Join the leaders July 26-28 for Transform’s AI & Edge Week. Hear high-level leaders discuss topics around AL/ML technology, conversational AI, IVA, NLP, Edge, and more. Book your free pass now!
[Editor’s note: American Robotics is a commercial developer of automated drone systems.]
Drones have been in the news a lot for two decades now. In many ways, this attention was warranted. Military drones have changed the way we fight wars. Consumer drones have changed the way we film the world. For the commercial market, however, drones have largely been a false start. In 2013, the Association for Unmanned Vehicle Systems International (AUVSI) predicted an $82 billion market by 2025. In 2016, PwC predicted $127 billion in the “near future.” But we are not yet close to these projections. Why is that?
Let’s start with the primary purpose of drones in a commercial setting: data collection and analysis. The drone itself is a means to an end – a flying camera from which to get a unique aerial perspective of the assets to be inspected and analyzed, whether it’s a pipeline, a fleet of gravel storage or vineyard. Accordingly, drones in this context fall under “remote sensing”.
In the world of remote sensing, drones are not the only players. There are high orbit satellites, low orbit satellites, airplanes, helicopters and hot air balloons. What do drones have that other remote sensing methods don’t? The first thing is: image resolution.
What does “high resolution” really mean?
One product’s high resolution is another product’s low resolution.
Image resolution, or more precisely Ground Sampling Distance (GSD) in this case, is the product of two main factors: (1) the power of your imaging sensor and (2) the proximity of the object you are imaging. Since drones typically fly very low to the ground (50 to 400 feet AGL), the ability to collect higher image resolutions than aircraft or satellites operating at higher altitudes is significant. Eventually you run into physics, optics, and economy issues, and the only way to get a better picture is to get closer to the object. To quantify this:
- “High resolution” for a drone operating at 50ft AGL with a 60MP camera is about 1mm/pixel.
- “High resolution” for a manned aircraft servicelike the defunct Terravion, was 10 cm/pixel.
- “High resolution” for a satellite service in low orbit, like Planet Labs, is 50 cm/pixel.
In other words, drones can provide up to 500 times the image resolution of the best satellite solutions.
The power of high resolution
Why is this important? It turns out that there is a very direct and powerful correlation between image resolution and potential value. As the computer saying goes: “waste in, waste out”. The quality and scope of machine vision-based analysis opportunities are exponentially superior to the resolutions a drone can provide compared to other methods.
A satellite might be able to tell you how many well pads are in Texas, but a drone can tell you exactly where and how the equipment on those pads is leaking. A piloted aircraft might be able to tell you which part of your cornfield is stressed, but a drone can tell you which pest or disease is causing it. In other words, if you want solve a crack, bug, weed, leak or similar small anomaly, you need the right image resolution do this.
Bringing artificial intelligence into the equation
Once the correct image resolution is achieved, now we can start training neural networks (NN) and other machine learning (ML) algorithms to learn about these anomalies, detect them, alert them and potentially even predict them.
Now our software can learn to tell the difference between an oil spill and a shadow, accurately calculate the volume of a stockpile, or measure a slight tilt in a train track that could cause a derailment.
American Robotics estimates that more than 10 million industrial asset sites worldwide use automated drone-in-a-box (DIB) systems, collecting and analyzing more than 20 GB per day per drone. In the United States alone, there are over 900,000 oil and gas well pads, 500,000 miles of pipelines, 60,000 electrical substations, and 140,000 miles of railroad tracks, all of which require constant monitoring to ensure the safety and productivity.
Therefore, the magnitude of this opportunity is actually difficult to quantify. What does it mean to fully digitize the world’s physical assets every day, across all critical industries? What does it mean if we can start applying modern AI to petabytes of ultra-high resolution data that has never existed before? What efficiencies are unlocked if you can detect every leak, crack and damaged area in near real time? Whatever the answer, I’d bet the $82 billion and $127 billion figures estimated by AUVSI and PwC are actually low.
So: if the opportunity is so big and clear, why haven’t these market forecasts come true yet? Enter the second important ability unlocked by autonomy: imaging frequency.
What does “high frequency” really mean?
Useful imaging frequency rate is 10x or more than people originally thought.
The biggest difference in performance between autonomous and piloted drone systems is the frequency of data capture, processing and analysis. For 90% of commercial drone use cases, a drone must repeatedly and continuously fly over the same terrain, day after day, year after year, to be of value. This is the case for agricultural fields, oil pipelines, solar panel farms, nuclear power plants, perimeter security, mines, rail yards and tank farms. When examining the complete operating loop, from setup to data being processed and analyzed, it is clear that manually operating a drone is much more than a full-time job. And at an average of $150/hour per drone operator, it’s clear that a full-time operational load on all assets just isn’t feasible for most customers, use cases, and markets. .
This is the main reason why all predictions regarding the commercial drone industry have, so far, been delayed. Imaging an asset with a drone once or twice a year is of little or no value in most use cases. For some reason this frequency requirement has been neglected, and until recently [subscription required]autonomous operations that would allow high-frequency drone inspections have been banned by most federal governments around the world.
With a fully automated drone-in-a-box system, humans on the ground (pilots and spotters) have been taken out of the equation, and the economy has completely changed as a result. DIB technology enables constant operation, multiple times a day, at less than a tenth of the cost of a manually operated drone service.
This increased frequency not only leads to cost savings but, more importantly, the ability to track issues when and where they occur and properly train the AI models to do so autonomously. Since you don’t know when and where a methane leak or rail tie crack will occur, the only option is to scan each asset as frequently as possible. And if you’re collecting that much data, you better build software to help filter out key information for end users.
Link this to real-world applications today
Autonomous drone technology represents a revolutionary capability to digitize and analyze the physical world, improving the efficiency and sustainability of our world’s critical infrastructure.
And luckily we have ultimately went from theoretical to operational. After 20 long years of flying drones in the Gartner Hype Cycle, the “productivity plateau” is reaching its peak.
In January 2021, American Robotics became the first FAA-approved company to operate a drone system beyond line-of-sight (BVLOS) without a human on the ground, a milestone unlocking the first truly autonomous operations. In May 2022, this approval was expanded to include 10 total sites in eight US states, signaling a clear path to national scale.
More importantly, AI software now has a convenient mechanism to flourish and grow. Companies like Stockpile Reports use automated drone technology for daily stock volumetrics and inventory monitoring. The Ardenna Rail-Inspector software can now evolve on the whole railway infrastructure of our country.
AI software companies like Dynam.AI have a new market for their technology and services. And customers like Chevron and ConocoPhillips are looking to the near future where methane emissions and oil leaks are dramatically reduced through daily inspections from autonomous drone systems.
My recommendation: look not at the smartphone, but at the oil fields, rail yards, tank farms and farms for the next data and AI revolution. He may not have the same pomp and circumstance as the “metaverse”, but the industrial the metaverse could be more impactful.
Reese Mozer is co-founder and CEO of American Robotics.
Welcome to the VentureBeat community!
DataDecisionMakers is where experts, including data technicians, can share data insights and innovations.
If you want to learn more about cutting-edge insights and up-to-date information, best practices, and the future of data and data technology, join us at DataDecisionMakers.
You might even consider writing your own article!
Learn more about DataDecisionMakers