Google didn’t lead the self-driving vehicle revolution, John Deere did | Hacker News Reader

Google didn’t lead the self-driving vehicle revolution, John Deere did

washingtonpost.com

181 pointsLibertatea136 comments

While I certainly agree that self-driving farm equipment is revolutionary, I'd argue that Google's technological advances are in the field of being able to understand how to drive on a road which marked for humans to drive on, and which is simultaneously in use by human drivers as well.

I'm sure that there are problems and issues which arise from navigating a field and manipulating whatever equipment you've got for whatever task you're working on that I'm just not fully appreciating. However, I really question if that task is anywhere near as complex as safely navigating traffic while obeying all the laws of the road when you can't even be sure of the state of the road you'll be travelling on.

It's just a totally different problem being solved. This is apples and oranges.

I agree. A farm is a type of factory, and thus has fairly controlled conditions compared to a public thoroughfare. Any number of solutions could work for a farm that would be untenable for self driving cars, such as marking the expected area with sensors to define the working area, outlining the area on a GPS map, running the equipment over the area one or more times to teach it, etc. None of these work as a general purpose solution to a self driving car, and they might not even be progenitors to the self driving technology we end up with.

A solution to a highly constrained version of a problem may not be applicable to the general purpose problem in any way.

Edit: mean to say "may not be applicable" in last sentence.

Yeah, agreed. The overlap technologies of position fixing and mechanical control have actually been solved for a long time now. The current interest in self-driving vehicles is based on advances in sensors and processing to detect environmental conditions and hazards on roadways and in places where people commonly drive cars.

You don't think farm equipment has to navigate unexpected hazards or coordinate with other moving equipment or people?

I'll agree that roadways have much more to do with avoiding traffic, but I don't think that the restrictions on vehicle operation or avoiding hazards on the terrain or having to understand and plan according to layouts devised by and primarily for humans is substantially different between the two, but rather, a difference of parameters. (Farm equipment maintains a lower speed and for a different reason, but still has to monitor and obey rules about maximum speed under various conditions, for instance.)

I think you're underestimating the degree to which the two tasks are fundamentally the same.

"You don't think farm equipment has to navigate unexpected hazards or coordinate with other moving equipment or people?"

Nope, they don't need to worry about those things. Self driving cars and tractors adhere to ISO-25119 and within that, tractors adhere to agricultural performance level which doesn't have the same risk. Self driving tractors are required to have an operator in the cab and they are not allowed on public roadways with "auto guidance" engaged.

I work for Case New Holland - John Deere competitor.

It would seem that at 10mph all you really need are some sensors looking forward that have the ability to provide enough warning that the vehicle can stop before it hits an obstruction. Kind of like the auto-brakes that some cars have.

That's a lot different than having a car which can at 70mph successfully swerve enough to avoid a wreck while still not swerving so far as to cause another wreck.

I agree with the larger point that robotractors have an easier job, but I really wonder how close we are to a robocar "which can at 70mph successfully swerve enough to avoid a wreck while still not swerving so far as to cause another wreck."

I think there was some discussion about the ethics of the software in google's cars because they would solve driving problems with a view to minimizing harm to all involved, not merely the car's occupants. So I think this kind of thing is very much in development, and indeed more sophisticated than that (e.g. The car might swerve into another vehicle to avoid hitting pedestrians).

Ah, forgive me. I wasn't suggesting that such a robotic car exists already. More that successfully navigating such a situation on a freeway would be an acceptance test that I would like to insist on before entrusting my life to one, and would DEFINITELY insist upon prior to purchasing one.

Were I in need of a self-driving tractor given the much lower speeds and whatnot I might be much more willing to accept a fairly primitive autopilot rather than something truly self-driving. Freeing me from needing to constantly be on the alert to keep a combine from straying from the optimal path is far, far more important that something that'll drive itself from the barn to the diesel pump and then onwards to the field. Were I wealthy enough to afford so much land to where owning such a vehicle was necessary then either driving it myself or hiring someone else to handle that would be entirely reasonable.

> but I really wonder how close we are to a robocar "which can at 70mph successfully swerve enough to avoid a wreck while still not swerving so far as to cause another wreck."

Automated cars can already do this far more reliably than your average driver.

>> You don't think farm equipment has to navigate unexpected hazards or coordinate with other moving equipment or people?

Coordinating with other moving equipment I did not address, because that is too broad a description. As far as hazard avoidance is concerned I don't know the state of the art for farm equipment, but I don't think they are as focused on fully autonomous operation, which has been the goal on the transportation side for some time now. In any event, there is obvious overlap in a number of areas, but I simply don't think the premise of the posted article holds up. Farm equipment is not the "real revolution" in autonomous transportation. It's merely one area in which some of the dependent technologies have already seen the light of day.

Even if the tractor has to be aware of such hazards, the degree is totally different. The tractor may need to identify that something unexpected has occurred, but it doesn't need to know anything about it other than that, most likely; it can likely deal with all unexpected occurrences by stopping and waiting for someone to come manually fix it. Self-driving cars need to make complex decisions about hazards, because stopping in the middle of a public street is probably not an acceptable behavior, it may not be practical given the speeds, etc. That means the car needs to know a lot about the hazard in order to figure out what to do about it (swerve? honk? alter speed?).

I think you're correct. I can vividly remember a local news story from at least 15 years ago that talked about self-driving cars being just around the corner. All it would take is putting sensors in every roadway and having dedicated lanes for them. I was pretty young and excitable, and I could not figure out why we weren't putting those sensors in as roads were replaced around the city. Google requires none of that insane retrofitting, and is therefore much more likely to be useful.

I think this is a valid approach, in theory.

IE, find some limited use case for a self driving cab service in a city and try to make the problem more like an agricultural field.

OTOH, it would require a city or state to actively pursue this kind of a bleeding edge technology, make decisions and possibly even develop the technology. Maybe there are some municipalities out there with the competence for this, but not many. I'm not as down on government abilities as some on here, but I think you have to admit that making such a technologically innovative, blue sky project a public one seems like a recipe for disaster.

The self driving cars can just develop on the free market. If the cars are good enough to go on normal roads people will buy them. I imagine that if they are prevalent enough, cities will then gradually tailor the infrastructure for them.

Still, there must be someplace where the smarter infrastructure could take some of the hard problems away for self driving cars and get a nice demo going. Some small town, large army base, some specific city hoping routes?

The vision of self-driving cars using special roads has actually been around since the 50s. Here's a commercial GM put together about it (admittedly using different technology at the time): https://www.youtube.com/watch?v=F2iRDYnzwtk

I can rember a local agricultural college in the UK (silsoe now part of cranfield) experimenting with self driving tractors back in the late 70's

In a somewhat tangentially related vein, Suncor Energy, one of the top Oil Sands mining companies in Canada, announced they're replacing their fleet of heavy haul trucks with automated ones[0] by 2020, in the artcle they interview a automated truck industry rep who basically calls the use-case 'simple':

>"Really, what the industry is doing here is picking the low-hanging fruit. It's a lot easier to design and operate an autonomous truck on private property like the oil sands than it is on a on a public road. And there are a number of motivations. Computers drive better and more safely and better than humans."

[0] http://www.cbc.ca/radio/day6/episode-238-charleston-s-ame-ch...

Exactly. Self steering sailboats existed before self driving tractors, but again... keeping a sailboat on a specific bearing is far easier than turning a tractor with any number of attachments towing behind it. Nobody came along and said what Deere did with tractors wasn't original because sailboats could auto steer. What about autopilot on airplanes? The comparisons are endless.

It's an interesting article on its own, but any claim to inventing self driving cars on the farm is rather shortsighted... every invention borrows from the past and contributes to the future. The same is true for art, literature, industrial design, cooking, medicine, and almost any other creation.

Lets not discount the invention of the self driving tractor, but lets also not discount the leaps that have been made since then in AI and computer learning for understanding road markers, traffic behavior, traffic signals, routing, crash avoidance, parking assistance, etc.

There's something to be said for starting with a simplified version of a complex problem. Driving on a road vs. a field is apple and oranges, but I'm sure it was a useful starting point for building road vehicles.

Exactly. Self-driving farm equipment is more in the category of the Roomba than it is in the category of self-driving cars.

A million dollar Roomba, but yeah.

More interesting is to compare the stakes. On a farm, you have a worst case of ruined farm land and equipment. On the road there are human lives.

I really would love to see how a Google car would handle chaos of third world metros, like Mumbai, Manilla etc...

I imagine a total fail. The car is programmed with an expectation of fairness - it assumes folks stop at lights, observer lanes and take turns at intersections, at least most of the time.

In those other cities the locals would learn to game the autocars, taking the initiative from them at every interaction, leaving them impotently idling at a busy intersection waiting for somebody, anybody, to give them an opening.

IIRC, they're working at "fair but not doormat" - yes, gameable, but not so supine as to wait until the rush hour is over before nosing forward to indicate a desire to turn.

Need to review films of rush-hour traffic in Delhi. Cars stream pell-mell through intersections as a living fluid. No such thing as nosing forward to indicate anything. Its a whole different world.

yeah but where can you buy a Google self-driving car? and how many other companies are also working on the problem? many.

This smacks of "Apple didn't invent tablets, Microsoft did!!" kind of thinking.

Who cares who invented it, especially if they were incapable of bringing the technology to the mass market themselves? The real heroes are those who make it useful for the mass market, not those who keep it on the 8th shelf in their "R&D Lab" or even those who manage to sell 10,000 units a year globally.

Also, as you say, there's a big difference between "self-driving farm vehicles" and self-driving road cars, just like there' a big difference between an unusable 2h battery life and $2000 XP "tablet" from the 2000 compared to the 2010 iPad. In other words, there was been a lot more "invention" added to the iPad since the Microsoft tablets appeared as well.

Who cares who invented it, especially if they were incapable of bringing the technology to the mass market themselves?

This would pretty much disqualify a large chunk of computing research (research in general, really) from being notable according to your criteria, as it wasn't brought to the "mass market".

The real heroes are those who make it useful for the mass market, not those who keep it on the 8th shelf in their "R&D Lab" or even those who manage to sell 10,000 units a year globally.

They're all important. The ones who keep it on "the 8th shelf in their R&D lab" are who laid the groundwork for the mass marketeers to prevail in the first place. The small-time sellers frequently come from hobbyist backgrounds who are attempting to distribute a more specialized (or more capable for complex use cases) product as opposed to tailoring for maximum units sold.

Those who bring it to the masses deserve praise in as much as the product has utility and how it improves over previous designs. All too often the result will necessarily have to be watered down so it can pass in widespread consumer circles.

Is Bill Gates a hero, but Gary Kildall a zero? Nonsense. Though that's what your logic ultimately concludes.

Quite. Yes, John Deere did create self-driving tractors designed to follow strictly specified paths on vast open fields having controlled access (i.e.: other vehicles & people could be denied entry to the area &/| informed of the situation) and well-defined path markers; this is a necessary but not sufficient precursor to developing a vehicle for use on arbitrary universal-access roads. Likewise, I was using tablet computers created by Wang back in the 1980s, ungainly monsters 3" thick with low-res monochrome screens and short battery life; this was a necessary but not sufficient precursor to the iPad. "Necessary but not sufficient" meaning "yeah, they made something like X way back when but somebody else had to do a whole lot more work to make it the viable & popular product we know today."

John Deere does sell self-driving tractors to a fairly massive market. Just because you aren't in the target market doesn't mean that they haven't brought the technology to market.

While there is a big difference between farms and roads, the tablet analogy doesn't make any sense.

The scope of hazard detection and avoidance is vastly different between a tractor in a field and a car on the highway. They solved a self-driving problem for sure but a different problem than the one of general purpose human transport.

Not the least of which is velocity. Tractors generally move under 12mph. And traffic density. The tractor is the only vehicle in the field, as a rule. And obstacles - there is absolutely nobody out in the field, usually. And noisy environment - a field is often table-top flat and empty of all obstructions. Kind of the definition of a field.

So the John Deere problem space was mostly dead-reckoning navigation to a matter of inches, to get crop spacing right. A very different problem from auto-driving cars.

> And noisy environment - a field is often table-top flat and empty of all obstructions.

Perhaps this is true in certain parts of the country, but certainly wasn't my experience growing up in the Southeast. The land there tended to be mostly rolling hills, basically no field was totally table-top flat. (And the 3rd paragraph in TFA mentions that the land there even in Kansas is hilly so he does the first pass).

Obviously you want few obstacles to maximize the useful land area, but fields often had ponds, creeks, patches of trees that broke things up, not to mention unknown obstacles like fallen trees or fences, wild animals, etc. that happen 'randomly'. I'm actually curious how much of this the software can handle on its own.

So it's certainly a different challenge, but to imagine most fields as perfectly empty, flat, continuous, homogenous spaces is to over simplify things.

Pardon my ignorance. I'm here in Iowa where many fields are most definitely empty, flat, continuous and homogenous. Not to mention very fertile, well watered and yielding among the best of any place in the world. Also the home of agricultural research including, I'm guessing, John Deere pilot projects.

Which biased my view. Interesting to know the technology can handle Kansas, which would be a challenge.

Exactly, which is why people who claim that Apple invented tablets still deserve to be corrected.

That's OK, I'm sure when you delightfully correct people about this oversight, you don't skim over things like the Newton or the Palm Pilot, or the prop from 2001 that was hugely inspirational, and that you go into great detail about all the technical and usability differences between iPads and early-2000s tablet PCs. I'm sure you give fair and equal coverage to all of these topics when you are correcting people, since you seem to care about history and stuff. Just giving you the benefit of the doubt here.

I bought a Compaq TC1000 when I graduated highschool for $750~. It was one of the very first "Tablet" PCs. It received a LOT of very positive attention when I used it on campus in college. It was a great machine and was very usable even with limited OS support.

Microsoft had a marketing problem (they still do) and not a usability problem.

Did Apple invent the usable tablet OS with a decent amount of software?

DARPA jump-started the industry with their three autonomous vehicle competitions a decade ago. Most of the entrants were university mechanical engineering departments. No one came close to completing the first challenge. But they got pretty good by the third one. Google and the auto indutry continued funding thereafter.

The DARPA grand challenge was certainly a pivotal event, but I wouldn't say it jump-started the industry. The two biggest teams in the second race, from CMU and Stanford, were both spinoffs of the Field Robotics Center at CMU, which had been doing this for a long time:

http://www.cs.cmu.edu/~tjochem/nhaa/Journal.html http://www.pbs.org/saf/transcripts/transcript705.htm

Disclosure: I was a grad student at CMU at the time, in the building, and saw a lot of this up close.

John Deere does not receive any credit for any innovations they may have made in the area of self driving vehicles until they acknowledge the right of vehicle owners to service their own machines and release all required documentation, source code, and diagnostic tools to facilitate this.

I have a hunch future retail self driving cars will have the same restrictions, unfortunately.

Who wants to actually own their own bare metal anymore? Why do you want to be tied to this specific server or car, this specific package of molecules?

The SPSD Car[1] will not be something that can be owned by individuals. You can use your smartphone to request a pickup, which will be fulfilled by any of a number of nearby cars. You'll lease a license to be transported in one car, and the app will automatically transfer that license to the car that's fulfilling your trip.

Your Leased License to be Transported in a Car will only apply for the duration of the trip, and gives you no rights or responsibilities to the underlying vehicle.

[1] (Solar Powered Self Driving)

I do. I want to store stuff in it (my assorted gear I want nearby wherever I go but won't be lugging around every time I leave the vehicle), I need specific features (say, child seats), I want to customize it (be it bumper stickers or off-roading bumpers), I want to know that there WILL be a suitable vehicle ready when I need transport ("whadayamean there are no vehicles available with two child seats? my girl needs a doctor NOW but an ambulance is super-expensive overkill!" or "it's 1AM and I have to wait 15 minutes for one to arrive? with my own car I could be home in 15 minutes."), and I don't want to deal with consequences of strangers driving it before me (everything from annoying seat settings to vandalism & bodily fluids).

Sounds like you're thinking of a self-driving taxi, overlooking the many reasons why people don't want taxis (not deeply relying on others being a huge one for many of us).

Yours is a use-case that I often forget.

For you, a car isn't just a conveyance. It's also a storage locker. And a second home, in some ways.

Such is a way of life for most American drivers. We spend a LOT of time in our cars. The freedom of "mine, now" transportation is...visceral.

I don't think anyone's trying to pass laws banning owning one's own vehicle once self-driving cars become ubiquitous. I imagine you'll always be able to own your own car, and store whatever in it.

(I also suspect your insurance will start to climb through the roof once self-driving cars do become ubiquitous, because computers will be able to be far better drivers than people, once the software's in good working condition.)

Who said anything about banning? Parent post just asked why anyone would want to own a car.

If 90% of the cars are taxis, there won't be much of an issue with availability. The cost difference should be huge too.

I do like driving around with loads of crap in my car though, I think some kind of standard secure box storage system would be helpful, a wheeled box you can lock up, like a bike rack, standard size so you can easily fit 4 of them in the taxis

We are a long way from that. In a cab, its somebody's job to keep it clean and safe. An auto-car may arrive with dog droppings, vomit, trash in the back seat with slashed upholstery and condom wrappers. How is that going to be managed?

I think this isn't such a big issue. I use Car2Go all the time, never had any of those things in a car I drove.

Maybe its different when you're driving, vs chilling in the back of an auto-car for a half-hour with your sweety. I think young folks will be very creative filling the time.

I think you're obsessed with this one exception. So I'll state your concern outright: How do you prevent people from having sex in an auto-car?

Well, how do you prevent people from having sex in rental cars? First, people that can get a rental car usually can get more comfortable accommodations for other activities. Second, the rental car is only in the renter's control for a short duration, and that duration isn't cheap. Third, the rental agency will hold the renter responsible for any damages caused by this activity.

I'm sure that Zipcar, Car2Go, Enterprise, Avis, and others have solutions for these situations.

Its not that one thing - its a stand-in for the issue of idle folks in a private space, that they don't own or have responsibility for.

There's no other transportation option in existence that combines privacy and anonymity without the requirement to drive. Nookie is only the beginning of what idle folks will get up to.

Rental cars are not the same - we're supposed to be using auto-cars as an everyday thing, to get to work or school or shopping or anything really. Not just on a business trip.

Actually, I can think of one similar place where people have time on their hands in a private space - public restroom stalls. And look how wonderfully that works out for everybody.

>n auto-car may arrive with dog droppings, vomit, trash in the back seat with slashed upholstery and condom wrappers. How is that going to be managed?

Civil fines.

Record the cars under use. When people don't clean them they receive a cleaning bill at the end of the month. When they don't pay it, they don't ride.

When the car is dirty have it dispatch back to the management facility for cleaning and insurance documentation.

It will be managed by whatever company owns the cabs and rents them out. They may subcontract that to car detailing businesses.

Sure, once a day or once a week. A cabbie is right on any problem. Or they don't get any custom. An auto-car could be disgusting for hours or days.

Or we can try to do OCR with baggies of trash and other refuse. When people get out, have cameras inside take a photo and somehow tell if the car is clean or not? Then maybe deduct a cleaning fee for leaving behind garbage?

The car could detect if there's garbage, and maybe use some kind of air sensor to tell if it stinks. If it's clean, it would go back into the motor pool, if not it would auto-drive to a detailer's.

(Who wants to help me start a robotic auto detailing company?)

Why focus so much on automatic detection?

Just do the simple way that Car2Go does: When somebody gets in, they give a "thumbs up" or "thumbs down" in regards to the car's internal cleanliness. If the car gets too many "thumbs down" ratings, then it gets put on the list to be cleaned.

How about if its just disgusting? Condom wrappers and stains on the seat. Hard to detect automatically. Cameras are missing the point - back to human monitoring which is far from automatic.

Treat any trash left as worst-case scenario. If it costs $1000 to clean a car when there is a bloody tissue left in the back, charge $1000 when there is a Snickers wrapper in the back. Not fair? Don't leave trash in the car.

Accidentally leave your wallet behind? $1000 fine! I'm sure drunk people will love it.

Yes, yes. Since we can't cover all edge cases right now, let's abandon the concept completely!

"Disgusting" is a normative, qualitative term. Can you put it into calculable, quantitative terms?

I fail to see how cameras are missing the point. Are you saying that all condom wrappers will be the exact same color as the upholstery? Are you saying the stains on the seat will be the same color as the upholstery? (Are they still stains if they colormatch the material?)

So we take some before photos, and do machine learning analysis. And we can have people rate the car, or even refuse a car, with the app. Yes, you'd have to have a motor pool with idle cars to get this to work correctly, but queue theory says you'd need idle cars during non-peak hours to meet peak demand anyway.

Whoa, sorry, didn't mean to push any hot button. Just exploring the problem space, no emotional content intended.

If a car service is automatic, there are no people guiding its normal operation. Putting cameras back in would mean, I thought, humans to monitor those cameras. That's adding back much of the cost of cabbies.

So maybe the camera can be automated too! That's a good idea. In fact, just their existence may be enough to quash most objectionable activity (snogging, urinating, allowing pet messes, littering).

In the end whatever the scheme has to be practical, not just possible. Its the details that sink projects. Like, how to keep the kids from using an autocar as a makeout pit. And so on.

Take a train in most of the world and it won't have used condoms, dog vomit or pools of urine. It may or may not have cameras.

Only Americans seem to have these problems with public transport — in which case, maybe the technology will be rolled out in Europe or Asia first.

A private car may be entirely different from a public train. Just thinking out loud. These auto-cars will be a new thing, dissimilar to what came before. Not an American thing (where did that come from?)

I wonder if most of them will have upholstery, or if they will have something a little easier to point a pressure washer at.

Wouldn't ZipCar or the smart car rental service have these issues? How do they handle them?

I use Car2Go (a car sharing program) extensively. I've never seen any garbage or body fluids. I once found a females sweater left in the back. I don't think it will be as big of a problem as you think.

Isn't that more like a rental? You're driving the whole time. Different from idle, unsupervised teenagers in the back seat for a half hour.

There's a passenger seat in a Car2Go.

They could just make the previous owner responsible for cleanliness (a fine or fee if dirty) and I'm sure it wouldn't be a problem.

transportation might well become a service -> you need a car ? use an uber-like app in order to have one arrive soonish. That would make that restriction moot.

That uber-like app will probably be Uber.

I understand where you're coming from and I think there is a middle ground, but there is certainly a decent rationale for locking it down: these machines cost $750K to $1M and enormously complex. Allowing an owner to inject some jury rigged code could cause hundreds of thousands of dollars in damage to the machine, or worse yet hurt or kill someone.

Honestly, what would happen if the implements you have hitched to the back of the tractor can't receive the proper instructions because the owner bought some counterfeit FTDI chip. You can set up your spray zones to make sure you don't spray pesticides near waterways, who is liable if that gets screwed up? The counterfeiter? The Farmer? Either way you know John Deere is going to get dragged through the mud for it.

The exact same argument could be made about cars. Or a house. Or any large, complicated and expensive modern thing, really.

I’m assuming that you don’t believe that similar restrictions are reasonable on cars or houses, etc. So what makes a tractor so utterly different? Is it more dangerous? Cars can certainly be more dangerous. And a house? Where people sleep? Wouldn’t that be even more important to restrict people’s ham-fisted “repairs” or the like? But we don’t. So why are tractors different?

Cars are similarly locked down, as are many (most?) consumer electronics devices. Buildings have an even more complex system of restrictions placed on them, called fire/building/electrical codes, albeit the enforcement is more nuanced.

If you want to advocate for a more open system of published regulations and a licensing/insurance system for professionals that's fine, but don't try to pretend that other facets of life aren't equally regulated.

But we do have similar restrictions on a house--you need licensed electricians, etc., as well as approved building permits to do many tasks.

This is an argument against allowing self-service, but it isn't an argument against opening things up and allowing certified third party service.

Preventing counterfeiting by not releasing source/documentation is a pretty textbook case at an attempt at security through obscurity.

Couldn't agree more; it's pretty insidious. To be fair, the other systems we've installed are similarly encumbered (at least so far).

That makes no sense at all.

You mean how Apple similarly doesn't deserve credit for the iPhone and iPad until they open up the platform?

How do we make this ideal compatible with the often-stated claim that, for insurance and legal purposes, the manufacturer should be liable for the vehicle? A car you can service only at the cost of losing your insurance and legal protection is practically a non-serviceable car.

I'd say Daimler and co. got there first:

https://en.wikipedia.org/wiki/Eureka_Prometheus_Project

Google has obviously taken taken that a lot further (no driver reacting every 9km)

> "We kind of laugh when we see news stories about self-driving cars, because we've had that for years," Poole said.

Yup, no wonder he is a farmer (to be fair: if you're running a tractor at 2 am you may well imagine there to be heavy traffic on the field...).

Daimler is also doing quite a bit with self driving trucks. I'd guess they have to be a leader in that domain. Admittedly they weren't really on my radar until the recent news about their tests in Nevada.

From a layman perspective, it sounds much easier to automate a vehicle moving in a field, where you have a low chance of hitting a pedestrian or a cyclist.

To be fair, they also started working on the problem before Google even existed, with technology more primitive than Google's first servers.

Yes, I know of professors who were doing field trials of self-driving tractors 30 years ago, and I'm sure they weren't the first.

The mining industry has dreamed of self driving vehicles to replace humans in dangerous work conditions for ages.

Nothing new here : http://mining.sandvik.com/en/products/equipment/mine-automat...

It is, although there are issues that self-driving cars don't have to deal with, like muddy conditions on hills, lining up rows (tricky on head rows), speed-matching during picker unloading (e.g., tractor and truck working together to evenly-distribute the unload), etc.

Self driving cars absolutely do have to deal with varying friction of different road surfaces and conditions, lane alignment, speed matching (or rather, not bunching), overtaking. There is a lot of overlap in the tasks. The only one you mention that is unique is unloading while picking.

Yes, but traction on pavement is quite a bit different than traction in mud on a 20deg slope with a Really Heavy Vehicle. There are varying degrees of "uniqueness". My point, which I think you deliberately "missed", is that looking at it from the "layman's perspective" is flawed.

(And lane tracking is generally more forgiving than row tracking, (except for on the Garden State Parkway) in the sense that a few inch deviation while driving is acceptable, but cumulative parallel row error can add up to quite a bit, both time and money.)

On the other hand, roads are designed for cars. In fields, it's hard for a computer vision system to distinguish between a thicket of brown grass (can easily drive over) or a pipe (will damage you if you try to drive over.) They're different problems, I don't think one is obviously easier than the other. And there are certainly people in fields.

Driving a car is a social exercise. The challenge is making something navigate through human chaos, not make it move.

Even the article itself hints that comparing the two is kind of stupid: "[...]where there are few pesky pedestrians or federal rules to get in the way." and then immediately proceeds laughing at the general population ignorance because farmers did it first.

I remember when I was a child, my father explained to me that depressing the right pedal makes the car go forward, the left one makes it stop and this round thing makes it turn. "WOW! Easy!" I exclaimed, "I know to drive now!" Yeah right.

Elevators are automated. Subways are automated. Even mother fucking planes can fly and land themselves (with the right equipment). Yup those giant flying monsters containing hundreds of human souls got automated before our cars.

But wait, Google is stupid because it didn't do it first.

EDIT: missing sentence

How about we all just agree this is a click-bait article and we've all gone chomp on the hook.

I don't want to write 'who cares' but realistically neither John Deer or Google did. There were gifted individuals at both companies who wrote the code and developed the hardware and one day someone will write a book and it might be an interesting read.

To argue for JD or Google is just playing into the marketers hands.

/rant

I think a lot of the commenters here are correct in pointing out that the problem John Deere have solved is very different from the problem that Google, Tesla, Uber are trying to solve.

That said, I think there is a "human acceptance" piece necessary for autonomous vehicles to be allowed on roads. A lot of people are going to fear, dislike, or be otherwise opposed to sharing the road with autonomous vehicles. A lot of people simply don't believe that it is possible for an autonomous vehicle to handle certain driving situations. To be able to point to tractors driving themselves around fields for the past X years is valuable in getting humans to accept robotic cars.

It's useful but I guess people will just raise that same reservation - just because they work in the field...blah blah blah.

My guess is that insurance will force people's hands in the switch. Most accidents on the road are due to human error (I assume, or rather had there not been human involvement the situation could have been avoided). Let's assume that self driving technology reaches the safety level that most of us here believe it will. Insurers, hedging against risk will have to financially penalise manual driving, since it will carry a much higher risk.

Hopefully that takes the irrational opposition out of the equation and let's the raw statistics take charge :)

Self-driving tractors are incredibly dumb. They just follow a predefined path using GPS. Most don't even have any obstacle sensing. They probably should at least have a basic obstacle detection radar, but at present, they don't.

I saw the original John Deere differential GPS prototype tractor in use at Stanford many years ago. There used to be an empty field next to the Stanford horse barn where they tested. It's now a parking lot.

This article is giving credit to a market leader, but I'm pretty sure there were a number of GPS technology companies and precision ag startups that were working on automating farm equipment first. John Deere has many times been a late adopter in terms of technology, waiting to see what pans out in the market, then relying on the penetration of their dealer network to distribute proven technologies to a much wider audience.

Yes, this a ridiculous PR puff-piece. JD saw something coming and got into the business along with a bunch of other companies. There are a lot of competing GPS technology companies that make products just as good if not better and sell well in the absence of the advantages that JD had from the start.

Hemisphere GNSS, whose (Outback) products I use, has a superior product in my opinion and started a few years before Deere got interested in the business at all.

Or just journalistic naivete and laziness of not thoroughly researching a sector. It looks like the author normally covers consumer technology, so she probably doesn't know much about the agriculture or GPS industries. I would give her the benefit of the doubt of not knowing (which in journalism isn't really an excuse), but either way it's bad journalism.

True of 99% of journalism, you know if you read articles about your field of expertise how wrong they are, they're just as wrong about everything else.

How about no.... Driving a tractor through a predetermined path in a predictable pattern is no where near the same as driving on our roads with human error and the like. Also I'm not about to get behind a company that won't let you service your own machinery.

Instead of self-driving cars, I'd say focus on eliminating the need for "going somewhere"/vehicles in general. Would be so cool if we could just 'beam' in to work. 3D virtual presence. But hey, its not my money they're investing.

How about 3D bio-printing you at work? Oh but at the end of the day, do you get rid of the copy, or get rid of you? Could be some trouble either way.

John Deere didn't invent the technology though, Autonomous Solutions Inc. (ASI) did. They partnered in ~2000.

http://www.asirobots.com/news/

http://www.asirobots.com/farming/

http://www.asirobots.com/products/mobius/

ASI were hardly the first, and Deere had their AutoTrac system before they started working with ASI.

> The Beeline Navigator was developed in 1993 by Australian engineer, Rob Mailer. [1]

[1] http://www.queenslandcountrylife.com.au/news/agriculture/mac...

Deleted comment

I don't have any inside knowledge of the Deere system, but I'd be very surprised if it was. My guess is that most of the Deere stuff comes NavCom. That is certainly the case for their GPS system.

> "NavCom's involvement with Deere & Company began with Deere's GreenStar precision farming system in 1994 and has expanded into many areas of Deere's product activities. The commitment by the most senior Deere management to information systems leadership motivated the NavCom shareholder-employees to join with Deere & Company,"

http://www.prnewswire.com/news-releases/deere--company-acqui...

You sure? Satloc was doing GPS guidance for aerial guidance for decades (crop dusting planes). They adapted it for ag guidance in the 1990s I believe. This technology has been evolving for years and saying that John Deere or some single company invented it sounds like a gross oversimplification.

Comparing this tractor to a self-driving car is like calling someone an ace pilot even though they don't know how to take off or land.

Spot-on but unimportant.

"Apple didn't lead the mobile device revolution. Blackberry did."

Technologies succeeding in the gauntlet of consumer electronics makes a bigger splash in public media---for obvious reasons---than those same technologies applied to the problems of specific trades.

In the most recent user guide (2013), they talk about "mapping".

https://stellarsupport.deere.com/en_US/support/pdf/om/en/PFP...

You can record a map with the GPS and replay it. It's a little bit like waypoints for drones.

There are other companies that make driverless tractors:

https://en.wikipedia.org/wiki/Driverless_tractor

It also reminds me the machines in Interstellar (the indian drone that can fly 10 years and the tractor).

Not actually. John Deer doest not build vehicles, he builds machinery.

John Deer machines work in a strict set of environments, and although they may be very sophisticated and react to a range of adverse situations, these situations are still much more predictable and the route much better known in advance than roads and streets.

If the autonomous vehicle problem were dealing with different roads, pavements and other pre-programmed vehicles, the self driving car would be a thing from the 80s, together with the robot car welder.

Will someone who works at Amazon and reads HN please ask Jeff Bezos to tell the Washington Post to grow out of this smartass fake journalism? Most people are well aware that agribusiness depends heavily on automation. Most people are equally well aware that it's a lot easier to drive around a field when you are the only thing in it than to drive on a public road. That's why you only need a driving license for the latter case.

The entire article is one giant straw man fallacy.

> Most people are well aware that agribusiness depends heavily on automation.

While it doesn't necessarily invalidate your criticism, this couldn't be less true. Most people in the Washington Post's reader base have exactly zero awareness of agriculture, full stop. Food comes in a box they buy at the supermarket, and that's the extent of their knowledge. If they have any thoughts beyond that, it's probably either an understanding of farming they got from a quaint children's book, or some kind of poorly-reasoned Monsanto/GMO FUD they heard at a fashionable dinner party. The idea that they know what kind of ag equipment is on the market, what it does, how it works, or what it costs is ludicrous.

Last time I checked I didn't see any farming in my local main street. Not sure to what extent technology from John Deere can be used for public vehicle. May be not much.

What's especially laughable is the article's mention of robotic lawn mowers. I happen to have bought one today, and those things (although admittedly there are models that are more intelligent) just ride around randomly until they hit a hard object or the perimeter, which is signaled by a wire that is dug into the ground. An 'autonomous vehicle' this is not (or only for the most basic definition of the concept).

The way robotic lawn mowers work sounds awfully like the beginner LEGO Mondstorms projects.

Yeah, might as well compare it to a Roomba too.

This is an apples to oranges comparison of technologies.

>But another is pure necessity: There's a labor crunch in rural America - young people move to the cities, leaving the average age of U.S. farmers at 58, according to the Department of Agriculture. Similar forces are pushing self-driving tech into other industrial sectors at a pace that outstrips the consumer market. >

Oh sure, that's the reason. All the job seekers dried up.

Many years ago, I dreamed of huge roadways with wi-fi embedded in the road. Cars would broadcast their destination and the network would adjust the speed of all cars to facilitate a smooth entry to the freeway and subsequent navigation.

Self-driving cars are really cool. However, if every car is trying to optimize for itself then traffic will still be an issue.

Forgive my laziness, but the last time I did any long research about self-driving cars, I understood that the problem of night-time driving and driving in precipitation still hadn't been solved. Anyone know if this is still the case?

It is more comparable to automated lawnmowers, than it is to selfdriving cars.

So did any of Google Moon-shoots actually had ANY impact on society so far besides being a cool news story? It looks like most of their cool projects got to the point, where they stale and cannot move ahead.

To be more clear, here's what JD equipment doesn't have to worry about that Google's does

- Pedestrians

- Other vehicles and obstacles (and automatically detecting them)

- Roads

- Traffic laws

JDs tractors are choreographed, it's not really 'self-driving'

These systems have no sensors beyond GPS and IMU, and typically only control the steering.

Typically they are following a predefined path, not even having the concept of where the end of the row is, relying on the operator for turning and avoiding any hazards. The most advanced functionality is Machine Sync [1], which is still only using the GPS to synchronize the machines.

Just because the functionality is basic, doesn't mean it's not useful. We are quickly coming to the point where these systems will be installed on every machine. The reduction in fatigue and other benefits are profound.

Another point to note is that getting to this point was a marathon, not a sprint. These systems have been around in one form or another for more that 20 years. Acceptance came in large part with the generation change of the owners. It will be interesting to see if the rollout of self-driving cars follows a similar trend.

[1] https://www.youtube.com/watch?v=GFhuJl3hYm8

Or even speed. What's the average speed of a tractor in exercise ? 15 mph ?[1] anticipation need, which was already pretty low in a field, gets even closer to zero.

[1] quick googling says some countries impose speed limit of 20mph but I don't know if it's on public roads and not private land. http://www.fwi.co.uk/machinery/road-rules-for-tractors-in-th...

Probably on public roads, speed is lower when harvesting

So I fail to see what Deere did. A 10mph, pretty stable, on mostly flat (curved, but locally flat) land, with almost nothing of its size to interfere (won't a beef move away if a tractor aims at him ?), except avoiding the occasional tree .. I fail to see what problem they had.

also cats

have seen a cat picked up by a corn harvester

Yeah, but I guess even a human has difficulty picking those in a place with tall vegetation

Lots of small mammals die this way :(

Self-driving tractor in 1978: https://www.youtube.com/watch?v=_3xS3eIHzLk

Sure, but none of the engineers who work for John Deere would ever be allowed to publish their developments, nor would they get any public recognition for them.

Time to buy $DE

Reply on news.ycombinator.com