Comment by abakker | Hacker News Reader

Comment by abakker | Hacker News Reader

I'm not really surprised by this. At a drone conference in 2016, I first saw presentations from them among many other interesting technology vendors. They had a "vision" that they carefully presented on stage of what they would do, but everyone else at the show was either demoing technology or taking orders. Airware had a "call us for pricing" kind of attitude.

Meanwhile, its been a year since Verizon acquired Skyward - which was a force in the drone industry pushing for better flight management tools, better collaboration with ATC, and an affordable product for small-time commercial operators that got most of their money as contractors/service providers.

At the enterprise level, Drones just didn't solve a lot of problems. At the small farmer, construction contractor, gravel pit, etc, drones solved a lot of problems, but drones were the anti-scale solution. They Reduced the work so drastically, and improved the result so much that there was no need to scale up software licenses. In many cases, it just did the work better, in a 10th of the time, at a much lower cost. But, there was not a giant amount of work in the queue for those companies. Just a steady flow of the same jobs that used to take longer.

Now, beyond visual line of sight (BVLS) flights might change that for the enterprise eventually, Power lines, rails, pipelines, and roads are all good candidates for remote, automated drone inspection. But, I suspect that it will be either DJI hardware, Custom Hardware, and some amount of custom software that really integrates drones into the enterprise process. Flight planning is only part of the picture. Maybe Airware had it all, but, if so, they were too early - they couldn't outlast the slow (and rational) pace of aviation regulatory change.

It has less to do with the space, and more to do with the odd engineering decisions and multiple pivots.

There wasn't a SDK or platform to build on and we didn't take any actual orders for anything (so we had minimal success stories or $$$). The hardware we showed was built by us, and ran a custom OS (not based on linux) that was nearly impossible to work with. We had a very talented engineer rage-quit after spending a week attempting to figure out how the OS handled the internal message bus system to debug an issue.

The hardware was also crazy expensive. We built literally everything: the vehicles, autopilot, the autopilot OS, the comms board that connected to motors (and enclosures), the sensor boards (GPS, gyro, etc), the radio that talked to the autopilot, the flight planning software, and later an entire enterprise cloud stack written in only the newest of the new JS frameworks and containers. The team often decided to do things such as not use off-the-shelf processors like the Nvidia TK1 or existing embedded software solutions -- nothing was 'good enough' so everything was invented.

Oh, and did I mention we also cycled through recruiters like candy?

So, a death by NIH?

Certainly sounds like it. I find this bit pretty funny though:

> entire enterprise cloud stack written in only the newest of the new JS frameworks and containers

I'm of the opinion that the JS ecosystem is one of the few spaces where a healthy dose of NIH can save you a lot of time and money in the long run. And it seems like the only place where they didn't go for a custom solution.

That people are choosing js for drones boggles the mind.

It makes a lot of sense for high level mission planning (more agile than other languages + it runs everywhere) and interacting with web apps.

They aren't. He said that was for their servers.

At the end of the day, a lot of what NIH ends up being is employees extracting experience and resume value at the expense of business goals. It's kind of funny, then, that the JS ecosystem gives you a better resume if you have keyword soup with trendy frameworks.

Sometimes engineers are just curious about new and exciting stuff and don’t make the right risk/reward calculations.

Yeah it doesn't need to be a conscious and deliberate choice. If some folks have more of a tendency to do this and wind up with trendier resumes, then simply hiring trendy resumes for senior roles gets you there.

“But developer experience!”

Takes 3 weeks to debug.

What’s NIH? Not Inough Humility?

Not Invented Here https://en.wikipedia.org/wiki/Not_invented_here

Hah, "Not Inough Humility" seems to apply as well :)

That is so right. We’ll all be giving talks about this soon. NIH: Not invented here, aka not enough humility. :)

"Not In-House" is another common interpretation, in addition to "not invented here"; both mean the same.

I'm absolutely going to add "not inough humility" to the list from now on, it works just too well :)

more like too much easy money to burn

Was the engineering leadership all from academia or something?

yes, both ceo and cto recent mit grads

The CTO is a recent MIT PhD, but I don't think either of the CEOs are academics. The original CEO & founder did go to MIT, but as an undergrad and years before founding the company. The most recent CEO has a business/MBA background.

Would Airware have attracted as much funding as it did in subsequent years if it didnt propose a pivot into hardware? Or do you think it could have eventually succeeded by focusing only on software?

Almost all those things should have been built in C

I'm not sure what conference you were at/what you saw presented, but there is drone software that creates a lot of value for construction, engineering, and mining companies. BVLOS may create a lot of value some day, but there are a lot more technical hurdles that will need to be cleared before it becomes safe and mainstream, just like Level 4/5 self driving cars. (unless you're talking about doing FPV inspections.)

Disclosure, I originally was the product manager for 3DR's flagship software product, Site Scan, which drove our transition from a hardware business to a software business, and I am currently VP of Marketing and Business Development here. Each mine that deploys our software saves, on average, 6 figures per year (after the product cost). This is on man-hour costs alone - the value is greater if you look at the increased safety of taking individuals out of active mines and increasing topography capture that enables analysts to identify issues before they turn into safety hazards on site.

That's just for one industry we serve - cost savings/time savings/improvements in project pursuit are all high for construction and engineering as well.

I think it's likely you saw some presentations from companies that had not done enough customer discovery or they were perhaps just very early in the process of identifying real problems and building the right solutions, but it's not representative of the industry.

>Each mine that deploys our software saves, on average, 6 figures per year (after the product cost).

Where do the savings come from?

They come from man hour savings from generating a topography of the mine and stockpiles. Our mining customers have found our product gives results within 1.1% of the accuracy of laser scanning the mine and is more accurate than the results from traditional methodology performed by an average ~3 person survey crew. It requires only a single person to capture the mine, and customers get the data back in a single day rather than 3 to 4 days later. Thus, they can capture the mine topography far more inexpensively and far for frequently than they could previously.

One other usecase with a lot of potential is inventory control and counting in warehouses. Not small articles, but for industrial warehouses and larger products drones combining bar code scanners and visual artical identification can improve things considerably. Fraunhofer ran some first test like two years in Germany. No idea what came out of it.

Does that really benefit from a drone vs a wheeled vehicle with a bunch of fixed cameras on it? I would think a gigantic planar environment is pretty much ideal for an industrial SDV; a lot of things get easier when you're no longer worried about weight— you can carry a full day's worth of batteries, not to mention having higher quality optics and safety lasers.

(disclosure: I work for OTTO Motors, which has customers in this space)

Why not both? A large, wheeled vehicle that also carries a small flying drone for the occasional inspection from above.

This is the approach I've been thinking about recently. The landing pad on the wheeled vehicle could have an inductive charger to recharge the smaller multirotor vehicle(s).

wheeled vehicle parked in the isles to take a picture of a space on top shelf would be blocking pallet jakcs and other vehicles that need to move between the isles. may be a railed solution hanging from the ceiling would be better.

Rails are a lot of infrastructure to install and maintain. People complain that sticking down magnetic tape is a pain in the butt.

Exactlly, also a drone is much more flexible in case you change your layout.

other way, you mound cameras ON pallet jacks and other vehicles

I could imagine both. In the case I mentioned the warehouse was euipped with highracks up to 13 meters if I remeber well. Also the drone flew during non-working hours, so yeah some ground based solution would work too.

I don't know why this is getting downvoted. years ago when we had multiple warehouses our employees would upload pictures of shelves into the cloud at the end of every shift and if a product goes missing we would look at the pictures and inventory record to find out when exactly the real quantity of items on the shelf and inventory level in the system became off. if a drone could fly thru shelves on schedule and AI could count the items that would be amazing

I always hear people raise this point but there are better solutions than a flying robot inside of a warehouse.

Out of interest, could you name a couple? I just happened to come across a couple students that are tackling this problem via drones.

A vehicle on wheels or a track is orders of magnitude more efficient. It’s not sexy but it’s more effective than having to power a flying vehicle and control it autonomously. If you really want something more drone-like, hang it from the ceiling.

What kind of solution is best depends to 100% on thr warehouse you are looking at.

Take you standard Amazon warehouse for small articles as an example. Maximizing untilisation, meaning the ratio between space used und spa e available, in that case you end up with a never ending mixture of differwnt small articles in the same location. Without RFID chips on every article, which is as of now prohibitively expensive, the only feasible way to count the inventory is using people.

For other warehouses and articles, like white goods or other large stuff that easily identified with out touching it whether a drone, a ground vehicle or something running on rails is better suited depends on a couple of factors. If you might run into situations that require a chenged warehouse layout, which occurs much more often than you think, fixed installations add a lot to the fix costs and rwduce flexibility.

Using ground based solutions runs into issues with high warehouses, storing up to 15 meters in height is not uncommon. Again, density is the main reason for this, getting more inventory at the se number of square meters. In that case a drone is both more flexible and can cover the heights as well.

But thats theory, not sure what came out of the Fraunhofer trials covering white goods.

I think automated drone inspection may be a more profitable niche than it first appears. My school had a couple grad students start up doing NDT analysis with drones. There seems to be a lot of interesting problems that could be solved with drones when they are treated as a platform for specialized equipment. And the specialization in equipment and analysis is good protection against commoditization. But I do agree with you that a lot of these problems don't really scale.

NDT - Non destructive testing? So coupled with "specialised equipment" I assume something like scanning the steel in a bridge for corrosion? Is that feasible - the noise in any feed from distance / stability must mean "usual" specialised equipment cannot be used?

You are probably overthinking this.

Simply being able to fly a high resolution camera with a bright light up underneath a bridge probably saves an enormous amount of time.

assert numBridgePieces == 1

Do you have a view of Airmap?

I’ve not personally used it, so I can’t comment on that part, but, it appears that they have gone with the open, extendable, and most critically - actually for sale - route. That would bode well in my book.

On my soapbox, the big missing piece in drones is a real platform for managing the concept of time series map data for a layperson. There is a lot of dimensions in the data that drones generate (lat, long, alt, color, time, camera stuff, angle of image, etc). Tons of data, and nobody has convincingly showed anything that goes into trend analysis. I’m not sure that any company is really delivering on the concept that this data needs to be controlled and analyzed. Especially in farms, the time series part is important, but it is mostly snapshots and intuition right now.

Back in 2015 Airware was pushing for its own technology stack, coupling an autopilot with mission management/ATC. Only later they pivoted into a software-only solution. Same goes for 3D Robotics.

Agree with your point, this is the vision we are developing at Sterblue !

Reply on news.ycombinator.com