Artificial Intelligence <> Drones

Drones have always been on the periphery of my working life.

Whether it be the early days back in 2017; when I was exposed to advanced drone concepts via the Air Force's SBIR program, during my zoo mapping days at Zoptiks, and now as I experiment with content creation & drone racing.

To be honest, I don't remember having to use drones when the industry didn't maintain all the 'helper' tools that modern drone operators have.

• Software tools such as DroneDeploy , Pix4D , ArduPilot , and SiteScan ( ArcGIS ).
• Hardware enhancements such as better cameras, neural chips, stabilized propellers, composite materials, and battery capacity.

Things have come a long way from my fragile Anafi which would maybe last me 15 minutes or be decimated in animal exhibits.

As my experience in working at the intersection of hardware & software grew (mostly through working with complex hospital bed data at Hillrom ) so did my awareness of how the complexity of data can be streamlined.

In fact, I'm shocked that there is yet to be an 'Apple' of drones where hardware and software are streamlined into one unified solution. DJI is getting there, but their software needs a lot of work.

Now let's fast forward to how artificial intelligence is going to change drones.

When Skydio released its first autonomous consumer drone, the R1, back in 2018, drone enthusiasts were amazed. Simply because we didn't need a drone operator. Peep the video from 2018 below:

Skip to 2023 and nearly every single aspect of drones is being affected by AI. Whether it be design and development, flight operation, in situ data processing or post-processing. As a result, no other hardware industry is primed for radical transformation than the drone industry.

Why is this change so prevalent?

1) Computer Vision & Cameras - The camera on the Anafi I had was 8mp, although it worked many commercial drones are hauling 48mp out of the box with advanced neural imaging to support video processing. If the camera can take in more information, there is more information to be used. In the case of AI, there's more data to feed the machine-learning models.

2) Realtime Data Processing - Advances in chips have made it possible to put neural engines inside of the drones themselves. This is incredible because it eliminates the need for data transfer and bandwidth problems. We don't need to export video files, analyze them and do all sorts of post-processing to create iterative cycles of development. Think of it as avoiding the hassle of moving a 100 GB file back and forth 3-4 times to achieve one task. Less 'labor' means faster iterations.

3) Generative Software Integrations - So much of the post-processing data wrangling is incredibly simplified by modern golems such as ChatGPT. I don't have to search for every tiny data reference, procedure, or 'how to' guide manually. It's spoon-fed.

• Need to clean up your 3D map in Pix4D? Used to take 16 hours.
• Need to identify points in the map? Pick a reference and it'll get 90%.
• Need aggregated pipelines for various data consumers? You get the point.

This is all on top of the need to even have a drone operator manually operate the drone in real time. Plan out your course of action (or ask AI to do it) and your drone executes tasks.

So where are we really going to see these changes be implemented?

To start we'll see changes in existing drone industries:

But as the utility of drones grows everything from last-mile delivery to public safety, and even healthcare will adapt.

Imagine getting packages delivered onto the porch of your high-rise balcony, a drone first responder to an accident that can relay the situation to officers who are trailing behind, or microdrones that fly around hospitals (or inside you) carrying material.

That last one may be a bit of a stretch but its interesting to conceptualize.

Drones are here to stay and they're changing fast! Follow for more.