How Iris Automation’s Alex Harmsen is building trust in drones
This interview is with Alexander Harmsen of Iris Automation which enables safer drone operation through intelligent collision avoidance. Since this interview was published, Iris Automation has raised an additional $1.5 million.
The Drone Series explores the drone industry via discussions with CEOs of leading companies, giving investors and industry analysts the opportunity to hear directly from emerging technology leaders. The series is lead by CEO Grant Canary of DroneSeed which is a drone company positioning to automate and dominate the forestry services vertical.
GRANT: What is Iris Automation up to?
ALEX: So far, a lot of the focus of drones has been in the military or in consumer space (photography, close inspection and toys — all line-of-sight use cases). The big problem with industrial drones is trust when they go beyond line of sight. Military drones are in a controlled airspace and photography drones are fairly close by. Industrial drones must fly completely automated missions, hundreds of kilometers long, over terrain, perhaps fairly close to the ground, perhaps much further up. In any case, they’re flying in uncontrolled national airspace. Right now, they rely 100% on the GPS systems and existing topo maps. Iris Automation is building a completely independent, real-time situational awareness system in a tiny box that drone manufacturers can integrate into their drones on-board. We can insure that the drones are collision-free and away from static objects, the ground, trees, power lines, or anything moving in the sky.
Regulators, insurance companies and governments are scared of drones hitting manned aircraft. Part of our complete situational awareness solution is being able to see anything else that moves, track it, find its trajectory, and make sure it’s not on a collision course. If it is on a collision course, then it can instruct the autopilot to take avoidant action.
GRANT: Let’s talk about your amazing background. You started out with Matternet. You’ve done Engineers Without Borders in Canada. You’ve worked at JPL. How did it all lead to the formation of Iris?
ALEX: When I first got my pilot license six years ago, I’ve been fascinatedby aerospace, especially satellite design. I worked for MDA and Airbus developing experiments for the International Space Station. I realized that space is very slow, and we could work on the same aerospace problems within aerial robotics.
When I was with Matternet, we flew drones in developing countries like Bhutan, doing vaccine and blood sample delivery in the Himalayas. Incredible. Then I moved onto NASA to continue working on computer vision robotics applications on adrone project for Mars called Mars Helicopter, which was full of unique and wonderful challenges.
Then, all my experiences clicked: the pilot license, space industry background, building reliable industrial-grade systems, computer vision expertise from NASA and the startup experience from Matternet. It all came together to form Iris Automation.
GRANT: What’s different about being a hardware startup versus a software startup?
ALEX: Actually, we consider ourselves a software startup. The core piece of our technology is the computer vision algorithms and how they interact as a world. It doesn’t really matter who provides the video stream or hardware. Right now, we’re packaging the Iris in a hardware unit with the camera and the processor in one piece to make it easy to integrate and sell for the first couple years of the product. But we imagine moving to software licenses over time.
Testing can burden a hardware startup because of the cost involved in actually building mechanical and electrical systems. It’s much harder to change hardware than a couple lines of code. On the other hand, it makes the demonstrations a lot more interesting. If you have a physical piece of hardware to show off, investors really like that. Any curious human wants to see hardware (especially drones) actually flying around.
GRANT Let me shift a bit and talk about drones as a much more general category. We talked to Jonathan Evans, who has talked about networking effects of being able to connect people, whether it’s Uber, whether it’s drones, whether it’s the internet. We talked to David Pittman at Converge, and he was talking about the democratization of flight. These are people who are very big on drones- equating them to the personal computer, as far as that level of revolution or that level of technological advancement. In your opinion, how revolutionary are drones?
ALEX: Drones are part of this growing Internet of Things movement. The personal computer was such a huge change in how we did business and how we lived our lives. Now we are multi-screen, with all these connected devices, so the computer just becomes one of the devices in our world. The PC was our first multi-purpose machine, but now the cost and availability of technology allow us to have more than one portal to the world. When we talk about connected devices, or the Internet of things, drones are a key piece of that because it allows us to get sensors in places where we normally wouldn’t get them.
So many current functions and jobs in the natural resources sector are way too expensive, inaccessible, or dangerous. Drones allow us to be able to cut into those markets. Even more than drones, I’m really excited about autonomous vehicles in general, whether it’s autonomous shipping or autonomous cars on the road, like aerial vehicles or mining trucks. When we no longer need human resources, autonomous vehicles can increase scale, speed, reliability and safety.
This going to be a huge shift in how we do farming, mining exploration, package delivery, movement, transportation, even planting trees.
GRANT: What do you say to someone who’s a skeptic of that position? One possible line of thought there is that it’s impossible to code like a human brain thinks. So, there will always need to be humans involved in the loop? What do you say to someone who’s a skeptic and makes that argument?
ALEX: I think there are two lines of thinking here. One line is: the initial tasks that we give these robots to do are going to be very simple and repetitive, like planting the exact same tree millions of times or flying over a farm every Monday morning. The drone is always flying the same map, which is why Iris’ situational awareness is important — in case something changes and we need to react to it.
We will encounter certain situations where human problem solving would be needed, but that’s not the case for most of the things that drones will do. Even self-driving car have fairly prescribed problems. The issues that do come up, humans wouldn’t be able to deal with either — that’s why there are so many car crashes.
The other line of thinking is that artificial intelligence and machine learning is actually making machines very smart. Most people don’t realize how intelligent these systems actually are right now. Deep learning and neural nets are already matching human intelligence in several areas, such as playing chess and facial recognition. I just read this paper the other day that was talking about an artificial intelligence network that was able to diagnose patients to a more reliable level than doctors. While it may seem very far away from matching human intelligence, there are so many different areas where artificial intelligence is already there.
GRANT: Where do you see the most opportunity in drones?
ALEX: A lot of the really exciting things that I would want to see in the drone space involve more trust and reliability in the systems. I like to think that that’s exactly the problem that we’re working on at Iris Automation. I want to see a drone simply take off from a farmer’s field every single morning, 6 a.m., survey crops, then come back, land in a box and send the farmer a report automatically. But we don’t yet trust it to do that. I want to see sidewalk package delivery, and though a number of companies that are working on that, there are still trust issues. We’re scared it will run into things or get stuck. I want one operator to launch a thousand drones off a truck to do eight-hour mining exploration flights over a huge area and report back with a map of the best ore deposits in that area. But we’re not quite at that level of reliability…yet. We’re scared they’re going to hit stuff, that the GPS is going to drop out, or that we need these human observers to be in the loop. Hopefully, Iris will build up that trust level and be the unlocking piece for the industrial drone market.
Have a question you want to ask, send it to email@example.com.
See other interviews in the series here:
From Kosovo to Sand Hill Road: A discussion with Skyward CEO Jonathan Evans
The Democratization of Flight by Drones
Drones and Simplifying the Future
Why Drones are Taking Off in Unexpected Industries