Last month we had the pleasure of interviewing Anil Nanduri, Vice President in the New Technology Group and General Manager of the Drone Group at Intel.
In that conversation we focused on Intel’s Shooting Star light show drones, which have been making headlines for the last year, most recently for their use during the promotion of Wonder Woman’s release on Blu-ray.
We ran out of time during that interview, and there was still the whole world of A.I., industrial / commercial applications, and the Falcon 8+ to discuss, so we scheduled a follow up to cover these topics.
Here we go—
UAV Coach: Our team was impressed by the Intel demo at InterDrone, in which the Falcon 8+ was used to perform an automated inspection, and then deliver actionable data by comparing that inspection to a prior one. Can you tell us more about the automation used in that inspection, and what’s next for the Falcon 8+?
Here’s a short clip of the demo referenced above
Anil Nanduri: In automated scenarios, like the inspection done in that demo, the pilot is there to monitor and maintain the machine, as well as stay compliant with regulations and safety protocols, as opposed to actually doing the inspection or surveying work himself. From an automation standpoint, what we were doing at InterDrone is showing the capabilities of the platform.
Whether you’re looking at construction, surveying, mapping, or inspections, automation can be incredibly useful in industrial scenarios, because all of these applications come with a number of factors that have to be considered.
To be able to complete an inspection, you have to know the protocol of the inspection object, whether it’s a bridge, or a building, or a land survey. Automation can take care of both repeatability and safety, which is why the Falcon 8+ was built with triple backups, so there’s no single point where electrical failures or rotor failures can occur.
If you’re trying to inspect the facade of a building, like we were doing in the demo, the biggest challenge is often just to hold a position so that you keep the same distance from the object being inspected. From a distance, when you’re looking towards the building or surface to be inspected, our human eye cannot understand the perception of depth, especially as the drone gets farther away from the object.
Automation gets rid of this concern. If the system can manage the distance automatically, then the pilot doesn’t need to worry about it. Add to this our obstacle avoidance, and you have a platform that can fly safely.
The other critical factor of course is really solid mission planning—that is the last key ingredient to fully unlock the potential of automation.
UAV Coach: Where are you seeing the Falcon 8+ being used? What industries are using it the most, and how are they using it?
Anil Nanduri: The Falcon 8+ is primarily being used in surveying, mapping, inspections, and precision agriculture.
One of our biggest partners is Topcon Positioning Systems, a geospatial company that makes positioning equipment, which uses the 8+ for construction surveying.
For inspections, some of the most common use cases for the 8+ are inspections of assets like wind mills, offshore oil rigs, and other oil and gas scenarios. We’ve also seen it used for airplane and air bus inspections.
Another use case we’re seeing more and more of for the 8+ is precision ag, especially with seed providers. They need very, very accurate information, including flight outputs with high resolution imagery, and we can provide that for them.
Because the 8+ is so sound from a safety perspective, and also because it can fly in high wind and other harsh environmental conditions, you really see it shine in some of these more rugged conditions.
A srvey of the Cathedral of Halberstadt done with the Falcon 8+
UAV Coach: Given that the Falcon 8+ is such a robust, high end platform, do you have a special training program to help companies learn how to use it?
Anil Nanduri: Yes, we have training programs. Usually we work with operators until they’re very proficient.
We have lots of complicated features, such as electromagnetic interference handling, so it’s important to train the pilot in the ones they’re going to need.
Sometimes we provide this training directly, and sometimes it’s provided by a distributor.
UAV Coach: Can you describe some of the complications that come with learning how to fly the 8+ as opposed to other high end drones built for industrial scenarios?
Anil Nanduri: Flying isn’t really the complication—flying itself is amazingly intuitive with the Falcon 8+, and it’s no different depending on the mode of the remote.
It’s more making sure that the users can handle the advanced features and capabilities of the system in scenarios where they may need them, and also making sure they know how to capture and process the data that’s important to them.
For example, if you’re flying near a power line and need to use the electromagnetic interference handling, making sure you know how to do that. What should I do if there is interference? And so on.
Watch the video to learn more about Intel’s Falcon 8+
UAV Coach: We’ve seen the Falcon 8+ priced anywhere from $25,000 to $40,000. Is there a base price point, or how do you price the 8+?
Anil Nanduri: We actually don’t publish pricing from our side, but our partners and distributors do. Pricing can vary greatly depending on the unique customer’s request, since there is such a wide range of capabilities for the platform.
You might have an inspection pay load, which has a floor camera and, say, a 20 megapixel RGB camera. So depending on how you’re configuring it, your cost will vary accordingly.
UAV Coach: During his keynote Intel CEO Brian Krzanich spoke about how the development of drone technology is pushing forward the development of A.I. Can you describe how this is happening for our readers?
Anil Nanduri: Regarding A.I. the key question is, How do we automate data capture and then also automate the production of actionable insights from that data?
If you think of a drone as a tool that helps perform certain tasks, the customer really doesn’t care about how you capture the information. What the customers wants are the insights produced from the data the drone captures.
So if you’re doing an oil rig inspection, the customer wants to identify any faults, defects, or cracks. He’s not looking for a thousand images from a drone, just the insights about his assets.
The same goes with inspecting a bridge, where you’re going to be looking for corrosion and cracks. Are there any deviations or any deformities, and things like that. Or if you’re looking at surveying, the customer will want to know the terrain model, and he wants it to a centimeter level of accuracy, so he can automate the construction flow and feed that back to the architects and designers—again, not the images, but the output.
We’ve proven that a drone can be a very effective tool. It’s more efficient, less expensive, and safer for all of these types of use cases we’ve been discussing. But the next step is, how do I make inferences from the data a drone captures? If it takes 50 people to analyze the data, that not going to help anyone.
You need to be able to automate this process, so you can get the inference and the actionable insights automatically. Once you have the data, wouldn’t it be great if your computer automatically told you, “Hey I detected 10 cracks on this bridge. Here’s the precise location of each one, and heres what they looked like the last 10 times you did that inspection.”
That’s where A.I. comes in.
UAV Coach: How do you see this technology being used five years from now?
Anil Nanduri: I see a day where drones are actually flying automated patterns—not just one of them, but many of them—and they are both capturing information and analyzing it.
For inspections, I can see scenarios where drones are finding corrosion and cracks, flagging them, and then delivering the information through a text message that sends the data file to the right person, and lets them know when something needs to be taken care of immediately.
This isn’t just for commercial inspections either. This same technology could be applied for search and rescue missions where you have a missing person, and it could be a fleet of drones doing this kind of work.
I think these capabilities will evolve rapidly. If you look at where we were five years ago to where we are today, you see drones getting much, much smarter.
And this technology actually already exists inside other industries. Machine learning is being used with data from our smart phones, our pictures. If you start using this same technology for crack detection, or rust detection, and you’re able to develop that domain knowledge, that is the path forward.
First we have to get a lot of data into the cloud, and as it gets analyzed it will be interpreted and classified, and the ability to classify it will continue to get better. At first you’ll need a human inspector to review the data, but as the system gets better it will become more reliable, and eventually it will become better than a person, and will even be able to predict changes over time, once it has enough information to see what has happened in the past and project the likelihood of degradation in the future.
This is where the systems we have and AI will come in. The goal would be for a whole lot of data to come in, and then be automatically classified into sections that say, okay I see seven potential problems. Three are critical, two are risks, and the last two seem to be okay for now.
The technology exists, it just needs to be adapted to this ecosystem and to these applications. I’m confident we’ll get there, and I think it will happen soon.