Vision-based AI drones outfly world-class human pilots

https://spectrum.ieee.org/zurich-autonomous-drone-race

by samizdis • 4 years ago

81 142 4 years ago

spectrum.ieee.org

At this point nothing AI does (or could do) should come as a surprise. Humans will be completely outclassed in every scenario where AI is able to participate. The trick becomes how to benefit from AI as opposed to just entirely replacing human value.

jvanderbot2• 4 years ago

I'd bet a lot of money they are following pre-optimized trajectories or policies, using onboard sensing to localize. I didn't get a good sense from the article whether the CAD model was only used at startup / online, or if they used the CAD and found routes or policies first (pre-trained) and followed / executed during.

A great achievement, no doubt, but the days of hunter-killer drones flying willy nilly through arbitrary environments are not yet arrived.

spoonjim• 4 years ago

The days of hunter-killer drones flying willy nilly through arbitrary environments is already here! What do you think Predator drones are doing in the Middle East?

melling2• 4 years ago

I imagine the military value will be too hard to overlook, and will get a lot of Defense dollars.

A drone air force is much cheaper than a traditional one with pilots.

Drone use is already evolving in warfare.

https://finance.yahoo.com/news/column-drones-old-shells-key-...

https://news.sky.com/story/ukraine-war-teenager-and-father-u...

Dozens of automated drones could make several quick surveillance runs on a regular basis, for example, using image recognition to report equipment location, etc

Buttons840• 4 years ago

A drone air force is much cheaper than a traditional one with pilots.

"Cheaper". This might be a reason it won't happen. At least not until wealthy militaries have learned some hard lessons.

hodgesrm• 4 years ago

How are AI capabilities different from the introduction of guided missiles and fly-by-wire control in aeriel combat? Humans are still there but they take on different roles -- the end result is a much more formidable machine using blended capabilities.

Framing the competition as AI vs. humans as in the UZH competition therefore seems to miss the point. Humans already can't fly high performance aircraft without machine help.

tintor• 4 years ago

Not always more formidable machine.

Aircraft without human inside can be smaller, more expendable, can perform longer missions and tolerate higher g-forces.

oldstrangers2• 4 years ago

In the near term yes, humans and AI will have to work together. Eventually the only limiting factor in the AI's progress will be the human component.

For stuff like these drones, where is the benefit of a human component? Picking the drone up off the ground? Being a spotter? Driving the drone to wherever you want it to fly?

Lets take self driving cars for example. The benefit there is we no longer have to drive, but you also lose millions of jobs (taxis, trucking, etc).

Capitalism isn't going to take a romantic approach to AI / human harmony. It will do what it always does, cut costs and turn profits at the absolute expense of everything that's not creating shareholder value. And to that extent, the military industrial complex will do exactly the same thing, just replace shareholder value with bloodshed.

submeta• 4 years ago

I wouldn’t want to work on these kind of projects as it is clear that eventually these technologies will be used for military purposes. They will lead to a kind of weapon that will get out of control. And will be used to kill many human beings. Let’s make no mistake: This is no innocent research in robotics but lays the foundation for weapons we don’t want to see or have.

yboris2• 4 years ago

Excellent point. Here's an important PSA: https://www.stopkillerrobots.org/

anigbrowl2• 4 years ago

This seems like an exercise in futility. Robotic killers don't need to be sophisticated or expensive; writing international law is largely useless when the technology isn't limited to nation states.

landryraccoon2• 4 years ago

Do you think it’s a risk that more authoritarian and belligerent states develop them first?

War has been a constant force in human history up to and including the present day. I would greatly prefer that if such weapons exist that my society has access to them if others are sure to follow.

For that reason I think it’s ethical to work on weapons technology. You could be a principled pacifist, but I think that’s self defeating. Ukraine gave up their nuclear weapons, that did not bring them peace. Arguably if they had nukes, they would not be at war.

melony2• 4 years ago

Put the engineers in charge of the executive state, then we can talk.

kwatsonafter3• 4 years ago

You're missing an important caveat of your own argumentation; if you believe in Liberalism you must by extension have some skepticism concerning the ability of authoritarian states to, "keep up" with freer states in terms of innovation and cultural robustness. Otherwise you're arguing for an arbitrary terminal struggle between competing world cultures. (try reading history). You're basically scared by the current war in Ukraine and are willing to take on an ugly foreign policy position to feel safe.

It's going to be a bunch of software developers wearing t-shirts and shorts that eventually bring us to nuclear war. You know, "to keep their daughters safe."

Fuck all of you.

falcolas3• 4 years ago

Do you think it’s a risk that more authoritarian and belligerent states develop them first?

It does not require the development and realization of a technology to create a defense against that technology. We don't fight bombers with bombers, nor do we fight ICBMs with ICBMs, nor nukes with nukes (OK, perhaps not the best example, but only because MAD is a political deterrant against rational actors (which one can argue Putin is not)).

Ukraine gave up their nuclear weapons, that did not bring them peace. Arguably if they had nukes, they would not be at war.

Only if they were depraved enough to use them. At which point they have to concede that they too will be annihilated. The threat of nukes alone does not prevent a country from being attacked - especially by madmen.

trhway• 4 years ago

The high precision weaponry and drones are already equalizing battlefield for Ukraine. The autonomous drones would be even bigger equalizers, somewhat similar to nukes, and thus leading to more peaceful world.

For example a small country can't maintain a top modern air force, and thus for example Russia dominates Ukraine airspace. Autonomous drones would make it a contested situation at least.

fny• 4 years ago

1. The same argument could have been made about flight itself or even GPS.

2. In general, the tech transfer tends to work the other way: military develops and uses the tech first then come the civilian applications.

Speaking of 2, Boeing has developed autonomous drones that are deployed in Ukraine.

[0] https://fortune.com/2022/03/29/artificial-intelligence-drone...

solardev2• 4 years ago

But any and all technology can (and will) be militarized. Do you similarly forego working on anything touching encryption? Radios? CPUs? Machine vision? Simulations? Weather? GPS? Automobiles? Flight? Chemistry? Where do you draw the line? Tool usage has always been a, um, double-edged sword... humans will be humans.

Targeted drone strikes are at least somewhat less brutal than WW1-style trench combat or hand-to-hand melee combat with swords and spears. Wars in general are getting more precise, with fewer fatalities, over time (https://ourworldindata.org/war-and-peace). For all this talk of killer robots, the Rwandan genocide killed more with machetes than our drones.

Is it necessarily a bad thing that we're replacing boots on the ground with robots in the air? Eventually, one would assume, our opponents would catch up as well, and maybe at some point in the future, battles will be more drone-vs-drone than fighter pilots dogfighting it out. Our killer robots against theirs.

Maybe wars are getting more high-tech, but if their body counts are lowering, isn't that a good thing?

Jolter• 4 years ago

I think there is an argument being made that the drone technology enables a distancing of “the person doing the killing” from the action. That being 200 miles, or 2000, from the person being killed lets one take the decision to kill more lightly. Drone operation becomes like a video game.

The worry is that such distancing increases the risk of collateral damage, or simply mistaking of targets. There were accusations during the wars in Iraq and Afghanistan, that the US were bombing weddings and civilian homes.

golemiprague• 4 years ago

What's wrong with military purposes and how do you know what exactly it will lead to? If anything the most effective weapons lead to deterrence at best and cold war at worst, actual war are fought with ak47.

rg1113• 4 years ago

While many think that they should not work on such projects, my take is the opposite.

If we don't work on these projects, then we will not have autonomous killing machines, and our rivals will.

Turkey already have autonomous predator drones [0], and they have been used in the battlefield.

If countries like the US, UK, France, Germany, India, Japan, Israel etc. do not have these, and China, Iran, and Russia do (they surely do or soon will), we are at massive disadvantage.

We will have to deploy humans on the field when they won't have to. That is a gap too wide to exist. It is more fundamental than gunpowder.

Like Archimedes created massive war machines, and mathematicians and physicists worked on artillery, weapons, and Manhattan project, AI scientists of today should help develop autonomous killing machines. Freedom isn't free.

[0] https://foreignpolicy.com/2021/07/05/killer-flying-robots-dr....

kwatsonafter3• 4 years ago

You might find this interesting:

https://www.theatlantic.com/ideas/archive/2022/05/ukraine-ru...

the_af6• 4 years ago

Interesting but puzzling at the same time.

I don't disagree modern wars will result in whole classes of equipment being found obsolete or not cost effective, but this strikes me as odd:

[quote from the article]

"The military we have—an army built around tanks, a navy built around ships, and an air force built around planes, all of which are technologically advanced and astronomically expensive—is platform-centric. So far, in Ukraine, the signature land weapon hasn’t been a tank but an anti-tank missile: the Javelin. The signature air weapon hasn’t been an aircraft, but an anti-air missile: the Stinger. And as the sinking of the Moskva showed, the signature maritime weapon hasn’t been a ship but an anti-ship missile: the Neptune."

Well, none of those types of missiles are fundamentally new, and have existed in one way or another for decades (e.g. the HMS Sheffield was sunk by an Exocet anti-ship missile during the Falklands War in 1982!). More importantly, they work because the Ukrainians are fighting a war of resistance against an invading force; were they to go on the offensive (hypothetical, imagine they -- like the Soviet Union in WWII -- were to advance into Moscow and decapitate its leadership) they would need tanks, and airplanes, and other allegedly "obsolete" equipment.

The Ukrainian defense works where they are given support and weapons by foreign powers and because of a Russian mismanaged offensive.

Say -- again for the sake of argument -- the invasion was a coordinated effort by NATO against an hypothetical Ukrainian rogue state. How long would they last with their Stingers and Neptune anti-ship missiles?

As a devil's advocate against my own argument, Afghanistan ended up "defeating" the US in the very long run, for reasons not directly related to the weapons they used. And what weapons did they use, anyway? (Good old) Stinger, RPGs and AKs? No Neptunes in sight.

antisthenes• 4 years ago

Turkey already have autonomous predator drones [0], and they have been used in the battlefield.

Turkey is part of NATO.

China, Iran, and Russia do (they surely do or soon will), we are at massive disadvantage.

Countries like Iran and Russia aren't able to sustain an electronics industry to produce something as complicated as AI Computer vision drones. I hope I don't need to explain why.

We will have to deploy humans on the field when they won't have to. That is a gap too wide to exist. It is more fundamental than gunpowder.

This is unsubstantiated fear mongering. The economies are too intertwined for either US or China to deploy their forces full-scale. If we ever get to the point of a full-scale confrontation, some AI-drones will be the least of your worries.

AI scientists of today should help develop autonomous killing machines. Freedom isn't free.

Yikes.

loudmax2• 4 years ago

The economies are too intertwined for either US or China to deploy their forces full-scale.

This was was a common belief about Europe before World War One. Unfortunately, that didn't prevent the carnage.

I tend to agree with you about the existential danger of full-scale confrontation. And to be fair, your concerns about these weapon systems are well founded. These things are really dangerous and we need to think very carefully about how they're developed and deployed.

But as we see in Ukraine, there's a lot of scope for limited confrontation. In this case we'll want access to deadlier weapons than our enemies have access to.

To those of us that believe in liberal democracy, this isn't just about one team vs another. This really is about living in a world where freedom is the norm rather than living in fear of an autocrat.

UberFly• 4 years ago

I completely agree with your post but say with only partial sarcasm that the US/EU will develop these first, followed shortly by China but only after they steal the technology behind it. Then a trickle-down from there.

yoaviram• 4 years ago

You do realise that right now on a Chinese version of hacker news, a Chinese version of you is making the same argument?

The messed up aspects of this world aren't inevitable, we all take part.

belval• 4 years ago

As they should, with 8 billions inhabitants, assuming that peace can be kept on good intentions is the pacifist fallacy. We just got a wonderful example with the invasion of Ukraine. You can be nice, you can play nice, you can apply sanctions, send strongly worded statement, and yet Ukrainians are dying and tanks are still rolling in.

You know what's stopping the Russians? Javelins and Bayraktars.

rg1113• 4 years ago

Yes I know. Basic game theory comes to mind.

But if you have closely observed China, you know that it is not some private hacker arguing in favor of development of AI drones in some private forum.

It is some Harvard/Stanford/Peking/Tsinghua educated scientific policy advisor deep inside the army/government doing this job.

tb0ne• 4 years ago

I don't think there is a moral debate about developing autonomous military drones in China.

These self-restrictions based on morality regarding the development of military drones is really something only some western nations do. Other countries, if they reach the capabilites to develop them, they just do.

mgraczyk• 4 years ago

The reason both sides are correct in their conclusion is precisely because there is a "Chinese version of hacker news"

Frost1x2• 4 years ago

I'd argue that negative aspects of the world are inevitable in competitive environments given survival instincs humans have. The fact is, a lot of things really are zero-sum, like land (e.g. Russia and Ukraine)--there's only so much of it the way we take ownership of it.

Ultimately this sort of environment leads to competition over resources vs collaboration between humans. So long as environments are competitive and people want to survive, the rest emerges inevitably I think.

I always found modern conflicts interesting because humans basically formed societies to reduce competition and dangers, to help their odds. We can now take turns taking watch at night and I sleep most days vs living as a nomand I'm always in danger for my life and resources. We then moved from hunter gathers to agrarian societies because it provided even more security and stability, we didn't have to compete for food we could contribute, grow it, and share it. We formed towns, built houses. We worked together.

Then somewhere along the line we decided to reintroduce survival dynamics abstractly in these safer societies. We reduced your food security and livelihood security based on valuations of your contribution to society, we add power dynamics and authorities of control, created the idea of private property rights, we setup market places where people compete abstractly through wealth transfers in the forms of money, then we tied things back securities like food and housing to these things.

Maybe we got bored and many want to be in a threatened survival state. I'd like to think we, as a species, could toss the people obsessed with accumulating wealth and power to the curb and try stripping away many layers of abstract competition we created and look at creating layers of collaboration. It doesn't need to be another failed attempt at socialism, but it seems like we can at least be far less competitive to one another and still grow.

archontes• 4 years ago

This is the tragedy of the commons. Its solution is a central authority. How do we get there?

imtringued2• 4 years ago

What if there is an easy and painfully trivial answer to achieve world peace and avoid wars?

Imagine you are one of a million people on this planet who know the secret. Everyone else won't believe you because it is too simple and hence very easy to abuse in the other direction. What if it was as simple as flicking a switch and just doing it?

Here is the problem. If it just about flicking a switch, how to make sure people don't switch it from utopia back to dystopia?

Your utopia needs to be a self preserving system even in the face of violence. A pacifist system that isn't self preserving will be destroyed and taken over by non pacifist systems that are self preserving at least in the short run.

A democracy needs the backing of a military force to keep it alive. If there is something better than democracy then it will also need a military to protect itself.

martindbp2• 4 years ago

Yes, the invasion of Ukraine hopefully showed how naive this type of thinking is, i.e. "I hate war, therefore I won't support or work on defense technologies".

tsimionescu23• 4 years ago

The invasion of Ukraine has shown precisely the opposite: playing military games gets youe neighbours antsy, and an antsy empire just invades to calm its nerves.

If Ukraine hadn't been slowly de facto integrating into NATO (NATO weapons, NATO training, and even joint military exercises with NATO, on Ukrainian soil), it would have been much harder for Putin to justify the extreme costs this invasion is having on his economy.

wing-_-nuts• 4 years ago

The end game of intelligent autonomous weapons are dangerous to everyone. Google slaughterbots. We've been pretty successful in stopping major nation states from using bioweapons so far. Maybe we can do the same for AI. Maybe we can't. That doesn't mean we shouldn't try and the very least we can do is not start a bloody arms race as to who can develop the best autonomous killing machine.

rg1113• 4 years ago

We've been pretty successful in stopping major nation states from using bioweapons so far.

I am sure everyone has it. Every major nations. Although they might not use it.

AI

I am sure it is already there. Many have it. Turkey, and at least China. Maybe Israel and the US.

And it will not come in clear paradigms, I think. Like FSD in Tesla. Use will happen for components first, then major things, then maybe one day, fully autonomous.

How do you even regulize it? What happens if they use Logistic Regression? That's just Stat, right? What if add one more layer? Then that becomes MLP, and in the AI domain. What about CNNs? Transformers? Where do you draw the line?

What if the missile truck uses something like FSD?

arms race

This isn't going to be like other arms races. This will fundamentally change all weapons. And should not be used where it is not needed, though.

Also, gone are the days when countries flaunted big missiles. How would even other countries know, if you don't want them to?

anigbrowl2• 4 years ago

It's just autonomous technology in general. Making an autonomously-operable object destructive is the easy bit. Widespread use is a lot closer and easier than people realize, but I'd rather not spell out why that is.

lowbloodsugar4• 4 years ago

We've been pretty successful in stopping major nation states from using bioweapons so far.

Name a major state that has experienced existential war since their invention. Same for nukes.

"We've been successful in stopping" - there has been no "we" doing any "stopping". There have only been "others" "choosing not to". There would reasonably be some point of conflict where the decision falls the other way.

Foivos• 4 years ago

Major nation states do not use weapons of mass destruction against each other out of fear or retaliation. If you do not want your any to use autonomous drones against you, the best deterrent is to build some on your own.

lwswl• 4 years ago

Since when has google been making slaughterbots?

Or do you mean those dogs they were building for the military?

lowbloodsugar4• 4 years ago

Hard agree. The only counter to an enemy AI swarm or a runaway AI swarm is another AI swarm. Sure, we'll probably fuck it up, but at least there's a chance for us.

The alternative is simply accept Chinese rule.

That's it. Those are our choices. And for the pacifists, China is still going to develop their own AI swarm, and they could lose control of it, so dying is still on the cards with the second one, too.

I love people here focusing on the "end game". Yes, this is the horror: the rational choice is to develop the technology that will likely kill us all. They argue "The end-game is so bad that we must lose the mid-game", as if that will stop other nations from developing the end-game technology!

Really, anybody who isn't already a monk who lives entirely off the land through their own labor, doesn't got to make any moral argument, or argument about "well, Chinese rule wouldn't be so bad - we're just fighting for culture". Give up your culture, then we'll talk.

Everyone who believes there are other choices are simply living in their "just world" fantasy.

tsimionescu23• 4 years ago

This same logic could be used to justify any atrocity. What exactly makes you so sure China is developing such dangerous technology?

And even if they are, why isn't it a better idea to negotiate an international ban on this type of research instead of driving full force into it?

Creating ever more powerful weapons is a fool's game, and every knowing participant is a human monster.

cjbgkagh• 4 years ago

“Mr President, we must not allow a mineshaft gap!”

stevenalowe• 4 years ago

Typo: should be “minecraft gap!”

bob1029• 4 years ago

If we don't work on these projects, then we will not have autonomous killing machines, and our rivals will.

My biggest fear is that the next generation of warfare is going to be lost only due to lack of imagination.

I am starting to see a new opportunity emerge for the skills and tools held by AAA game studios and fintechs. Building a UX/UI that allows for you to interact with your next-gen weapons system is going to be the biggest challenge in my view. How does 1 person manage a fleet of 10k drones in a strategic conflict? How do you integrate hundreds/thousands of these fleets into a cohesive force? You are talking about potentially millions of battlefield actors. And billions of discrete events per second. How do you process all of this data and then make a rational decision by the time the next loop comes around? Every millisecond that goes by without a decision is a substantial step closer to defeat.

mosfets• 4 years ago

Very impressive racing performance indeed! Though, the "vision based" wording in the title is a little misleading. Human pilots can use vision to develop an understanding of the whole environment, while the drone in this project still need to consume a CAD model of of the track. The vision is probably used to replace the motion capture system.

quantumduck2• 4 years ago

Yeah, vision is indeed used only to localize, not really to recognize the track. Low-key this is what I don't like Waymo/Cruise approach to autonomous driving - they use HD maps to localize themselves and I see that as finding an easy way out instead of tackling the harder problem.

danuker• 4 years ago

A humorously-named method I like is LSD-SLAM: Large-Scale Direct Simultaneous Localization and Mapping.

It is fast, and it produces the richest point clouds I've seen, from even monocular video. The "Direct" comes from directly using pixels, not features such as SIFT/SURF.

I suppose it could be used in environments with not too many reflective or transparent objects.

black_puppydog3• 4 years ago

Wow literally the only thing people seem to be putting up as an argument for working on this is "someone else will do it otherwise."

I mean, yes this tech like most others is dual use. But let's for the sake of argument assume that we know this tech will mostly be used for military purposes.

Yes, others might still develop this. But in the meantime "we" (whoever this is, reminder: HN users aren't just from the US) get to miss out on other things we could have invested millions/billions into. That infrastructure project? Gone. That research into new/better energy storage? Gone. That quirky project about preserving natural habitats in the third world? Gone. By the way, that one might have actually contributed something to a lasting peace in the region. But hey it doesn't matter because we'll win any armed conflict with our drones.

That might sound bitter and/or idealistic, but ultimately if we accept that the funding of whole branches of research depends on how much benefit it provides to the military, then ultimately we can say all day long how we value nature and human rights and all these things, but we're really just dressing up a really fancy warrior armor with some Greenpeace badges.

quantumduck2• 4 years ago

A couple of points:

1. We doesn't have to be another country/military. It could just be another research lab in US, China, or anywhere else in the world. Some people just like pushing tech to the limits, they seek the next hardest problem. The hardest problem post robust drone controls/motion planning era (e.g. drones like DJI) is to make them pure vision based.

2. Millions/Billions? Lol what are you even talking about. It takes about 100k to fund a graduate student for one year. I wouldn't be surprised if this entire research work was less than 1 million, maybe even 500k.

black_puppydog3• 4 years ago

1. Believe me I know the appeal of a hard technical problem. PhD in Vision here. :) 2. Military drone tech as such? Yes, billion-dollar industry.

martindbp2• 4 years ago

That infrastructure project? Gone.

And do you know what would be gone right now if Russia managed to build killer drones first, or HIMARS? That's a rhetorical question.

echelon• 4 years ago

If we don't built war machines at parity, we become Ukraine. Toothless. Ripe for bullying at best, hostile takeover at worst.

Everything in nature evolves to consume the energy of other systems. Our unnatural Kumbaya/Disney society just plasters over that fact. Despite the bubble we live in, at the most fundamental level, we're evolved to consume.

Nature is metal. Look at orcas playing with food to teach hunting. Hornets laying larva in live prey. Lions eating out distant relatives alive.

the_af6• 4 years ago

If we don't built war machines at parity, we become Ukraine. Toothless. Ripe for bullying at best, hostile takeover at worst.

That's only one part of the equation. It could be that the perpetual arms race also leads to situations like Ukraine's now, or many others in the recent past.

It's a way of thinking everything is a nail, since you are holding a hammer.

Nature is metal. Look at orcas playing with food to teach hunting. Hornets laying larva in live prey. Lions eating out distant relatives alive.

Nature may be metal, but we aim to be better. After all, when a male lion takes over a pride, he often kills the cubs sired by the previous male; do we want that in human society? The good old "is-ought" [1] problem of philosophy applies here.

[1] https://en.wikipedia.org/wiki/Is%E2%80%93ought_problem

loudmax2• 4 years ago

These things aren't mutually exclusive. It isn't the case that money saved on weapons research would necessarily be spent on something positive, like cancer or green energy research.

I do value nature and human rights, and sometimes that means preparing to fight those who don't value nature or human rights.

black_puppydog3• 4 years ago

But in terms of budget, whatever your budget is, as long as its finite it will always be a zero sum game between "mostly civil" and "mostly military" applications, fuzzy as that line is of course.

seydor• 4 years ago

get to miss out on other things we could have invested millions/billions into.

Things like the internet, space flight, nuclear power etc?

Unfortunately, defense spending is so successful and brings high rewards exactly because there is a strong motivator behind it. I am very very very doubtful the same amount of money would have high return of investment if it was instead given to cookie-cutter science grants.

planetsprite• 4 years ago

this however saves lives in the long term. MAD becomes predictable when you have AI v2 vs AI v4. The countries resolve peacefully before the battle even starts and exchange resources. The computers can ascertain losses before reality proves it.

tsimionescu23• 4 years ago

That is dangerously bad thinking. MAD is a pipe dream, and the world is still much closer to annihilation today than it was before nuclear weapons were invented. There have been far too many close calls to count.

Secondly, no country will back down on the calculations of some unproven AI. And those calculations could only be proven in blood.

bickett• 4 years ago

I wonder how this will change in the future.

gcanyon• 4 years ago

“…and it absolutely Will. Not. Stop. until you are…” a shot rings out, and Sarah Connor drops to the ground with a bullet hole dead center in her forehead as a terminator drone whizzes by.

Yes, I know xkcd did it years back. I suggested this concept to an Air Force general at a 2002/3 planning meeting for the DARPA Grand Challenge. He looked at me cross-eyed and said “We don’t deploy systems like that.”

jonathankoren• 4 years ago

Ironically, I think flight is much easier than driving. Simply because there's less things to hit, and it's all controlled (or at empty) airspace.

It's just the low level navigation in residential neighborhoods that's hard. But a drone helicopter at a heliport or a field? Why not?

Vision-based AI drones outfly world-class human pilots

Related Stories