Plenty of companies are using outdoor drones to inspect pipelines, power infrastructure, and other hard-to-reach or dangerous locations. But when those locations are indoors or require access to confined spaces, the challenges become different and more difficult.

Swiss startup Flyability, which makes aerial inspection indoor drones specifically designed for confined spaces, is leveraging its recent $11 million Series B funding round to develop a set of next-generation drone-mounted sensors for indoor industrial inspections. Through a new initiative called Flyability Labs, the company is recruiting partners to work with engineers to develop new sensors for industrial inspection indoor drones. While the company is already working with partners on gas, chemical, and radiation sensing, they are not limiting the ideas to those areas.

The company has more than 300 customers worldwide that use Flyability’s Elios drones for infrastructure, power generation, and public safety inspection scenarios. Robotics Business Review recently spoke with Patrick Thevoz, Flyability’s CEO and co-founder, about the company and the unique challenges for indoor inspection drones.

Q: Does Flyability sell drones directly to companies, or do you also provide services and/or work with integrators to get them into the hands of companies that need to do the inspections?

Thevoz: We do both. Our main business is mostly product sales, with a lot of training and onboarding until our end users can use the products. But we also work with a number of inspection companies that provide the service. We don’t have a large team of pilots doing missions all over the world – we have two to three pilots that help us with doing onboarding customer types of missions.

Q: How long does it typically take to train an end user in the piloting of these drones?

Thevoz: We have mission assessment metrics that we ship with our drones that really help understand how hard it is to do certain missions. In general, someone who is not a pilot, after one day of training, they can conduct a mission. Now obviously there are some cases, like with air draft, or you’re flying beyond the line of sight, or are in a very challenging dusty, dark environments – this requires a bit more practice.

Q: Is there a particular area where you’re selling these drones around the world?

Thevoz: North America is definitely our biggest market as of today. Canada is really big. Then it would be more industrialized nations, including China and European countries like Germany, U.K., and France.

Q: Most of your customers are using these are to conduct infrastructure inspection in small spaces, correct?

Thevoz: Not necessarily always small spaces, but confined spaces, which is not exactly the same. It’s more a matter of how dangerous the space is in terms of how you get inside it, but infrastructure inspection is definitely the biggest market. Number two would be public safety, and it’s all about visual inspection in those hard-to-reach places indoors.

Q: What do indoor environments have that make them different than using drones outside?

Thevoz: You do have a lot of different problems. First, you don’t have any GPS coverage obviously, which is a big challenge. Then you have a lot of obstacles. You have things everywhere, you have people, scaffolding, beams, internal structures, so there’s a lot of risk of collisions.

You’re not going very far usually indoors, you don’t cover kilometers. But a lot of times you are behind something, or you’re entering inside of a tank, or inside a chimney, stuff like that. Then you don’t have lights, it’s usually pitch dark. You have dust, and the environment is thus fairly different from outdoors. You often want to inspect things that are above you and not below you. When you fly outdoors, mostly you’re looking down — in our case most of the times you’re looking up.

Looking for sensor ideas

Q: Tell us about the program where you are looking for partners to provide ideas around the next generation of sensors. Can you explain what you’re hoping to achieve?

Thevoz: We’ve built our first product, and we’ve got about 300 customers around the world using it. Now we’re looking at going deeper into different verticals where we have activity and looking specifically what would be of interest to our existing customer base and prospective customers.

We’re doing visual inspection, we’re doing thermal inspection, and there’s a lot of interest around how can we detect radiation, how can we detect concentration of some gases. We’re trying to pull together the interest to shape different projects going forward. We’re really opening the discussion with people that have requirements for other types of sensing technologies beyond visual and thermal inspection.

Q: Does that mean that the types of sensors out there aren’t good enough? What is the current state of some of those other sensors? Do you need to build your own or are you just looking to partner up with someone that can then fit your drone?

Thevoz: There are a lot of sensors out there. In terms of engineering, everything is possible to integrate new things on our product. It’s a bit sensitive to add things to a flying machine. You need to do some work to ensure that you’re not compromising the flight characteristics.

The question is what exactly are we trying to solve? It’s actually more of a business question than a technical question, about building the right business case. For example, is there a scalable business case that requires a radiation sensor? So in nuclear power plants, perfect. But then what type of radiation, what is the use case? And really trying to dig deep into what the needs are, in detail, what is the user story so we can get the ball rolling with the right product. Right now there are so many opportunities and so many things that we’re pushing forward, that we need to establish the business case before we start doing development. That’s really what we’re trying to identify.

Q: How long do you anticipate this call for products, or assistance and partnerships? Do you expect to do this in the next two to three months?

Thevoz: On those particular topics, we have three different tracks, so today we have one about gas sensing, one about radiation sensing, and one about fully automated drones. On the sensor project we are probably going to keep the conversation going with customers until we have identified what we need to get to press the button and go with the deployment. It can stay open for a year if we don’t get to the bottom of it, or if we launch a product, then obviously we will, we will stop the call for projects.

Autonomous flight goals

Q: Talk a little bit about getting these drones to fly autonomously. Are you working on something where a drone can fly in and then map an area and then fly autonomously, where you don’t need a pilot, or the pilot has less to do?

Thevoz: Definitely, this is one of the big areas of work where most of our team is focused towards automating drones. It is highly important, and that was one of the first questions you asked about it. How long does it take to be efficient with our machines? We’re trying to bring this down to a point where you don’t even have to care about the drone itself and about how the data will be acquired. It will get acquired fully autonomously, and then it’s all about pushing this data into the right systems that the company is currently using for their infrastructure maintenance, and having access to this up-to-date data without having to care about the drone, piloting the drone, and avoiding obstacles along the way.

Automation is a huge enabler for the scale-up of this industry, from something that requires quite a bit of onboarding, of ensuring that your people are able to operate those machines. That scales much slower than what the technology could unleash if it was automated. So we have currently built our automated machines to be able to do those tasks.

But when we talk about automation, we talk about switching from something that is very broad in terms of usage. We go inside natural caves, inside glaciers. There are hundreds of different use cases, and now we’re really narrowing down to places where automation makes a big difference or a big change, and needs to be very focused on one particular task. It also means focusing towards more higher volume type of applications. When you’re operating in a nuclear power plant, or you’re delivering a half a million dollar savings in one single mission, you don’t care too much about [spending money on] an operator to do that. When you’re doing something that is in much higher volume, and where you’re not saving half a million dollars every time you’re flying, you’re probably looking at further reducing the cost and having something very repeatable. This is the direction that we’re taking for the future of indoor drone flights.

Q: This feels like if you were going to do kind of an automated inspection of a ventilation shaft for example, you would want the software to basically alert you whenever there’s an anomaly, versus just kind of having someone watch the video screen during the flight.

Thevoz: Exactly. The automation is upstream, from the data acquisition (having the flight pattern and having the drone go in the right places to acquire the right data), and downstream – what do I do with this data? How do I organize it and how do I analyze it? This is where we’re going – very industry specific and use-case specific, and we’re not talking about something very generic anymore.

Q: How long do you think that’s going to take either your company or the industry in general to get to that point?

Thevoz: This is a great question. We’re always looking at the outdoor drone market, where automation is a much easier task because you don’t have as many obstacles, and you have GPS. This is really made to use things like the senseFly eBee. It’s fully automated, you just shake the drones, throw it in the air, and it delivers you a 3D map of the area you defined on your computer beforehand. So now it’s fully automated. It took about five years probably from the first hardware to the moment where it was used as a seamless, fully automated experience.

In our indoor industry, where the conditions are more challenging, we are going to start looking at the first implementations as early as this year, and the big question is when does the scale-up really happen? Are we able to go from the first pilot projects to something that is really widespread inside the industry? We feel the maturity is there, according to the number of different large industries that are contacting us for those types of needs and have this in their roadmap. I think the time is now, and within three years I believe we’ll have numerous large scale implementations of automated devices. This is what we’re seeing right now with our manually flown drones.

Drones vs. crawlers vs. ground-bots

Q: Do you think there’s going to be a battle between the drones versus other kinds of inspection robots, either ground-based, or ones that can stick to a wall, or does it not really matter?

Thevoz: A lot depends on the environment. If you have a clean metal pipeline, it probably will not make any sense trying to fly inside an area like that, and a crawler would solve that use case very well. When you start having places with variable geometry, that are dirty, that have T junctions going on in all directions – at some point you really need to fly. It doesn’t make sense to fly unless you need to.

But in a lot of those indoor places, you’re working at height, you’re working in environments where a crawler cannot access. So this is probably a very complementary technology. The advantage of flying robots are definitely more versatility and bigger tolerance to unknowns in the environments. This originally triggered our work on flying robots, looking at catastrophes like Fukushima and looking at ground robots that couldn’t enter the place because of things like debris on the stairs, and the robots couldn’t overcome the debris. With flying, it’s just amazing at how easily you can access anywhere in a space. That’s something that is complementary to magnetic robots that can do an amazing job in a more structured and known environment.

Q: Are there any advances in the drone space on the technology side that have allowed you to ramp up the scale? Is it just that components are cheaper, or have there been sensor improvements? What are you seeing that is driving the scale for the company?

Thevoz: The commercial, non-military drone space emerged when a lot of different technologies came to maturity at the same time, whether it was batteries, motors or cameras. Right now there are a lot of opportunities driven by autonomous car developments in terms of sensors that can detect the environment, whether it’s radar, lidar, or other types of visual SLAM vendors. This is what we’re currently leveraging as we move forward in drones, and the huge investment in autonomous vehicles will drive the next generation of technical improvements. The money that is being deployed for autonomous cars is orders of magnitude ahead of what’s being done in robotics so far. So it’s definitely moving the sensor space fast forward.