Hacker News new | past | comments | ask | show | jobs | submit login
Unmanned grocery delivery is underway in Arizona (detroitnews.com)
121 points by evo_9 6 months ago | hide | past | web | favorite | 139 comments

I think this is one of those examples of how unless regulated there will be 10x more cars on the road if/when self driving cars are ubiquitous.

Lots of people envisioned less cars because the with self driven cars the current use of cars can be covered by less cars but self driven cars enable more uses. Send your kids to their friends. Grandma goes shopping. Need to borrow a drill pop it in car. Want to run a local cake store. Deliver the cakes in the cars. Right now in my city there are billboard trucks that drive around. I can imagine 10x more of them. And of course with no driver and lots of competition prices will come down. Why take public transport when I can be delivered door to door for less.

I enjoy public transport here in Tokyo and I'm happy for the excersize I get walking to and from the station but, for example my office is a 20 minute walk from my apartment. I usually walk it but if it's raining or too hot I've used the Taxi app to have a taxi pick me up and drive me that short distance. The cheaper, easier, and faster (time to pickup) things get the more likely I'll take advantage of that more often.

It will be only a matter of time until we have autonomous vehicle spam: cars with billboards driving around (unless stuck in traffic) simply to be seen.

It could be a huge problem because of competing goals. The spammers won't care about traffic flow as their only goal will be to have cars on the road and visible, whilst people in cars will want to get places in a reasonable time.

I hate to tell you.....this is already a thing with drivers. Las Vegas is infamous for this, and I have seen them in other large cities too. And the issue you describe happens as well. The Car with billboards want to drive as slow as they can for that reason, and it directly competes with folks who are trying to get from point A to Point B

Fortunately the regulatory way out is easy. Forbid ads on vehicles.

Like all politics, it's not so simple. What qualifies as an ad? Should Walmart and Amazon discontinue their corporate logo (and sometimes produce) on their semi-trucks? Many large companies put a lot of thought into the branding on their vehicles: a law forbidding that will not get passed without a fight.

Here's some real-world examples:




Respectfully, I think that would cause a lot more trouble than the problem. Would that mean the license plate ads are illegal? Or the Dealers stickers? Or would branding for cars be ads? And many commercial trucks have ads for their trucks. While I despise seeing those Ad trucks, I think making them illegal would cause more problems than it solves.

Just Dutch auction off the limited quantity: road space. Modern technology allows ubiquitous tracking and recognition of vehicles. No reason to subsidize at that point. We can just charge everyone

We already have a system to turn road usage into government revenue. Collection is easy efficient, and wastes zero time for the people paying. It does a good job correlating not only to road use but also pollution produced. In fact it's so easy to pay that many consumers rarely even think about it when they're paying it, yet even anonymous cash payers can't find a way to skirt around it.

It's called the gas tax, and it does a good enough job charging people for their negative externalities that everyone seems to hate it.

This would be true if fuel usage was linearly related to wear and tear on the roads, or road usage.

Fuel taxes almost never tax heavy usage vehicles enough, compared to the amount of damage they do to the road.

A fuel tax also doesn't solve the problem grandparent post is trying to solve, excess usage of 'high value' roads.

If road usage, and road wear and tear, was evenly distributed then a fuel tax would work, but the reality is that is simply a convenient and fairly effective proxy solution. As technology improves a solution similar to grandparent, some form of usage tolling, will become feasible and provide a more effective outcome.

I don't think it's wise to encourage even more tracking of people/objects moving around in the real world.

it probably wouldn't be that bad if government could track autonomous semis making deliveries to large shopping centers, but I wouldn't want it to have fine-grained data on who goes where and who delivers to their house.

That's easy enough to solve, simply ban advertising on vehicles. We ban billboards or regulate signage in many communities, so there's certainly precident.

Very few communities in the United States put quality of life ahead of the needs of businesses.

I think the "Rebound effect" may help you in your argument. https://en.wikipedia.org/wiki/Rebound_effect_(conservation)

> The cheaper, easier, and faster (time to pickup) things get the more likely I'll take advantage of that more often.

I think this balances out naturally though in (y)our situation -- there's limited space for taxis near the station, so if a lot of people are taking taxis, the benefit of taking a taxi approaches zero unless the benefit is "sit" rather than "save time."

A good example of this is rainy days or typhoons.

So in the other cases you list, I wonder if it might not be self-balancing there as well -- as the roads get congested, the benefit of popping a drill into a car to send to a friend or whatever will approach zero as the costs rise due to congested roads.

I think you're saying the problem is as more people take taxis the lines for them get longer at the station. But I don't need to wait in that line if I'm going door to door.

As for congestion, if all the cars are AI maybe they can pack a few X more cars on the streets? Not saying I like that idea

> I think you're saying the problem is as more people take taxis the lines for them get longer at the station. But I don't need to wait in that line if I'm going door to door.

I'm saying that the immediate area around the station gets incredibly congested, so it's difficult to get close to the station in anything that approaches a timely fashion. It prevents "door to door" since it takes longer to get to the station "door" with the congestion than it would have taken to simply walk.

But I have no desire to go to the station door. I want to go to my destination. From my door to my office door. Or from my door to my friends door. I have zero interest in going to some station to use an outdated mode of transport when I can have a modern and cheap AI car take me directly to where I'm trying to go

My bad; I completely misread your initial post as wanting to replace a 20 minute walk from your house to the station with a taxi.

Yes. Roads widened -> traffic increased.

Also: when you drive, you probably schedule a bunch of errands to do on the one trip. Soon, you mightn't bother arranging to avoid this bother. Corollary: Then, why collocate businesses? This could be the beginning of the end of the mall.

> This could be the beginning of the end of the mall.

Malls started dying a long time ago. Articles like this[1] have been written for years.

[1] http://time.com/4865957/death-and-life-shopping-mall/

>Roads widened -> traffic increased.

This is only true when there is unmet demand. I have a lot of roads to show you in Western Nebraska that won’t get increased traffic no matter how big they get.

Rather, it's true when the existing roads have a lot of traffic.

The reason why widened roads increases traffic isn't just unmet demand; demand actually increases. The extra capacity means that people can shorten journey times by using the road, and they then locate their homes and businesses at either end or along the road. The road itself increases demand.

I think the OPs point is that this is not true. Roads do not ever, themselves, increase demand in any way. (See any rural area anywhere, full of roads way under max capacity).

> The extra capacity means that people can shorten journey times by using the road, and they then locate their homes and businesses at either end or along the road

Exactly. So if people are shortening their journey times using the road, then whatever previous journey they took no longer exists. The new road did not increase demand, it relocated already-existing previously-poorly-serviced demand to a new more efficient route.

This is generally a good thing, it's what a successful road project is intentionally supposed to do, and why they are worth spending money on.

Exactly. So if people are shortening their journey times using the road, then whatever previous journey they took no longer exists. The new road did not increase demand, it relocated already-existing previously-poorly-serviced demand to a new more efficient route.

This is a mendacious misphrasing of what I wrote.

People move. Businesses move. People are born. Businesses are created. There isn't a static amount of demand. People make decisions based on costs; whether they create a business, whether they have children, where they live, where they work. Roads change costs. If you have a clue about economics, you'll know that when costs change demand changes in all elastic markets.

Roads don't simply move demand around. People make more use of roads when they're cheaper to use. In other words, demand increases.

If you don't understand this, please consult an economics primer. Your argument seems to come from the same place as the ignorant people who think immigration lowers wages.

It is you who needs an economics primer, specifically in supply and demand curves.

People are willing to pay a certain amount of time to drive on a road. This is your demand curve. Everyone willing to spend any amount of time driving makes up all of your demand.

Your supply curve is the speed the road can provide. Increase the average speed by removing stop lights, decreasing congestion, etc and you capture more of the demand.

The only way a road increases demand is if people move to a location they refused to consider before the existence of the road. However, non of the analysis on induced demand for roads analyzes this (that I’ve seen on the academic side). All of it is mistaking capturing more of the demand curve for increasing demand because the effect is always immediate.

That’s not increased demand, it’s just meeting demand that was only willing to pay a lower price in the form of time spent on the road.

Not quite.

The unmanned delivery vehicles will be followed by a “shadow car,” which will be driven by a human with the ability to stop or control it.

There was a Silicon Valley startup called Starship Technologies which had a little robot vehicle running around downtown Redwood City sidewalks making deliveries. It was always followed by a guy watching it.

The guy following it (and also remote "pilots") are/were part of a training program for the eventual deployment of self driving delivery robots. The startup is from Tallinn, Estonia, with offices in several locations

The same guy for four months?

I sense the plot for a tech drama, when he gets too attached but the project is axed...

I saw multiple people in both Redwood City and Palo Alto.

same thing in san jose, they eventually did go fully autonomous this summer though. those dumb robots used to get stuck on curbs all the time.

Yeah I don't think we have had any truly autonomous land vehicles released into the wild yet. Any company that claims to do it had better provide proof.

That is an excellent point. Beyond a very few fixed route systems, there's not much. Anyone know of anything?

Fifteen years ago I ran a DARPA Grand Challenge team. At the time, I thought about commercial applications. Starting at the slow-speed end looked better from a safety perspective. Most problems below 25MPH can be dealt with by slamming on the brakes.

The first area to look at was rental car return. When you're done with the rental car, you get out at the terminal, and it drives itself, slowly, back to rental car turn-in. When you order a rental car at the airport, it drives itself slowly to the terminal and picks you up at the curb. Then you drive away manually.

The numbers didn't look good. You're competing with minimum wage labor, and the hardware costs too much. That's still the case. Also, back then, everybody was so paranoid about airport security that getting this past TSA would be really hard.

Second area was shuttle buses. Parking lots, campuses, amusement parks - all slow speed applications. Local Motors and Navya are doing this. Navya had a backup driver on their Las Vegas shuttle bus for the first year or so. Not clear if they finally got rid of them. They have several test installations, but it's not clear that they have trustworthy real autonomy. They have a demo route in Switzerland, but "VBSH has in addition hired and trained three attendants who are always also on board the bus during trips." Their CEO was just fired for underperforming.[1] Local Motors does demos, but still seems to be searching for a serious customer. Baidu has a self-driving bus program, and they've built about a hundred of them. Not clear if they need a "safety driver".

It's rather disappointing that even the slow speed shuttle buses on fixed routes are not yet really autonomous.

[1] https://venturebeat.com/2018/12/19/navya-ceo-fired-after-big...

What definition of autonomous? Various military groups have deployed aircraft that do indeed operate totally independently, at least in some circumstances. A Global Hawk drone doesn't fall out of the sky if disconnected from controllers. It can continue flying totally on its own, as do cruise missiles. Lots of autonomous boats (small ones) have crossed oceans. A modern torpedo is essentially an autonomous submarine. But no, nothing on the streets of SF.

I think the grandparent forgot to quality autonomous with safe. John Deere (My employer) has been doing autonomous tractors for more than 20 years, there have been videos of dogs driving tractors around while the farmer sips lemonade in a lawn chair. John Deere lawyers will use any legal trick they can to get them taken down though because it isn't safe to not have a human sitting in the tractor watching for trouble (many videos remain because there isn't much to go on). The tractor could kill someone and keep right on going with no care so it isn't safe. Likewise drone aircraft can go one for a while because presumably any other airplane in the area will avoid it. Boats can cross the ocean because odds are it will not encounter another boat (relative to the size of the ocean).

What is hard about self driving cars is we want our kids to be able to ride their bikes on the same streets those cars drive on. Limit self driving cars to just hiways with only other self driving cars and we could have had them 15 years ago and there would be near zero accidents. Dealing with human drivers (who might not be sober), kids, wild animals, and the like is hard. (I'm sure those in self driving cars have a much longer list of things things that are hard)

Some of the autonomous boats know to keep away from the bigger ships, the ones with transponders. Some airborne drones will react to TCAS alerts (Traffic Collision Avoidance System) and change course/altitude in the same manner as a human pilot would. I would assume that the largest military drones are capable of flying home, or at least back to friendly territory, if they are disconnected from controllers.

There are two hard parts of hazard avoidance: identifying the hazard in the first place, and dealing with large numbers of different hazards moving around. Transponders, TCAS, and maps solve the first problem well enough for boards and airplanes.(there are time where they fail but they are rare enough to ignore). Except at ports/airports traffic is generally light as well meaning drones and boats only need to worry about a small number of hazards which makes this easier.

Autonomous cars are hard because children, and wildlife do not have transponders, and worse have a habit of jumping out from cover right into your path. Killing children is obviously bad. Even though who morally would argue that wildlife don't matter still do not like the damage hitting the wildlife causes to the car.

The parent did specify land vehicles.

What are the benefits over the traditional safety driver on standby whithin the otherwise self-driving car?

I can see that this is more future-compatible, the "drone" is smaller than a car, so, when you can lose the driver you will be able to shake off the whole shadow vehicle and have less mass overhead.

Edit: also, PR. The pretty pictures of the obviously driverless vehicle are almost inherently misleading.

Anyone can see anything else?

When I read the Bob Shaws book "Vertigo" in 1979 about personal flight-vest, I designed navigation rules of all aerial vehicles big and small. It is as follows:

If the vehicle is X meters from the ground, the only allowed direction is (X mod 360) degrees, and the only possible speed is X/10 meter per second. Hence the vehicle moves along helical path until it reaches the only height where the desired speed and direction is allowed. There cannot be any major collisions only some slight grazes.

Ingenious, nicht wahr?

The problem of what is the "ground level" and how passenger jets behave is beyond this brief introduction.

The problem with your proposal is several-fold: First, your hight from the ground changes very rapidly as you fly over terrain - you don't want to have to jot to the left every time you fly over a canyon. Second, it would require that you always take off and land pointed to the north, most airports would need to be reoriented, and you wouldn't want to land or take off when the winds are blowing from the south. Third, accurately measuring your distance from the ground is somewhat challenging, and the technology for accurately measuring the course you're tracking has only somewhat recently become available. Fourth, flying craft have maximum and minimum speeds, sometimes very close together, and maximum altitudes, so creating too tight an artificial coupling between speed and altitude can be very limiting.

In the US, there's a similar, but much more relaxed, system in use:

If you're more than 3000 feet above the ground, then the altitude above MSL in feet you should be at depends on whether you're flying under Instrument Flight Rules (IFR) or Visual Flight Rules (VFR) and your heading, as follows:

- For VFR with a magnetic heading 0-179: (1+2n) x 1000 + 500 (for integer n)

- For VFR with a magnetic heading 180-359: (2n) x 1000 + 500

- For IFR with a magnetic heading 0-179: (1+2n) x 1000 + 0

- For IFR with a magnetic heading 180-359: (2n) x 1000 + 0

Magnetic heading is trivial to measure, and under the specific rules used to determine MSL altitude, thats also really easy to measure (but might produce a large error relative to the correct answer, but a small error relative to other aircraft).

In addition, below 10,000 ft, the speed limit is generally 250kias or 200kias, depending on how close to the ground and other airports you are.


> There cannot be any major collisions only some slight grazes.

There's not necessarily much difference between those two.

Pretty ingenious, might steal this idea at some point in the future.

So the idea is that, in a given XY plane, all equilibrium flows are at the same direction and speed. You can never go exactly 26 m/s due north, but you can get close enough. By definition, vehicles in the equilibrium state can't hit other vehicles on the same XY plane, and vehicles in equilibrium at slightly displaced XY planes will only hit others at arbitrarily small angles. So far, so good.

Going up and down is simultaneously handled in the XY plane, because at a given height, a vehicle's direction of travel is tangent to the helix and going along with the flows. Also good.

There needs to be a strict speed limit on vertical speed. A vehicle hitting you from underneath or above at 30 m/s is pretty terrible.

A representative table for the lazy

  Height  Direction  Speed (m/s)  Speed (mph freedom units)
  0       N          0            0
  90      E          9            20.1
  180     S          18           40.3
  270     W          27           60.4
  360     N          36           80.5
Though I might make the coils of the helix a bit tighter, so you can get within 5 mph of your desired speed at whichever direction you want to go.

Other problems: - It may not be possible to arrange landing areas in the desired direction - Vehicles may not be able to make sufficiently tight turns - Fixed-wing vehicles cover quite a large vertical distance, especially when rolling as they turn due to wingspans. There's plenty of room for converging paths, and your rules may not allow enough leniency to avoid such paths. A "slight graze" that takes off half a wing is not acceptable. - Parallel lines converge on the surface of the world. Two planes flying due north (X is a multiple of 360) may collide at or near the poles.

> Parallel lines converge on the surface of the world

The scheme can be trivially modified to deal with this, once you've figured out a way to create a smooth tangent vector field on a 2-sphere.

Choose the points of non-smoothness to be in the Bermuda triangle. Everyone knows it's not safe to fly there.

The Bermuda triangle is an extremely heavily traveled area.

I have to imagine that one of the perhaps unintended side effects of lots of driverless vehicles buzzing around is that they'll act as a kind of traffic calming mechanism. I assume these things will have to obey speed limits and traffic laws to the letter.

The biggest reason I support self-driving cars is because on a bicycle you really see how often drivers roll through stops signs, don't use blinkers, and break traffic laws.

100% selfish reasoning, but less human drivers makes the road safer for cyclists (assuming robots use their blinkers).

I would like to see self driving bicycles, for the same reason ;)

Not trying to get into a bike vs car debate, they're all fine with me... but man some bikers I come across are just flying through intersections / stop signs too. Scares the hell out of me when I'm driving.

No disagreement here, plenty of idiots on bikes too!

Yeah, there are absolutely assholes in every mode.

It's just that assholes in cars are a lot more dangerous than assholes who are biking or walking.

Also roads are primarily designed for cars. You have to adapt the rules for bikes.

Maybe where you live. Roads can easily be designed or redesigned for other modes of transport, like bicycles. We've been doing that in the Netherlands for decades, with great results for road safety.

I'm sure you are aware that the Netherlands is more or less unique in having good cycling infrastructure.

It took us decades though. Any country can do it.

Yes, although I'd argue it's less about "rules", per se, and more about just having different infrastructure (protected bike lanes and separated bike paths).

They are equally dangerous in the sense of creating a potentially fatal incident.

No, not exactly. It's cars that generate the danger. When cyclists are on separated bike paths, the chance of a fatality is almost zero. When cars are in their own separate system (freeways), the chances of a fatality are still fairly high.

This is a relevant distinction, since it informs how you handle reducing the danger (e.g. road diets and traffic calming, at least for urban areas).

No, it's cyclists that typically die in a collision with other vehicles. But the behaviour of both contributes to the risk of accident.

I've seen cyclists drive the opposite way of the roundabout (at night), not using the lights. Breezing through crossings on red or where they had to yield. Seen them cutting into a traffic despite a broad, elevated, well maintained bike path along the same motorway. Taking pedestrian crossings without unmounting, and on and on.

So sure if the cyclist is not on the motorway the risk of fatality is indeed reduced - but not eliminated, as they still have to cross traffic. Pretending there is no responsibility on cyclists side is not solving anything.

> No, it's cyclists that typically die in a collision with other vehicles.

Yes, but it's cars that are generating the danger.

Suggesting otherwise is like saying that if a bulldozer is in a park and runs over some toddlers, the toddlers were the real source of the danger for being so darn vulnerable.

Besides, cyclists themselves are not really any different from pedestrians in this regard. It's just that there are physically protected walk lanes -- aka sidewalks/pavement -- nearly everywhere you go, whereas physically protected bike lanes are a rarity. Imagine if, for walking around, you only had "painted walk lanes" on the road, right next to cars going 40 mph. That's what it's like being on a bike.

> So sure if the cyclist is not on the motorway the risk of fatality is indeed reduced - but not eliminated, as they still have to cross traffic.

That's what protected intersections are for, and they help pedestrians too!

> Pretending there is no responsibility on cyclists side is not solving anything.

Of course, like all road and street users, cyclists ought to be responsible. But the primary problem here is extremely poor infrastructure that is both inherently dangerous and encourages bad behavior (and to a lesser extent, poor enforcement that lets off drivers easy when they seriously injure or kill other people).

On another message board, there was a Dutch dude who moved to California. He said that at first, he was horrified by the behavior of cyclists in his new home. Then two weeks later, "I was one of them". He didn't suddenly become an irresponsible jerk, it's just that in America, the road system encourages, if not outright requires an aggressive attitude from cyclists.

I have personal experience with this since I went the opposite direction, from California to Munich, which is much more bike-friendly than any major city in America. Here, the system is reasonably respectful of cyclists, and so cyclists generally respect the system in turn. There's no need to bike around like an aggressive asshole, because you can get around just fine like a normal person.

Contrast that with the SF bay area, where I got hit by cars twice the last year before I left. One time was with my son on my bike too, and the cop that came out didn't even give the driver a ticket for t-boning me as I crossed the intersection. And keep in mind, the SF bay area is actually a very bike-friendly area by US standards.

We should just make walking automated only, never have to bump into somebody or be stuck walking behind them ever again.

I always love the "but we have to maintain momentum" argument. A car also has momentum. In fact, it has quantitatively more momentum.

> I always love the "but we have to maintain momentum" argument. A car also has momentum. In fact, it has quantitatively more momentum.

But it's easier (in the sense of human effort expended) for a car than a bike to regain its lost momentum, right?

But that’s more of a personal problem isn’t it? No one forced that mode of transit, and you don’t hear joggers or skateboarders complain.

One can’t yell, “We are traffic!”, but then cry about how you want to be excempted from traffic laws you don’t like. I hate red light too, but I stop at them.

I always felt worse about stopping and accelerating again in a car because the cost didn't go to my muscles but to the environment in the form of lots of extra pollution.

I support self driving cars because I'm not arrogant enough a driver to believe I'm better than all the other idiots on the road and I see all the mistakes the other guy it making. Most people, when someone else cuts them off think that other guy is an idiot, when they cut someone off it is a minor mistake. In reality most people make about the same number of mistakes: every driver is a bad driver even when sober.

A self driving car is the only hope we have of making the situation better.

The reason why is because cops do not enforce those laws. How many times are people given tickets for not using blinkers in the USA vs Europe or Canada for example. It's a typical complaint for expats from those countries going to the USA.

Making the road safer for cyclists makes it safer for everyone. It's not that selfish.

I'm not sure that will actually calm traffic in the short run. Till every car is self driving I anticipate the mix to let to a lot more reckless overtaking abd weaving back and forth between slow autonomous vehicles.

My hope is that with autonomous vehicles having known reaction times we can actually get rid of fixed speed limits entirely.

Well, if you add enough slow autonomous vehicles to traffic, you won't have to worry about traffic obeying the speed limited because it will be much more slowly than that.

I'm really hoping for that.

I like that their delivery vehicle isn't any bigger than it needs to be.

An interesting idea out of this: if self-driving vehicles reduce the collision rate, road vehicles can (for some applications) get much lighter. Part of why cars are so big and heavy is safety (along with optionality for cargo and high speeds over mediocre roads). Shedding the steel safety cage could lead to improved fuel economy, which would result in a tremendous reduction in carbon output if scaled.

Excellent point! It's not just the reduction in space for the human passenger, but also a reduction in weight and resource usage from not needing (as much) safety features on the vehicle.

I'm guessing this would also make the cars safer to crash into, or be crashed into by, given that passenger cars would continue to be heavier and sturdier.

It's a chicken-and-egg problem. You'd want to be among the last people on the road to adopt a lighter car, because being in a lighter car while other people are still in heavy cars makes it much more likely for you to get killed in the event of a collision.

I wonder if it will take legislation to solve.

> You'd want to be among the last people on the road to adopt a lighter car, because being in a lighter car while other people are still in heavy cars makes it much more likely for you to get killed in the event of a collision.

There are many reasons to adopt a lighter car. Collision safety isn't one of them, but if that was your only concern you'd be insane to have a car at all.

The problem is that for many people (in the US), a car is simply a necessity. And car safety is an important part of that. A lighter car would fare poorly against an older, heavier car in the case of a collision. Therefore, the lighter car is less safe.

That's why it's a chicken and egg problem. No one wants to be in a less safe car when there are so many accidents (and deaths from them) every year.

People only want safe cars if they don't interfere with regular conveniences. I'm sure I've posted this on here before, but if people really wanted safety, we'd have roll cages, harnesses and helmets. But we want safety only when it's convenient.

I think people talk a lot about safety, but I don't think that is as primary of a concern as everybody says that it is for them. How many people actually go research and learn about all the cars they are shopping, vs people who just go to a dealership and listen to a salesman and buy what's in their price range?

Most people don't even read the damn owners manual, they don't even know what that big yellow light is (TPMS or CEL) that has been on for months. They won't buy winter tires in the winter, even though you can get a set for like $700. $700 is too much to spend to increase winter safety by a large margin on a $30,000+ car. Hell, people don't even use turn signals. The little stick that is less than 2 inches away from where your hand should be, and you don't even have to take your hands off the wheel to activate it. The problem is you'd have to decide which of your 16 cupholders to put the cup in, then reach way over and put it down before you could signal.

> But we want safety only when it's convenient.

Regardless, safety is still wanted. Just because people don't go over the top doesn't mean they don't want any at all. Having a lighter vs heavier frame would make a significant difference in a collision between regular-style cars and new, lighter ones.

The rest of your comment is attacking a strawman.

Its a non issue, I would buy and use a lighter car right now if it let say significantly cheaper.

A lot of the safety-related weight is non-negotiable, at least in the US. Every new car sold has a lot of airbags, and while they aren't mandatory, it's impossible to earn a good safety rating (from the NHTSA, a government agency) without them. So you could say the added "safety weight" (and drag, since the airbags make the car bigger on the outside to preserve cabin space) has been partially regulated into existence.

At some point in the future, I hope the NHTSA takes in "probability of crash" into their safety ratings for self-driving vehicles. Otherwise the safety bloat in modern cars will stick around. The equivalent of wearing a motorcycle full-face helmet while riding a bicycle.

Once safety improves enough, other considerations like fuel economy, available parking spaces, acceleration etc become more relevant and slowly drive weight down.

> You'd want to be among the last people on the road to adopt a lighter car

Make all the delivery vehicles that have no people in them really light first

> I wonder if it will take legislation to solve.

We can barely get common-sense prison-reform (or anti-surveillance) legislation on the US for fear of being labelled soft on criminals; imagine the terror of being the legislator who made cars less safe (without even winning favour from the auto manufacturers for doing so).

> I wonder if it will take legislation to solve.

Some people simply need large cars - how are you going to legislate that?

If I routinely carry 9 people, I need a large van. I don't have the option of a lighter car.

Kind of like how drones are smaller and cheaper than fighter bombers.

The vehicle also looks kind of top-heavy. If it is, that might not be such a problem because it can be programmed with limits to how much it can turn at a given speed (keeping the limit high enough that it could still swerve to avoid an accident).

Has no one here hit a dear before?

Or a moose. Every time one of these self-driving threads starts heading towards "and they can all communicate with one another and fly along the road at 200 MPH" — no. Braking distance is still going to matter when a deer bounds out of the woods.

My primary concern with autonomous vehicles is crime and abuse. People are going to vandalize, damage, destroy, and salvage these vehicles. There's already little being done about this in manned vehicles, so what's going to be done for unmanned vehicles?

USPS mail trucks sit on city streets unattended every day while the carrier walks house to house, leaving the truck chock full of people's paychecks, Amazon boxes, money orders, credit cards, and tons of valuables. And they don't get stolen or robbed often enough for USPS to do anything about it.

There are plenty of ways for criminals to commit super easy crimes without autonomous vehicles being in the mix, but somehow society continues to function.

Vandalizing mail service trucks carries quite a bigger punishment than Philadelphia citizens destroying an autonomous robot (which happened some time ago), in fairness. People know not to screw with the Postmaster General.

That's because the delivery person isn't far away so there's a decent risk of getting caught. If you know nobody is coming to save the robot for 20 minutes...

If they're deployed in areas with continuous LTE coverage, robust anti-tamper measures are entirely feasible. Cameras (which most autonomous vehicles have for perception) will witness and record the crime (and should be able to upload the footage in near-real-time).

Stealing an unmanned vehicle-- and getting away with it-- should be very difficult.

We've got one of these that roam around our corporate campus: https://www.knightscope.com/knightscope-k5/

Visitors like to walk up and take selfies with it, but no one messes with the thing because it's uploading 360 HD footage to some control center somewhere.

Some drunk guy tackled the one that roams the company that makes thems parking lot. He was promptly arrested based on all the excellent, well lit HD footage of his whole run up and contact with the robot.

Glitter bombs and fart spray.

> Stealing an unmanned vehicle-- and getting away with it-- should be very difficult.

As witnessed by the difficulty people have dealing with package thieves, "getting caught doing it" is unfortunately different from "not getting away with it", even when 'it' is illegal.

We've already got unmanned vehicles all over the place - in parking spots. Surely they'd be easier targets for vandalism etc. than a moving unmanned vehicle with a bunch of cameras recording?

There are 100 million non autonomous cars which do get vandalized, damaged, destroyed, and salvaged, but not at a rate which has obviated the concept.

Probably the same as in any dystopian book.

Sentry guns on top or inside.

Are regular ride-sharing cars vandalized and parts stolen off of them? I see lots of Reach One BMWs and Mercedes parked on the streets. Car To Gos, etc. Not sure this will really be any different.

Damaging automated infrastructure will reduce your social credit score. You won't be able to get into the automated convenience store. Won't be able to buy a plane ticket or a high-speed rail ticket. Won't be able to get a loan. Won't be able to get into a good school. Won't be able to get a vehicle license for a congested area.

All this is operational now in parts of China. They've finally figured out how to automate karma.

Looks freaky, which is part of the UPS truck model. No used market, no willingness by the owner to buy used parts.

They could still make it too easy to vandalize on a whim, but AFA theft, it is probably better than owning a Toyota.

> He said a vehicle at the demonstration didn’t drive because its battery had died.

I don't know what is worse: If that's true or if it's a lie.

Ive been following grocery delivery automation since the days of "Publix direct" back in the late 90's in Seattle. I believe it was microsoft that partnered with them to order groceries straight from a smart fridge, the MS Home had one of these. I was recently reading an article on Bill Smith, CEO of Shipt which was acquired by target for 550 Million a year back. What I could not for the life of me figure out, was how that company was profitable, they charge 99$ a year, for unlimited grocery delivery, how is that profitable? After looking at their glassdoor entry, I believe its due in part to labor rates being very low which wont be sustainable in the long term and I believe is why he sold to Target. Anyhow, this automated delivery by NURO could be something although a bit ahead of its time, as cities just arent ready for autonomous vehicles just yet. I dont see why the grocery stores havent caught onto this though, and adopted a take-out system only where orderes are placed online and robots put the orders together, for bulk pickup. The battle of grocery stores isnt the time to pick up the groceries, but the time it takes to shop for what you need, parking, checkout out.

I wonder why most autonomous vehicle projects are piloted in Arizona ? loose regulations ?

Having lived in Arizona, it is a perfect place to train autonomous vehicles. Most of the Phoenix region is a grid with wide roads. No inclement weather. Very few pedestrians. Clear road markings, isolated turn lanes. Basically no crazy scenarios like turning right on market street in SF. Given that autonomous cars are struggling in the most friendly environment tells me that they are decades away from ubiquitous rollout

All of those make it sound like a pretty terrible place to train autonomous cars, given how narrow the space of driving situations is. Cities like SF, where driving is kind of miserable, seem a lot more suited to training generalized models.

Though I do agree with you: I recently took a job with a self-driving car co, and am starting shortly (due to the technical challenges and great pay); but the assumed timeline most people are operating under seem pretty optimistic to me. That being said, I think much of the industry isn't expecting fully autonomous vehicles until at least the 2040s.

There's plenty of difficult places to train the robocars and I see them there often.

I rarely (if ever) see them on the westside where the roads are nice, logical grids but mostly Tempe, Scottsdale and Chandler where they aren't so nice and neat. Nothing like driving around SF admittedly.

I actually happened to spend the week of Thanksgiving in Scottsdale (incl a lot of time in old Town), and yea, it doesn't really compare. Weather, hills, pedestrian density and behavior, complexity of street signs and road rules, etc etc etc.

Didn't their governor sign an executive order a couple years ago to try to make the state the leader in self driving cars?

Edit: Yeah, here's[0] an article that mentions it. The governor came under fire for it when that uber car killed the pedestrian.

[0] https://www.bizjournals.com/phoenix/news/2018/03/20/ducey-up...

I'm curious to know this as well. I always assumed it was to do with comparatively sparse population, good visibility, lack of 'old' infrastructure, lack of weather incidents (rain/snow/etc).

Good weather

This and no hills... which means you have lots of options for routes, because everything's on a damn near perfect grid. Which also makes most of the situations the car encounters much simpler. No strange 5 way stops, no angled or offset intersections. Not even any curves to have to navigate. More opportunity to make 3 rights instead of a left, which these cars seem to be struggling at.

Since when does Arizona have no hills? Most states in the US are flatter than Arizona.

Phoenix has interesting geography. I struggled to describe it, so I checked Wikipedia, which has this succinct description:

"Other than the mountains in and around the city, the topography of Phoenix is generally flat, allowing the city's main streets to run on a precise grid with wide, open-spaced roadways."


Possibly, the flat image might come from Arizona being flat where a large number of the people live, like Phoenix and its outskirts.

Grandparent's "damn near perfect grid" comment also jives with the layout of much of Phoenix.

When the weather gets bad, it's just even hotter. No rain (some short summer monsoons albeit), no ice, snow, tornados, floods, earthquakes, fires, etc. All it ever does is get even hotter, which everyone is prepared for for the most part.

Don't forget the occasional haboob.


I doubt our technology is ready for that.

How could i forget the haboob! I was actually in Phoenix for this, it was wild. Like something straight out of the Mummy movies.

Unmanned, but "shadowed" by a car with humans. The autonomous vehicle industry excels in one thing, for sure: moving the goalposts.

A quick thought: when a customer enters the code and the doors open, can they somehow mess with other people's food? How do they prevent this?

Looking at the photos: the vehicle itself is tiny, as is the cargo space for groceries. There are two cargo areas with separate doors, so presumably each customer's delivery could have an isolated cargo space.

I could imagine this concept scaled up to van size or even truck size, integrated with something like a refrigerated amazon locker. Exciting stuff!

Who is Kroger and Nuro's insurance carrier on this if something bad happens?

What's the liability if their autonomous grocery kart runs over someone?

I have often thought:

Why isnt there an unmanned transport infrastructure which could be built between endpoints.

Take Hyperloop as a model-example. Instead of hyperlooping people about - why not only freight?

Imagine a mini-hyperloop enclosed tube about 10-feet in diameter, which has delivery sleds between hubs, which interconnect...

Then do a 20' diameter tube with sleds, and you have a trucking system which can be managed with BGP level routing experience...


Everything old is new again.

As to why it's not currently done, if somebody comes up with a significantly lower cost method of doing purely underground construction (cut and cover trenching), directional drilling or small tunnel boring that doesn't cost a huge amount of money per linear meter, they'll become a gazillionaire. Not just "loops" but all electric, gas, water, sewer and telecom utilities will clamor for the tech.

There's been a theory circulating around for a while that Musk's Boring Company is well-positioned to construct utility tunnels, and their seeming focus on transportation is essentially a loss-leader PR effort to drum up support from investors and the general public. Utility tunnels aren't exciting enough on their own, but there's a fair bit of money to be made in that industry.

Musk has been long mum on the subject but he recently said, at a National League of Cities event in Los Angeles on 2018-11-08 [1][2], that's he's open to the idea.

Of course, the big issue with tunneling, and construction projects in general, is often utility relocation, and not necessarily knowing where other utilities and surprises lie ahead of time.

[1] https://www.forbes.com/sites/alanohnsman/2018/11/09/its-not-... [2] https://arstechnica.com/cars/2018/11/elon-musk-on-double-dec...

The problem with those tubes is that when things get stuck underground it is very expensive to repair them.

Haha - Yeah - thats the basis.

Whats funny is when I was twelve years old - I designed a magnetic lev train built in a balsa-wood frame circle, with a cabin which was suspended in the tube-frame...

I was really pissed when Musk came out with hyperloop in that I was like "Any twelve year-old can design this!"


It was thought of, and built, in the past [1]. Most infrastructure-scale systems were abandoned for cost, maintenance, and operational reasons. And this was back when urban areas were more compact relative to their population, and not in an age like today when sprawling networks of government-subsidized, high-quality roadways reaching from an urban core far out into low-density areas are available for all to use for minimal cost.

[1] https://en.wikipedia.org/wiki/Pneumatic_tube

One of the most useful large systems of that type is the garbage pneumatic tube system on Roosevelt Island in New York. That was installed in the 1980s and has worked well. But big systems are rare. Disney World and the Grand Mosque area in Mecca have them, both being places with very large numbers of visitors.

>abandoned for cost, maintenance, and operational reasons

Because they didnt have the methods, systems, etc that we have avail today... We should look at this again.

I would pose, that the "random scooter rental" algo would do well within this space...

These are common, and called "Trains"

Most people don't realize that those massively long freight trains they see crossing the country often have one, or even zero, people onboard.

When I lived in Wisconsin, Wisconsin Central Railroad was lobbying the state to allow it to run trains with only an engineer. It pointed to the safety records of freight trains in New Zealand running with no crew at all as a safety precedent.

I think in the end, it was allowed on routes where the railroad paid to upgrade the grade crossings.

You just reinvented railways.

Another overhyped autonomous deployment. Just a gimmick like the "autonomous" Lyft ride I had in Vegas, where the car had two people instead of one in a regular run of the mill yellow cab

these overhyped autonomous deployments are how we slowly progress to accurately-hyped autonomous deployments.

A killer feature

Guidelines | FAQ | Support | API | Security | Lists | Bookmarklet | Legal | Apply to YC | Contact