What Would the World Miss If Perceptual Robotics Didn't Exist?

In the story of Perceptual Robotics, there was no single dramatic “Screw it, let’s build this” moment. They actually began at the University of Bristol, where the three co-founders, Kostas Karachalios, Dimitris Nikolaidis and Kevin Driscoll-Lind, were deep in research when they were approached by Det Norske Veritas (DNV) , the global risk management and quality assurance company.

The wind industry was facing a growing and largely unaddressed problem. Blade damages were increasing year on year, but inspections were still carried out using handwritten notes, fragmented reports, and manual analysis. There was simply not enough structured, high-quality data to identify patterns, anticipate failures, or make informed decisions at scale.

If blades could be inspected consistently and repeatedly, machine learning could finally be used to detect trends, prioritize repairs, and improve turbine performance. But it raised one fundamental question: how do you capture blade data accurately enough for AI to process in minutes rather than weeks?

The answer was autonomous drones.

In 2016, together with DNV, Perceptual Robotics carried out the first automated drone inspection of a wind turbine generator (WTG), with one goal: to transform inspections from slow, error-prone reporting exercises into structured, AI-driven asset intelligence.

The Road to Getting It Right

According to Lucia Roca Fernandez-Vizarra, Head of Strategic Marketing and Development at Perceptual Robotics, there wasn’t one big failure that stopped everything at Perceptual Robotics, but many small, frustrating ones. "Early on, the challenge wasn't flying drones. It was making them reliable, repeatable, and usable by non-drone experts", she said.

The automation had to reach a standard where a field technician with no prior drone knowledge could operate the system, every inspection would produce consistent results, and data would be ready to use within 48 hours, not weeks later.

That took quite a lot of iteration initially. Because “almost works” is useless in the field. But now customers know how reliable the systems are, and trust them to perform consistently.

Changing the Conversation Across the Industry

Before Perceptual Robotics, inspections were a cost to manage. Blades were smaller, repair costs lower, and the default mindset was to delay for as long as possible.

That conversation has shifted and now they’re increasingly seen as a strategic action. The question is no longer "Can we delay these inspections one more year?" It has become "How can we carry out our own on-demand drone inspections?"

The same shift has reached repair management. Where teams once manually reviewed hundreds of category 3 damages and relied on intuition to prioritize, the conversation is now strategic: "I will review these high economic impact damages and identify which adjacent ones to repair at the same time, so we are cost effective."

The Technical Innovation That Makes It All Work

Achieving consistent, perpendicular imaging of blades, regardless of turbine position or conditions, and the ability to continue a mission exactly where it left off even if the blades had moved, took months of refinement and ultimately became a patent.

This is what allows anyone to fly by simply pressing Play, produces high-quality comparable datasets, and delivers AI-driven analysis that is repeatable across sites, countries and teams.

"Consistency is everything," says Kevin (CTO). "If two turbines are inspected and processed differently, you cannot compare them. If you cannot compare them, you cannot prioritize repairs properly. And that is where costs quietly spiral".

Perceptual Robotics has spent considerable time refining its approach, including comparing human analysis with and without AI, to deliver consistency across every fleet they work with.

The Use Case Nobody Expected

When the company developed Eve, their portable drone solution, the original vision was a flexible tool for occasional, on-demand inspections, ideal for operators needing quick access to blade data without a large campaign.

That changed when a client in India shared a result that stopped everyone in their tracks. "With Eve, we reached up to 10 wtg/day. It was a goosebumps moment for all of us here."

The team had always associated high-volume campaigns with Dot, capable of inspecting up to 20 turbines per day. Eve was meant for lighter needs. Instead, it became a highly effective productivity tool for fast-moving field operations, opening use cases nobody had originally anticipated.

But the real innovation is the end-to-end workflow built around the drone. Capturing images, ingesting data from external sources, analyzing damages, prioritizing and managing repairs, all within a single platform. One that is in permanent evolution, shaped daily by customer feedback, from adding a button to helping operators forecast the evolution of unrepaired damages.

Real-World Impact

The biggest savings come from prevention. In one case of wind turbine inspection, a 30 cm crack doubled in size within weeks. Caught early through frequent in-house inspections, it was acted upon before becoming a serious structural issue. In Scandinavia, this year, two significant cracks were spotted just as a drone landed and images were being reviewed, allowing the team to stop the affected turbines immediately and avoid escalated damage, extended downtime, and significantly higher repair costs.

Inspecting the ‘As Somozas’ wind farm in Spain had previously taken weeks with too much wind for rope access, too little for ground-based cameras, and technicians blocked from other work for days. Perceptual Robotics completed the same task in two days, inspecting 40 turbines with two drones. The results arrived in time to carry out repairs that same summer, rather than allowing damages to evolve for a full year.

Built for Extreme Conditions

One of the most demanding deployments the system has undertaken was in the Faroe Islands, a remote location, harsh coastal conditions, and complex logistics including transporting drone batteries separately. Still, 24 turbines were inspected in four days.

The system has since been deployed across multiple continents, from working through extreme heat waves in Greece and Spain, where turbines were stopping due to their own temperature limits, to frozen mornings in France and Scandinavia, to offshore environments. Beyond convenience, it is about enabling inspections where they were previously impractical or simply could not be done.

This is why the company's advice to someone starting in this space is not to underestimate how hard “simple” is. Making something usable in the real world, by real teams, under real conditions is far more difficult than building a prototype.

What the Industry Would Lose

The myth that blade inspections are just a checkbox exercise no longer holds. They are one of the most powerful levers for increasing energy production, reducing costs, extending asset lifetime, and reducing financial risk.

Without Perceptual Robotics, the wind industry would still largely rely on manual flights requiring three turbine stops, outsourced inspections with delayed insights, fragmented data spread across different systems and suppliers, and a limited ability to compare assets consistently across an entire fleet.

Most critically, asset owners would still have no real ownership of their inspection data and the processes involved would remain slow, manual, and prone to both errors and accidents.

The real shift Perceptual Robotics brings is about control. It’s the ability to transform blade inspections into a consistent, scalable, data-driven decision system that is designed for continuous improvement and owned by the customer.

Did you know? Every software release at Perceptual Robotics is named after a cheese, from Gouda to Idiazabal, Parlick to Halloumi. They are, by their own admission, running out of options. As a closing note, they welcome suggestions.