical infrastructure applications, but can also do a lot more.
What have we learned?
- Firstly, it’s about the homework. Doing a lot of homework is nice, but it does not necessarily create useful solutions, and it does not boost innovation.
- Single-purpose devices are very important for the IoT space. But let’s make sure we define them as such, and when we are developing them, we focus exactly on that.
- On the other hand, we can let our imagination go wild with Edge IoT devices, which are low-power, which are capable of a lot more, and with machine learning we can keep extending them indefinitely.
By combining these two aspects and by really being sure which one of the two you are building, you can be very successful in IoT projects.
Looking forward to seeing you all in person at The Things Conference 2022!