Iris Automation has been in our peripheral vision for a few years now. So we thought it would make sense to take a look at what the company does, as its products offer the ability for a drone to detect (and avoid) other crewed aircraft.
Iris Automation is all about air safety. Specifically, ensuring that a drone you might be operating can detect and avoid crewed aircraft. In a nutshell, that’s what Iris Automation does, which is a very important thing as the industry evolves toward a future where Beyond Visual Line of Sight flights are the norm, rather than the exception.
Let’s look at how the company does it.
An eye on the sky
Drones are often referred to as an “eye in the sky.” But drones are generally built to use those eyes (regardless of what type of sensor we’re talking about) to capture data on a mission. The eyes of a drone, with rare exceptions, are generally focused on ground-based objects or assets. They’re not looking around for other things in the sky.
That’s where Iris Automation comes in. It will do the looking around on your behalf. And, if it detects an aircraft that could pose a problem, take evasive action.
Automating the process
The company name is the first clue: It’s Iris (as in your eye) Automation. The drone field is heading inexorably in the direction of automated, smart flights. And a fundamental component of the Iris system – called Casia – is intended to be part of the solution. Why? Because, as CEO Jon Damush puts it (and we agree), humans can be a weak link in the system.
In our briefing with Jon, he pointed out there are lots of pieces of the puzzle already out there, each making contributions toward Unmanned Traffic Management. For example, pilots file flight plans. There are flight path altitude separations depending on the direction the crewed aircraft is flying. There are tools like Automatic Dependent Surveillance – Broadcast, or ADS-B (this is a transponder carried by many crewed aircraft that pings out ID and location information on a regular basis). Radar, too, can be part of the equation. But the weak link? That’s the human being.
“The reason I’m terrible (as a pilot),” explains Damush, is because “I’ve got to do a lot of other things. My view of the sky is occluded. My efficacy as a human to be the primary detect and avoid (mechanism) is really poor.”
But think what might happen if you could integrate an onboard camera and detection system, and connect it with the other pieces of the puzzle. “You now have the capacity for completely automated aircraft control,” says Damush.
A watchful AI
A more appropriate term, given that the system involves cameras, machine vision, and more, might be “A-Eye.” That’s because the system uses a camera (or multiple cameras in the 360° configuration) to watch the sky. CEO Damush says cameras were chosen because they’re light, relatively inexpensive, and are constantly sampling data.
“An optical system is high rate – 15 fps. That’s 15 samples per second, and that’s a hard rate to match with other modalities,” he says.
The Iris Automation home page outlines some of the key value propositions. We’re going to borrow this straight from there:
Proven Detect and Avoid: “Extensively tested with more than 16,000 real-world encounters and 50,000 encounters in simulation. Over 600 terabytes of flight data recorded.”
Comprehensive Compatibility: “Works with nearly all industrial drones and integrates with the most popular commercially available autopilot software.”
Completely Onboard: “Install once, fly forever with no ground-based infrastructure or visual observers required.”
What is the system?
Well, there are two varieties. One involves a single camera connected with the Casia unit:
Or, if you’re flying something bigger and are after 360° detection, you’ll be looking at a five- or six-camera system weighing about 2.2 kilograms (nearly five pounds). The lowest weight of the system, with a single camera, is 400 grams.
And what does that look like, mounted on a drone? Well, something like this, seen here integrated on a Saxon M14. Obviously, this is a system designed for larger drones and not units the recreational user or hobbyist is likely to be flying (unless you’re a really serious hobbyist).
Getting the picture
You’re probably starting to get the picture. The Casia system is constantly scanning the skies and trying to detect what’s out there and if there’s a potential conflict. Because it has AI and machine vision baked in, it can even identify what’s out there. Is it a Piper? A Cessna? A helicopter?
As you can likely imagine, there’s a lot of computing going on behind the scenes. The system has to know precisely where its cameras are in time and space, along with the speed and trajectory of the drone. Plus it has to identify what’s out there, what it is, and where it’s headed.
And here’s the real secret sauce: Because the system knows, for example, how big a Cessna is, it can also quickly calculate how far away it must be (given its known dimensions). Here’s Jon Damush:
“Once we have a candidate we run it through AI – those classifiers determine if it’s a helicopter, a small plane, a large plane, a hot-air balloon, etc. Once we have that, we know the real-world size. And we can determine range.”
That’s a lot of information to capture and process on the fly. And then what? Well, if it detects there is a potential conflict, it will automatically perform an evasive maneuver to get the drone in question out of harm’s way.
What kind of maneuver?
The default would be simply to send a command to descend. But that’s not always the case; Damush explains that drone manufacturers sometimes have their own preferred avoidance moves:
“The avoidance thing is a tricky one. Most of the manufacturers and operators want to take charge of the avoidance manoeuvre (such as) vertical vs. horizontal. For this category, typically the descent maneuver is the best maneuver.”
The BVLOS advantage
Obviously, a larger drone equipped with 360° vision is going to be far safer to deploy on a BVLOS flight than an operator relying solely on ADS-B signals or other pieces of the detection puzzle. But when you can start integrating this additional layer of safety, regulators are more likely to feel satisfied that there’s significant risk mitigation. That means BVLOS approval is likely to come more easily for a UAS with the Casia system on-board (though the operator, of course, still has to prove they know what they’re doing).
In fact, Casia-equipped aircraft have received BVLOS approvals in five countries and counting, and the company has collaborated with regulators on 14 different BVLOS test programs.
“We seek to achieve 100% probability of detection,” says Damush. “Today we’re at 95% probability of detection. You have a very good chance of detecting a collision threat from 1.2 kilometers away.”
Well, these aren’t exact figures, but a basic system with a single camera will set you back about $9,000. There’s an additional licensing fee per aircraft per month. And yes, there are customers who are flying these systems right now – it hasn’t all been testing.
One intriguing thing that popped up in the conversation: Could this system be ground-based, looking up in the sky for potential conflicts if your mission(s) don’t involve flying far afield? This could be particularly useful if you’re operating smaller drones that couldn’t handle the weight of the Casia, or you had a fleet of different craft operating within a limited mission zone.
The answer is yes; Damush says Iris has been running some tests for this use case.
CEO Jon Damush points out the company is but five years old, and that these are still “early days.” But there are already 50 customers, and regulators appear to like what they see. As more regulators see the system in action, more BVLOS approvals will come. As for the weight, Iris Automation is already looking at ways to produce the same results, or better, in a lighter package.
And the bottom line? The system, says Damush, works.
“Everybody’s looking for a silver bullet. And for us? We’re going to keep you from crashing into another airplane.”
And you can’t put a price on that.
FTC: DroneDJ is reader supported, we may earn income on affiliate links