Interview with Ubuntu boss: A rich ecosystem for robotics and automation systems

Share this story

Exclusive interview with Mike Bell from Canonical, the company behind Ubuntu

There’s a lot of talk about the Robot Operating System. Understandably so. But the most widely used operating system in robotics and automation systems development is actually Ubuntu. 

In fact, ROS is not actually an operating system at all – it’s a set of software frameworks, or a software development kit, to be installed into an operating system like Ubuntu.

As Mike Bell, executive vice president of internet of things and devices at Canonical, explains in an exclusive interview: “It’s a bit confusing because it’s called Robot Operating System, but the reason is because if you’re developing robot applications, you don’t need to worry about the fact that it’s running on Ubuntu.

“You just write your application to interface into a ROS SDK and then deploy on Ubuntu.” 

ROS can be installed on a variety of systems, but Ubuntu is probably the best platform it – Ubuntu is at the top of the list on the ROS download page anyway.

Ubuntu is one of the most widely used distributions – or versions – of Linux, the open-source operating system, which itself is based on Unix.

And although Ubuntu Linux is a free, open-source operating system, many companies – like Canonical – make money from the services they build around it.

Canonical, however, is different from other companies, many of which in fact make quite significant sums of money with Ubuntu Linux – Canonical is actually the company that maintains Ubuntu Linux.

In old fashioned terms, Canonical sort of owns Ubuntu Linux, although it must be said that no one actually owns open-source software like Ubuntu Linux – that’s why it’s called open-source.

The whole open-source movement probably started when Linus Torvalds created the Linux kernel in the first place, back when he was in university in Helsinki, Finland.

A kernel is the central, core, or most important part of an operating system, and the Linux kernel started out as Torvalds’ hobby.

After being the principal developer of the Linux core for a long time, Torvalds invited other programmers to help develop it, and many accepted the invitation.

And so, not only was Linux development accelerated, an entire global software movement was established, resulting in a massive amount of software development in many fields, one of the fruits of which is the Android operating system, also based on Linux.

Linux is essentially the ecosystem for the huge variety of smartphones, tablets and laptop computers, as well as many other devices that we see in the world today.

If the old system persisted, all we would have is macOS this and Windows that, and maybe one or two others, which would have meant a limited range of devices, and far, far few people having the opportunity to develop their own software and hardware, either by themselves or with other developers.

Torvalds is also credited with the invention of the git system of revision control used in software development. And maybe he could also be credited as being the facilitator of ROS as well, albeit indirectly.

Linux was first released in 1991, and the latest version was released at the weekend, on 17th September, 2017.

In any case, as Bell says, “ROS is a set of toolkits that run on Ubuntu, it’s not a separate operating system.”

Mike Bell, EVP devices and IoT at Canonical / Ubuntu
Mike Bell, EVP devices and IoT at Canonical / Ubuntu

An audience with Canonical 

Canonical calls itself the company behind Ubuntu. It was established at the same time as Ubuntu to provide the necessary support to help the OS reach a wider market, a market which now includes many governments, industries and maker communities.

The flexibility and adaptability of Ubuntu Linux is probably key to its success – there are versions for any system, from a tiny device built by an enthusiastic maker, to servers at a large-scale data centre or public cloud.

Bell explains: “Canonical is the company founded back in 2004 around the Ubuntu Linux operating system.

“Since then, we’ve divided the company up into two divisions: one that I run, which is the IoT focus; and the other, which a colleague of mine, Anand Krishnan, runs, which is cloud part of the business.

“The client service side of the business has been accelerating over the last couple of years, and we’ve seen great momentum.”

“For example, within the public cloud – say, AWS, Google, Azure… we are now on over 60 per cent on all of those operating systems running on those public clouds.

“We’re now seeing the traction within the internal data centres that those large enterprises have, so we’re starting to win large enterprises.”

Ubuntu is the the reference OS for OpenStack, an open-source software platform to manage cloud computing infrastructure. Recently, OpenStack was updated to make it easier for companies to change cloud services and infrastructure providers more easily. This might accelerate the business growth of Canonical.

The main downloads available at Canonical are:

  • Ubuntu Desktop
  • Ubuntu Server
  • Ubuntu Cloud; and
  • Ubuntu Core

Then there’s Ubuntu Kylin, which is created especially for Chinese users. Also, there is a variety of “flavours”, with different applications and settings.

Support services are also available, and given that the operating systems are free, it’s where the money is.

The Desktop is the platform for growth 

Bell says he manages the division of Canonical which deals with all the compute that is “not within the data centre”, and that includes the Ubuntu robotics platforms, which the company categorises under internet of things.

As well as its many other notable successes, such as being the OS for the Robotis humanoids in the main picture, Ubuntu was used in the majority of robots which were entered into the last Darpa Robotics Challenge.

One of Canonical’s main offerings, and within Bell’s area of responsibility, is the Ubuntu Desktop, the free, open-source alternative to Windows OS or macOS.

Ubuntu Desktop is currently one of the main drivers of Canonical’s growth. “In many ways, one of the reasons for the success, in both the cloud and the growing we see in the IoT space, is really centred around the fact that many of the developers and innovators of the world are using Ubuntu Linux Desktop as their developer platform,” says Bell.

“Right now, we estimate that there are at least 3 million people developing on Ubuntu-based platforms. So what we’re starting to see now is the adoption within the embedded space.

“We kind of use the term IoT – it’s a bit lazy in many ways it’s just a nice catch-all term. But really it’s for embedded computers.

“And the compute ranges from everything from robots, drones, through to home gateways, appliances in manufacturing sites, edge gateways, and so on. There’s a whole range of different use cases that we address.

“And for that, we then took our Ubuntu operating system and did a targeted version, knowing that people are cost-sensitive and the compute generally is a lot less than you would get on a desktop or and server. So we have a minimal version of Ubuntu called Ubuntu Core, which we target at those embedded devices.”

And while “cost” may ultimately refer to money, the amount of processing required to perform tasks is becoming increasingly important. The more efficient a computing operation, the less cost in data storage, transfer, and processing, which may ultimately lead to smaller, more energy-efficient devices.

Helping developers develop things that just work

As you may have guessed by now, Ubuntu plays an important role in the maker community, being as it is on the Raspberry Pi and easy to install on Arduino.

But the commercial success of Ubuntu has been in the enterprise market, where it’s the OS of choice for companies developing advanced technologies in transportation, manufacturing, logistics and many other sectors.

“One of the things that’s important to us is to make it easy for developers, says Bell.

“We want to make sure that they don’t have to get into a whole bunch of really difficult hardware enablement issues, with device drivers for Linux and things like that, so we work with companies like Intel and the ARM licensees like Samsung, Nvidia, Qualcomm and so on.

“And what we do is work with them to make sure that on their key platforms for embedded compute … that we enable Ubuntu on there.

“Now, there could be some products where they actually ship Ubuntu on the device.

“Sometimes, we’re enabling a chipset that someone can download and it will just work. Other times, like in the case of the Raspberry Pi, which is a single-board computer so you’ve got everything running there, we enable the whole lot.

“So it varies.

“What we find is the maker community tends to focus on boards like the Raspberry Pi, and what we’re seeing now already is people trying to take that to market on the CM3 [Raspberry Pi Compute Module 3] board.

“But we’re also on various Intel platforms for robotics applications, such as [Intel] Euclid, where you have a RealSense camera, and then Ubuntu installed on it.

“So we have a range of different ways that we enable.

“Our main thing is really to get on as much hardware as possible, whether it’s just the silicon itself, or onto systems deployed by a manufacturer onto a single-board computer, so that when someone buys it, they can literally just wire it up to a bunch of displays and whatever and then suddenly it’s running – they don’t have to install anything.”

The Nvidia Drive PX, the autonomous driving platform, which uses Ubuntu as it OS
The Nvidia Drive PX, the autonomous driving platform, which uses the Ubuntu OS

industryOS 4.0 

Canonical can boast an impressive array of corporate clients, most notable of which, perhaps, are Nvidia, Dell, and Bosch.

Industrial giant Bosch uses either a custom version of Linux that the company has built by itself, or it uses Canonical’s Ubuntu distribution. Moreover, Mayfield Robotics – a division of Bosch – has built the Kuri home robot on the Ubuntu platform.

Through the chipmaker Nvidia, Ubuntu can be found on the vast majority of new cars which have self-driving features. Auto manufacturers buy Nvidia’s chipsets, onto which Nvidia builds its own toolsets using Ubuntu as the base.

As Bell explains: “Right now, the primary platform that you see most of the manufacturers – the likes of Volvo, Daimler, Tesla – going to market with is the Nvidia Drive PX – and that is enabled by Ubuntu.

“So when you go onto their website you find that it’s actually our distribution.

“We worked with Nvidia, they’ve slightly modified it, and we support what is basically an Ubuntu operating system to enable people to do self-driving car development.

“Also, for Nvidia, they also have something called the DGX workstation – and that is an artificial intelligence machine learning platform. It only comes with one operating system, and that is Ubuntu.

“So there’s a couple of interesting use cases in terms of self-driving cars, AI and machine learning.

“I think Nvidia were recently voted one of the most innovative tech companies by MIT, so it’s great to be able to partner with a company like Nvidia.

“If we go to an industrial applications, Dell have a product called the Dell Edge Gateway 5000 and 3000.

“Those products are not your typical kind of Dell PC or server. These are designed to go into harsh environments, whether it be monitoring environment in a utility, or a manufacturing site, or even construction management.

“So the edge gateway there is to bridge the operational technology environment – between the information technology and the OT.

“Typically, in those kind of industrial applications, you can have segregation of the network – separate the things that are very operational from your typical IT network.

“The edge gateway is a really great way of taking the vast amount of data, doing a level of processing within the operational environment, and then having it either go straight to the cloud or back into the enterprise IT network.

“Dell has been specifically targeting the industrial market for that edge gateway. It’s a very robust product – big heat sinks down the sides, but in a form that enables it to be located in a factory or in a machine room, rather than a typical data centre application.

“Another example in the industrial space is Bosch,” continues Bell.

“Bosch Rexroth, which is one of its subsidiary companies, has a PLC product – that’s a programmable logic controller – controlling a low-level input-output between machinery in typical plants. They could be manufacturing plants, they could be utilities, or oil and gas installations.

“And that PLC is running Ubuntu Core, similar to the Dell Edge gateway.

“So both products have enabled Ubuntu Core to run as a secure embedded operating system controlling industrial applications.”

And while driverless cars are a new thing, and a relatively niche market, the industrial sector – manufacturing, logistics and so on – is one of the oldest, which may explain why it’s one of the most conservative.

Until the past few years, most factories were not even connected to the internet – and many still aren’t; none of their machines had sensors on them connected to internet gateways, and so on. Now, with Industry 4.0 and the industrial internet, the machines are talking and moving to the cloud.

“Traditionally,” says Bell, “the world of operational technology that you would normally find in a manufacturing plant, a utility or an oil and gas installation, is very locked down, lots of firewalls, and doesn’t change.

“Asset life is quite long in those environments. When you put a plant in, a manufacturing line in, you’re expecting that to have a life of the seven to 10-plus years.

“And they expect, therefore, the devices monitoring those also to have a long life. So you’re not talking about consumer, three-year maximum lifetime, you’re talking about much longer lifetimes.

“As those environments have started to take advantage of cloud-based services, that also means that they have internet and the enterprise IT networks which reach into those networks, which means security becomes a paramount priority for the people running nose OT environments.

“And one of the things we do with Ubuntu with every platform we run is that we are extremely adept at keeping our operating systems up to date in terms of security patches.

“So everything is secure at that point in time.

“Then, days or weeks later, someone finds a vulnerability, and then you have to fix it. There’s no operating system that can say they are completely secure forever, so the important thing is how quickly do you update and patch your operating system and then how do you get the update out in the field?

“What we have done with a Ubuntu Core is we’ve made all aspects of the operating system upgradable over the air, so the devices can take an update as soon as it’s given.

“So the devices can see that there’s a new version and download it straight away.

“That’s how we address security, and that’s why we’re getting traction in the industrial space because people are concerned about security and Ubuntu Core is focused on security.”

The OS for cloud robotics 

For quite some time, many experts have been talking about cloud robotics and cyber-physical systems, which is where you have humans and robotics and automation systems all working together in an integrated environment.

The idea that you can connect up humans, robotics and automation systems across an entire factory or even a whole series of factories around the world appeals to many companies, large ones in particular.

At the moment, there’s no easy way to do this, and most industrial companies which do this develop their own solutions, working with computer technologies companies like Canonical and others.

But the progress towards an adaptable, flexible and easy-to-install cloud robotics system is accelerating, and Ubuntu will almost certainly be one of the fundamental software components in such an infrastructure – it already is to a significant extent.

And the cost of creating such cyber-physical systems will be within the budgets of even small and medium size businesses, suggests Bell.

“If you think about Moore’s Law,” he says, “the compute power is increasing but the price point is decreasing, and the devices and components are also getting smaller.

“So, all those three factors mean that actually putting hardware inside a device – whether it is a robot or anything else – becomes more cost effective.

“A lot of the reason robots were controlled with a tablet computer was because it was too expensive to put the computer in the robot because of the complexity of what would need to be done.

“Now what you’re seeing is hardware of the right physical size and at the right price point to enable that to be fitted into a device.

“I think this trend will become increasingly important.

“What would have been unaffordable only three years ago is perfectly affordable today. I mean it depends on the device but still.

“This is why we think there’s a real opportunity for Ubuntu running on IoT embedded devices, because we’re able to treat the device as a platform and have multiple applications running in an isolated, contained secure way.”

Bell talks about the possibility that industrial robots will increasingly integrate the functions of a PLC, or computers. “The traditional industrial robot space is conservative, but what’s been happening in all of these spaces – a bit like PLCs – is that people will have proprietary or expensive embedded operating systems running on it, running very bespoke developments.

“What we’re able to see is people deploying much more open platforms that they can access in the cloud as well as on the device. So they can use the same toolkit. And we see this happening a lot.”

The new Apple home automation device, the HomePod
The new Apple home automation device, the HomePod

Siri, tell my driverless car to exit the garage and meet me at the front of the house 

Arguably the area of robotics that will bring forward many of the new innovations is service robotics, or personal robotics – the little toy-like things that are meant for the home, such as Kuri.

One could also include smart home devices such as Amazon’s Alexa-enabled Echo and the Apple Siri-enabled HomePod in this sector, since they are devices which contain virtual robots – although the devices themselves do not move.

The key is connectivity to the cloud, where these robots or devices can access huge amounts of compute, which is necessary for one of the most critical functions – natural language processing.

Voice recognition and natural language processing require massive computing resources even for people who speak eloquently, in correct grammar and don’t mumble. It’s a challenge which has not yet been completely resolved, but Bell thinks a solution is not far away now, though he accepts there’s still work to be done.

“There’s much more development to go on,” he says. “But what will happen is that it will accelerate. People will suddenly see that there’s a market opportunity where they can do something, then people will innovate very quickly.

“They have limited capabilities at the moment, but it’s a bit like voice recognition. If you think now how quickly it’s moving on. In the past, you used to do all the voice recognition on the local device, and it was never brilliant, whether it was your PC or something else.

“What’s happening now, with all of the big tech companies, is that all of their speech recognition applications are running in the cloud. So that enables developers to do their processing in the cloud, where they can do a huge amount of machine learning and improve speech recognition.

“If I have a robot I can talk to, and if I keep talking to it, the speech recognition and the comprehension will just get better and better. The hardware will keep talking to the cloud and understand what it needs to say and do.”

Bell points out that a typical mobile phone was the size of brick in the past, and could not do a fraction of what much smaller smartphones of today can do. Speech recognition and natural language processing has also made progress at a similar rate.

And although it’s mainly a software challenge, such progress will have implications for hardware such as robots and home automation devices, which Bell believes will be big growth driver for the robotics and automation industry as a whole.

“I think one area we’ll see grow much quicker is service robots,” says Bell. “Service robots can be in every sector, from hospitality to catering, restaurants and hotels and so on.

“We did some work with a company that makes hospitality robots which they deployed at hotels in China. They adopted Ubuntu for those service robots.

“Those are new areas that are creating new marketplaces for devices that never existed before.

“In terms of service robots, I feel that it’s just three to five years before you start to see big adoption of them.”

Bell highlights the software-oriented advances made by tech giants such as Google, Apple, IBM and others in the area of natural language processing. And he notes the increasing number of computing hardware platforms – of tiny dimensions – which enable much more of the processing to be be done locally on the machine.

“A really great example is the Intel Euclid,” says Bell. “It’s a device with a RealSense camera built in, and then a development kit to enable people to build a robot. It enables you to prototype robots.

“So it is really great news in both consumer and industrial applications.

“This is a really great example of a product that ships with Ubuntu and ROS. And with the RealSense camera installed and the support, the developer can start building robot-based applications straight out of the box.”

The global phenomenon that is the Apple iPhone may never be matched by anyone else, but there’s plenty of space and commercial opportunity for innovations and inventions outside of the Apple world.

In fact, in some accounting periods, Samsung sells more phones than Apple.

And that’s just smartphones.

No one knows what devices will be invented using the new tools – the speech recognition, natural language processing, the Raspberry Pi minicomputers, the Arduino boards and Intel Euclids, to name just some basic tools available, and without even mentioning the augmented reality and virtual reality toolsets now available.

What can be done with a computing system which can understand everything you say? Pretty much anything, probably.

The only limitation might then be with the hardware.

But even that limitation has a time limit. It’s just that no one really knows what that time limit is. It probably depends on the ingenuity and vision of engineers, developers and entrepreneurs.


Share this story