The Trouble with IoT

In this podcast, we take a look at the common headwinds in IoT deployments, and how to overcome them for adoption of IoT solutions.

Steve Carr speaks to Warwick Taws, CTO of WiTTRA Networks AB about why the business started, and how they have overcome this challenge.

Transcript

Copy Transcript Copied

Steve: Hi, my name is Steve Carr. Your host on talking IoT and in today's episode called the trouble of IoT. We'll be looking at some of the headwinds that we've experienced over the last decade in terms of IoT deployment and bringing IoT to the mass market. I'm very pleased to have with us today, Warwick Taws, Warwick is the CTO of a company called Wittra which looks to be changing the face of IoT. Maybe you could give a brief introduction to yourself and also to Wittra networks. 

Warwick: Thanks, Steve my name's Warwick Taws CTO at Wittra for the past eight years, through the journey from a very small company to something a bit more substantial today, we provide hardware and connectivity. For lots of sensors, some of our own sensors, as well as masses of third-party sensors, connectivity via a number of different wireless technologies in our unified gateway product, which we'll get more into and then up to a portal and then we work with system integrators typically to deliver something to the end customer. 

Steve: Sounds awesome. You mentioned that you've been with the company for about eight years and now you've got the unified gateway, I've seen some information on it and It looks fantastic, it really does look like it will solve some of the key issues that we've been seeing What was the reason for the company being founded?  

Warwick: The founder Hakan was running in the forest near his home one day and he saw a bicycle, a new bicycle, propped up against the tree.  It was there every day for the next week afterwards and he had not long ago sold his previous company and was looking for something to do so he started thinking about the bicycle, how could he get that back to the owner?  So, he started looking around at asset monitoring, and asset tracking devices that were on the market this is 10 years ago, as we all know technology has changed a lot in that time, back then the market was predominantly cellular and GPS tracking devices.

He thought that they were just not suitable, big, and expensive, and they drain the battery. He started. Investigating the alternatives to actually find this bicycle, that's what started this whole journey, 10 years later, here we are. 

Steve: Sounds awesome, I get the challenges that were there 10 years ago and maybe that's a really good place to start. We've seen lots of historical data from the likes of Cisco and many others showing that a high proportion of IoT projects that go to proof of concept, never quite get off the ground. They never really deliver the value that they threatened to. I know there are lots of different reasons that are cited such as cost, lack of knowledge fragmentation in the ecosystem and many more. What do you see as the biggest challenge? 

Warwick: I think one of the biggest challenges is if you take the customer's perspective and put yourself in their shoes, and you look at the IoT market, the vendor side what is available. If you're a typical end user, you might not know that much about IoT or the deeper technology involved. And you look at the landscape and it's completely confusing because, through the different layers of IoT, you start at the application layer at the top. There are tens of thousands of applications out there on the market. There are hundreds of IoT platforms like Microsoft Azure, AWS and smaller ones. 

There are a few different connectivity standards. Maybe not so many like WiFi, Bluetooth, LoRa and others, and then if we look at the hardware, there are tens of thousands of choices, vendors, and different types of devices. So, if you take that as a whole and look at it as a customer you really have no idea where to start or even, what path you should go down. As you just mentioned, it's a very fragmented landscape There are very few end-to-end providers from hardware through to connectivity and up to the application layer because it's very complex and it's hard to scale a business where you're trying to deliver end-to-end solutions because while 80% of customer needs might be the same in a vertical or even across different verticals, the last 20% will kill your business. You just get tied up in endless, different integrations and customizations. So that's one reason why the landscape is fragmented in itself because it's complex. 

This is one of the huge challenges that customers have. They don't really understand the tech. They have a problem; they just want it solved. But, it’s very difficult to know where to start.

Steve: It's a great overview in terms of how you see the market, but that fragmentation's existed for a while and from what you've said,  If you pull together, the hardware and you make the hardware much more, intuitive, easy to use or easy to deploy,  there is still a  challenge at the backend in terms of the data, I fully agree that people are looking for data and data is really the oil that comes out, but actually your oil doesn't have any value unless we actually refine it slightly further. So, I guess that was the point you were making about also trying to include system integrators. So, do you see the market as, providing the data and not worrying about where the data comes from in terms of the protocols or the radios? And then worry about, the backend refinement of the data, which would come from the SI’s or directly from the end users.  What I'm really asking is how you see your play in that market. 

Warwick: Yeah, we focus actually on the lower layers here. As I said, there are a lot of players out there doing a good job of writing various, end-user applications, AI and BI and all the rest of it. There are lots of platform providers.  And there are lots of technology companies pushing various flavours of wireless connectivity out there as well. What we see is that, if you look at the connectivity layer, in particular, that's one of our areas of expertise.  It's a very siloed landscape. You have Bluetooth salespeople trying to get their sales bonus, pushing Bluetooth products to the customers. Saying that Bluetooth is going to solve their problems.  Look, I'm not Bluetooth bashing here. It's a great technology. It only serves a certain percentage of use cases in IoT and likewise for WiFi, likewise for LoRa, likewise for other various wireless connectivity standards. So, yeah, we're focusing here on the bottom layers, in this whole ecosystem. 

Steve: So, pulling together the various protocols to make that decision easy.  That was really going to be my next question. There are so many protocols. In the marketplace. When I say so many, 10, to 12 different protocols that have dominance, each one trying in its own right. To make the customers choose them. What is the right solution and how do customers get away from the amount of noise that's being made in that marketplace? 

Warwick: Yeah, that's exactly the issue, Steve, one of our recent messages, because we looked at this, we've been seeing this problem for a while in the marketplace and thinking,  what can we do about this to simplify the whole customer journey and make it easier And we thought if we can unify somewhere in this, the multiplicity of layers, and different standards and things If we can unify some of these and bring them together into one product with one management portal, one system, one delivery, one company, using open standards. We have a bias against proprietary. Those companies tend to die out over time.  Open standards are much easier to support and tend to survive for a long time in the market. So, we thought, okay, let's choose a layer in this stack where you have the least amount of variability and that's actually the connectivity layer. There are only a few different wireless standards that are suitable for IoT, so what we've done is created what we call a unified gateway, where we actually run several different wireless technologies or protocols. In a couple of different radio cards and they exist together side by side, working in parallel at the same time on a site, by doing that, we actually can combine all the strong points of each technology and eliminate the weak spots. So, for example, if you take. collecting a lot of sense of data on a site and you look at LPWAN technology.

Unlicensed LPWAN like LoRa or mioty they're great wireless protocols for delivering low bit rate, and lower amounts of sensor data over a large distance. But they are limited in the amount of data that you can send per hour and that's in the EU, that's a legislative thing. It's not a limitation of the technology itself. It's legislated. 

Because you're operating in licensed free bands the legislators quite rightly want to make spectrum and time available for any IoT supplier that wants to be in there. So, they do that by limiting the amount of data you can put over the air per hour in a given channel.  So, you have data limitations with unlicensed. LPWAN. 

But if you combine that in one system, with another technology, that can send a lot of data because it's doing it in a different way, perhaps at a lot shorter range, by combining them, you get the best of both worlds. And that's exactly what we're doing in a unified gateway. And we see that we've really hit the nail on the head 

It's really solving multiple use cases that might exist on one site with a customer. We can pretty much solve any use case they can throw at us now through this combination of different technologies in one system, one deployment.

Steve: That sounds amazing, so in essence, what you're doing is offering the infrastructure that is needed for an IoT deployment. 

Warwick: The classic use case here that I can talk about. We deployed our system in an international airport. And their statement to us from the very beginning is we are looking for one IOT system, one network that we can install that covers all our use cases. And so far, we haven't found it. We've been looking for six years. So, we said take a look at this. And we did a proof of concept or as we like to call them proof of value because we can very quickly show the value in the system. 

It became very apparent to them. After only a week of operating this, it was exactly what they were looking for. We were able to, track a hundred-plus wheelchairs in real time around the site. So that rather than wasting a lot of staff or personnel time looking for empty wheelchairs they knew exactly where they were just by looking at their iPads. 

We were monitoring escalators to know if they were stopped, or in which direction they were going in so helping with pedestrian traffic flow through the airport. Monitoring the one-way doors through passport control; often they get jammed and it creates a safety hazard. Monitoring heaters to say that they're blowing hot air down the walkway, the gangway into the aircraft.  And eventually, they want to also be tracking baggage trolleys across the tarmac to the aircraft and there are over a dozen use cases on this one site. We wouldn't be able to solve those dozen-plus use cases, without a combination of different connectivity options. 

Steve: I love that approach. You were talking about the airport one of the things that just came into my mind is you've got more than one environment and clearly, we've all been used to understanding how we track in the external environment, but it's also tracking internally. So are you suggesting that you're tracking internally and externally through the same deployment?  

Warwick: We can track, or locate, our sensor tags indoors, and outdoors within a site, we are not talking about wide area asset tracking across the whole city or across international borders. We were talking about onsite connectivity and we achieved that through deploying a mesh, a radio mesh. So, a mesh has anchor points in it and the anchor points mesh with each other to create a single or multi-hop network to deliver data from a mobile sensor tag through to our gateway, and, the mesh anchor points have a dual purpose, they also serve as fixed location beacons for tracking through triangulation of the sensor. We can deliver all kinds of sensor data and positioning data with one network, and one infrastructure, we can go down underground, and up in multi-level buildings with good floor-level resolution. It's a great solution and very low cost because we using sub-gigahertz frequency bands. They penetrate buildings, and structures extremely well and go much further in a single hope if you compare them to other higher frequency technologies like Bluetooth, WiFi, ZigBee and others. So, this is what allows us to position the assets and then if you have other use cases that require a very long transmission distance for data from a static sensor, for example, a liquid level sensor in a glycol tank way out on the runway at the edge of the airport, then that's where the LPWAN technology like LoRa or mioty comes in because you can get that data back to the gateway in a single hop over several kilometres if necessary. So, by combining the different wireless technologies, you leverage the strengths of each without having the disadvantages in the equation.

Steve: So, one of the things that I've come up against before and when talking to other people, and I've seen the deployments is when you want to get to decimetre tracking It is a UWB ultra-wideband type solution. 

I'm guessing from the system that you've described at the moment that it is not ultra-wideband, or can you get down to decimetre?

Warwick: Okay. Yeah, good point. So, with our sub-gigahertz mesh, we cannot get decimetre accuracy it's in the order of a few meters. So, let's say similar to what you would get out of, a true GPS system. So, you can think of it as outdoor GPS or very, very power-efficient outdoor GPS plus or minus a few meters. For those use cases where the customer demands, let's say sub-one-meter, sub-half meter accuracy positioning, through our modular hardware system we can have click-on devices that click onto the sensor it's a bit like Lego. 

Next year we will have a click-on UWB module. So that means in those parts of the site, those areas of the mesh where you need high accuracy positioning. You simply click on a UWB radio to the existing mesh routers in that zone or that area. And for those sensor devices, those sensor tags that you want to position accurately, you also click on the same UWB device onto those sensor tags and when they come into that zone with the UWB radios on the anchor points, you will get your sub-half meter resolution and then when you move away into the other areas of the mesh, you'll get your several meters of resolution where you don't need such precision.  What I'm saying is you only need to spend the extra money on ultra-wideband where you need that level of precision in the system. 

Steve: Awesome. There's a whole different episode that we could talk about, which is about tracking, location and positioning but from what you've said and from where we are today it is very much positioning technologies indoors and tracking RTLS real-time location systems haven't really been that effective just because of the cost. From what you're saying you might have the ability, going into next year to offer an RTLS tracking solution as well as a positioning solution inside. Is that correct? 

Warwick: Correct, and you're right. We need to be fairly clear on the difference between let's say a positioning solution and a tracking system or RTLS, typically an RTLS has a reasonably high update rate on the position. The way we look at a positioning system is that it's not really in reality, real-time, the way we do positioning over our sub-gigahertz network is we actually take a few snapshots of position, if a device is moving so you can show some kind of historical trail on a map interface.  Where we do the high accuracy or higher accuracy, part of the positioning uses time-of-flight triangulation when the device stops. So, if you're in a warehouse, and you have, some stock sitting on a shelf typically that's going to sit there for at least a day or maybe a week or even a month, or maybe more depending on the site. So, therefore we can do a sub gigahertz position when it stops and then we don't need to waste battery. If you need high accuracy or you need more real-time, then that's where UWB comes in because it has very, very short on-air pulses or pings. and you can get a much higher update rate than what we can get over the sub-gigahertz network. 

It's about picking the strengths of each technology and leveraging those. And then if you have a weakness, how can you supplement that with a different technology in the same system? 

Steve: So that's a pretty unique approach because as you mentioned earlier, a lot of the Silicon vendors are really trying to push or have tried to push their own standards, their own protocols. I just want to go back to the comment you made which was to do with open standards. In some instances, people talk about open standards, but they ended up having a proprietary system, or proprietary network. 

Is your network open? Does it allow for third-party devices to come into the system? And if so, how does that work? 

Warwick: Yeah. So, one of the things we saw when we very first released this product into the market, 18 months ago, as we have developed some sensors as I mentioned, we have this Lego concept. So, we had a click-on LPTH sensor, which is light, air pressure, air temperature, and humidity, all four sensors in one click-on. But, the feedback that we got is, yeah, this is great, but it's kind of proprietary so we thought, yeah, we need to do something about this and open up our system. 

What we developed where a series of interfaces, click-on interfaces. So, they're not sensors in themselves. They are interfaces to third parties, for example, one of the click-ons is a 4-20mA loop sensor. There are literally thousands 4-20mA’s sensors of all kinds on the market from hundreds of third-party manufacturers and 4-20mA loop is a way of expressing an analog value from a sensor. It could be temperature, it could be humidity, it could be liquid level, it could be pressure in a pipe, whatever, there are hundreds of these on the market. So basically, you can take any of these third-party sensors. Wire it into a click-on, click it onto the sensor tag, and then you're carrying that analog data over our wireless network, whether that's by sub-gigahertz, mesh or LPWAN, mioty or LoRa or something else. It doesn't matter. The point is you get that data from a third-party sensor, non-proprietary over the network to the cloud, and then you can draw down the data from there. 

Steve: So truly open in terms of the approach. Just touching on some of those data elements, what is your view on edge-based data solutions, decisions maybe could be made at the edge. What is Wittra doing or thinking in those areas? 

Warwick: One of the original approaches, when people started doing AI, was to have massive compute power in the cloud because that's where it can exist efficiently at a very low cost. But then you've got to get that massive amount of data to the cloud and that's the problem, especially when you start talking about wireless devices, sensor devices that have wireless and particularly devices that move you can't wire them. You can't cable them. 

So, they have to exist with a wireless transmission medium and that by design means that they're battery-powered.  You try streaming a hundred kilobits or 10 megabits per second of data from a wireless device into a network and see how long the battery lasts. It's not feasible in IoT. 

Consumer devices like mobile phones, everybody's used to charging those devices daily, but it doesn't work in IoT. You need 12 months plus, preferably several years of run time between charges. So, it's not an option to stream lots of raw data to the cloud for analysis. 

So that's where in more recent years, people have started looking edge processing. In our case, we actually doing lightweight AI on the actual sensor tag itself. So, in other words, you have a low-power AI recognition engine running on your sensor tag. 

Looking at various sensors could it could be an accelerometer, that's a very power-efficient sensing device. So, you can look at accelerometer data. It could be, for example, detecting vibration on a portable machine, let's say a forklift or something like that, driving around a warehouse. And you want to look at the preventative maintenance measures, the vibration characteristics and the electric motor that's driving that forklift. So, you're sitting there with your tag, analysing vibration data from the motor and this lightweight engine thinks it picks up a match to some kind of an anomaly. So what it does is captures a window of raw data, and then it transmits that window of raw data up through the low bit rate, low bandwidth network up to the cloud and the cloud can then do a much deeper analysis on that and say, yeah, well that's a match to this known anomaly. So yeah, you should go and have a look at your forklift or no, that was a false alarm. So do nothing. 

That's the approach we're taking at least, with fog, and edge, there are different terminologies here, basically, we are taking a distributed AI approach where we do lightweight processing at low power on the sensor device itself and then send windows of raw data up to the cloud for deeper analysis 

Steve: Seems to make a lot of sense. In essence, it is exception reporting, you are only reporting on the data that you need to understand and we see too much data and we can't consume it all anyway. Right.  Thank you very much for your time on the podcast today, but before you go. I do have a final question. 

What do you see as the next emerging? Killer application within IoT. Big question. 

Warwick: Sure, yeah, I guess, one of the things is an increasing use of AI in IoT. IoT produces tons and tons of data and especially in a system like ours, where you are producing not only simple sensor data, like temperature data. You can produce a fair amount of data just from a hundred or a thousand temperature sensors. But positioning data, especially when you look at a more real-time location system that produces gigabytes of data per month, making sense of that really lends itself to AI. 

So, AI is a big developing area and everything that goes along with AI, like edge computing and TensorFlow and all these different aspects of AI and machine learning. So that's a fast-developing area. 

I would also say that we've actually started now to see the first green shoots in this unification theme that I talked about. We were at a conference and, we saw a Silicon vendor who has now released, a unification product in LPWAN. So, they've released a silicon radio, which incorporates in a single chip both a mioty radio and a LoRa radio and these LPWAN technologies they compete with each other in the unlicensed LPWAN space, but here's a Silicon vendor hedging their bets and merging both competing technologies into a single piece of Silicon. So, we see this as a really great thing for the end users, the consumers in IoT because it removes, the complexity of the decision about which technology, and which vendor they should choose if we can pull that out and just make it simple for the end user the whole industry is going to move forward at a much faster rate. 

So, I would say, yeah, for me, two big things are AI and everything involved with that and actually unification. 

Steve: That’s awesome, you mentioned very quickly mioty so for the listeners I believe that you have information on both LoRa and mioty on the Wittra website, which is www.wittra.io so people can go there to learn a little more about them. Thank you very much for your time and we wish you and Wittra every success. 

Warwick: Thanks a lot Steve, my pleasure