Comment by joshvm | Hacker News Reader

Comment by joshvm | Hacker News Reader

Parent Full thread

Depends what you mean by "AI". Autonomy in aviation is far ahead of what you can get in ground vehicles. Autopilots are commonplace, even in light aircraft. There are established systems like TCAS for traffic avoidance. Aviation remains incredibly safe, despite huge levels of autonomy, because of the regulation involved.

What (I think) we need is a regulated system where every drone uses a standard transponder and negotiates airspace either with a controller (e.g. IFR) or other aircraft (e.g. VFR). The future is not everyone flying drones with proprietary AI collision avoidance. One existing product in this space is FLARM which is used by a lot of glider pilots and light aircraft, they also make a UAV variant: https://flarm.com/products/powerflarm/uav/

Autonomous drones themselves exist, are inexpensive, and can be set to fly a survey pattern based on a number of pre-selected GPS waypoints. Ardupilot will do this. The data is used for mapping at a much higher level of data than online maps. Think agri, ecology, general site surveying, etc.

That said, drones are not ideal for replacing aircraft in the large scale mapping. The flight time is poor. An aircraft modified for surveying can cover hundreds of square kilometres a day. A consumer drone with an hour's flight time (good luck) is going to struggle to compete.

Collision avoidance at the local level (e.g. are you going to hit a wall) is a different issue, and that can be solved by standard depth sensors.

>That said, drones are not ideal for replacing aircraft in the large scale mapping.

You are comparing consumer drones to airplanes. A quick google search tells me you can pick up a drone with over an hour of flight-time for 3.5k-ish (I don't know much about drone pricing so someone may have better options)

Another quick google search tells me cheaper planes go from 25k (crappy) to 40k (reasonable)

You can purchase 10 drones for the price of one airplane. Which I can't imagine an airplane could out map/survey 10 drones working together. I don't know much about drones or planes though so there are probably other restrictions.

> You are comparing consumer drones to airplanes.

Sure, because the OP called out Google maps and that's how Google maps (at high resolution) is created. Drone mapping does tend to be much higher quality because you get a higher image cadence at a lower altitude.

Fixed wing drones tend to be better in terms of flight time, that's certainly true. And you can get gliders which will last even longer if you manually thermal them.

> Another quick google search tells me cheaper planes go from 25k (crappy) to 40k (reasonable)

Add on at least another 20k for survey quality cameras (up to hundreds of k) and more for getting your plane modified (hole in the fuselage) and re-certified. Pilots are expensive, fuel is expensive and getting your annual COA (Certificate of Airworthiness, in the UK), insurance etc is not cheap. Google's surveying aircraft have at least five cameras so they can use photogrammetry (structure from motion) to create their 3D maps. As soon as you introduce surveying LIDAR into the equation, add a zero or two onto the cost estimate.

I don't disagree that you could fly a drone fleet instead of an aircraft, but for country-scale mapping, aircraft make a lot of sense.

At the practical end of the scale, I was involved in a project where we were quoted in the region of $25k for a 4 hour LIDAR flight, and my colleagues considerably better versed in the subject than me were reasonably happy this was (i) better value for the region than they expected and (ii) cheaper than digital elevation mapping the same set of discrete areas by specialist drone with the drone operator driving in between. Limited range of current generation drones is the real downside...

The flip side of this unhappy mapping equation is that if you're happy with lower resolution remote sensing data the government satellite programmes are mostly free and libre.

Depends what you want to do with your drones, but 3.5k would be really low end. I think 5k is more realistic, and you're looking at 15-20k if you want high precision (rtk) drones.

On the other end, it's probably not just any plane for mapping, so cost could be higher. Plus the gimbal and sensor cost is going to be non negligible.

Airplanes have to fly higher though, and thus at lower resolutions. The newest satellites are also getting closer to what they do, and they obviously have better coverage, so the market is shrinking.

Core difference is /pixel resolution. Drones can easily produce 2-3cm/pixel. Airplanes don't do that well because they can't get that close to the ground and move too fast. For larger areas, though, airplanes are MUCH faster. Realistically, at those larger areas, though, you run against satellites, too.

> What (I think) we need is a regulated system where every drone uses a standard transponder and negotiates airspace either with a controller (e.g. IFR) or other aircraft (e.g. VFR).

Collision avoidance when in visual conditions is "see and avoid", which is all aircraft's responsibility. CFR 14-91.113.b [0]

There is no solely electronic negotiation of airspace permitted (at least not currently). It is unlikely that the FAA would change the law to require all aircraft to be so equipped and, even if they did, it would be close to two decades before any such deadline would come to be. Look at the slow adoption of ADS-B, which is only required in a select minority of airspace starting in 2020.

[0] - https://www.law.cornell.edu/cfr/text/14/91.113

Have you seen the size of a transponder? You're not going to put that in a drone doing 1h flights. They're about 1/2kg. 4G is more likely, but coverage can be an issue. The bigger drones that could carry one are much bigger, and they fly for several hours, but they do not have the same ease of use as the smaller ones.

Flarm and Adsb (mostly in for now, out is more difficult/heavy but there are a few prototypes) are seeing some use. Companies are also trying to standardise around existing aeronautical standards, or to create new ones where needed. But it's a free market and it's not clear which solution will win in the end. The regulators are also closely following.

> What (I think) we need is a regulated system where every drone uses a standard transponder and negotiates airspace either with a controller (e.g. IFR) or other aircraft (e.g. VFR).

Before that happens, a good first step is to require all drones to transmit an ID linked to a person. I suspect that just that step would help to curb the worst excesses of amateur drone pilots.

Maybe someone should name the AI "Autopilot - now with 99% less head chopping."

Reply on news.ycombinator.com