It almost doesn’t need to be said just how important the maintenance and repair of infrastructure is for the successful operation of towns, cities and countries. But the sheer scale of infrastructure – a single city can have hundreds of bridges, wastewater pipes, stoplights, roads, sidewalks, streetlights, curbs and everything else that keeps a city moving – can seem insurmountable to keep up with. Tackling that massive need in a cost-efficient, yet detailed way is where drones have risen to the challenge.
At Commerical UAV Expo, a panel of experts discussed not only this opportunity, but the future needs, and challenges of putting drones to work on infrastructure inspection. The panel featured Jason San Souci, Enterprise Drone Architect at Cognizant; Russ Ellis from gNext Labs, LLC; Joshua Sexton of Stantec; Gagan Kanwar from Skydio; and Tanner Cook of Raptor Maps. The session was moderated by Nitin Gupta, CEO of Flytebase (see our recent interview with Gupta here), who lead a spirited panel discussion after a series of initial presentations about the panelists’ work. In that discussion, many of the biggest questions facing the industry came up.
What can we really expect from AI when it comes to infrastructure inspection?
Any time a session has “advances” or “future” in the title, you know that artificial intelligence is going to become a topic of discussion. There’s often a bit of a chasm between what the media portrays AI as being capable of and the reality of its use. However, in the drone space, there have been a few places in which AI and/or automation have made a big difference. There are now several software platforms that have some kind of functionality that can automatically detect changes and others that can identify defects in things like concrete or asphalt.
Ellis, from gNext labs, outlined several use cases where they are applying AI to specific types of assets. “Our first catalog was concrete and asphalt, we’ve now moved into cell towers and communication towers, where we’re automatically collecting the equipment and filling out an inventory database associated directly with that tower and all the pieces around it. Next will be the utility substations and infrastructure around those types of structures.”
“Wayne Gretzky was famous for saying, ‘I don’t skate to where the puck is now, I skate to where the puck is going.’ And when we think of where that puck is going for AI – it’s really going to predictive analysis and predictive maintenance suggestions.”
When you are able to have several, repeated snapshots of something, there will be ways in which to extrapolate that to see where those changes are going. This could inform safety implications, but also how municipalities spend limited budges for repairs and remediation, and how those dollars can make sure that infrastructure remains safe and stable.
Where to “drone-in-a-box” systems fit in infrastructure inspection?
One of the primary advantages of using drones for infrastructure inspections is the ability to conduct high-frequency, repeated flights. This can illustrate where problems grow over time – for example, is the crack on a bridge deck growing or is it stable? – but also can be used to create digital twins that are updated as new information comes in. But the repetition of these flights and whether it’s a fit for a “drone-in-a-box” system really comes down to frequency and reach. For assets that need continuous monitoring, or even for emerging situations, having a drone already on site (or nearby) can be game-changing.
As Kanwar explained, “If you need to inspect something once a year, maybe once every six months, once a quarter, maybe you don’t need a fixed site there,” he offered.
“Unless you use that location to inspect multiple assets around there and use [a docked drone] a staging ramp to reach those assets.” Panelists also offered that hard-to-reach places could be a good spot for some kind of docking system, especially when the trip to bring the drone out to a site is more arduous than the flying itself.
What’s still ahead for drone inspections that would make this work more effective?
This shift towards greater autonomy has the potential to unlock new possibilities, from streamlining data collection to expanding the reach of infrastructure inspections. Sexton shared that he thinks that greater autonomy and functionality of sensor payloads are still on the horizon.
“We’ll see better integration of lidar, photogrammetry and merging those two datasets together to get a holistic picture of our assets. We’ll be moving more into GPS-denied environments, so we’ll be moving from all outdoor GPS environments and going into indoor environments.”
Circling back to AI, Kanwar also pointed out that one of the technologies enabling those flights in GPS-denied environments or other challenging spaces is to use AI to aid in navigation. Skydio drones have been utilizing navigation cameras (three pointing up, three pointing down) and runs that information through AI programs on the drone “to build a complete 3D picture of the world around the drone,” says Kanwar.
“It’s almost like the human is right there – they can see in day or night what’s in front of the drone, what’s below it, what’s above it, what’s behind it. And they can use that to navigate. We see autonomy and AI as being really critical to powering things like beyond visual line of slight flights, and that’s where the industry is trying to move to.”
With all this data that’s being collected, what about standards?
While many of these innovations are making the process of data capture easier – enabling more data to be collected than ever before – there is still a massive challenge relating to how to get that data to work together, or to what standard those data must meet in order to be useful, or able to be ingested into other systems or merged with other datasets.
“There are no standards related to most of the data we’re collecting, or there are very very few,” said San Souci, “We’re accelerating with data capture very quickly, we’re automating – but no one is having a conversation about metadata.”
“It’s a massive opportunity because we’re taking a variety of different systems and capturing a variety of different modalities of data and putting these things together. You can give it to your GIS guy and say, ‘Good luck!’ or you can have a full system where you’ve built it to be able to ingest, process, disseminate, collate, exploit everything within that system build on that standard base.”
That challenge is something that San Souci views as a huge opportunity for the industry, even though we haven’t really gotten there yet, beyond a few bits and pieces and small conversations.
“A lot of folks are focusing on, ‘I have this equipment, I have these sensors, what can I go do with it?’ whereas the better conversation is – here is what the industry needs, here’s what our customers are actually looking for, and work backwards from there.”
What’s next?
The future of infrastructure inspections is being shaped by a convergence of drone platform and sensor technology, coupled with AI, and automation. As these technologies continue to advance, we can expect to see even greater efficiency, accuracy, and safety for infrastructure inspection, helping to process. However, challenges such as data standardization and the need for further technological advancements in autonomy and sensor integration remain. By addressing these challenges and focusing on the needs of the industry and its customers, the drone industry can move us towards a more comprehensive plan for our growing infrastructure needs.