Hacker News Reader

Comment by amarka | Hacker News Reader

I agree that there's a case for improved efficiency, but I'm not sure if the economies of developing all the tech from scratch adds up. One accident can endanger the entire company, a problem they're still reeling from with the bicyclist fatality. The car they were trying to automate has much better stopping (and overall safety) performance than any truck, yet they're still experiencing all the woes of developing new tech. Now imagine an automated truck accidentally punting a Grayhound buss full of band kids off the road.

View on HN

This venture kind of scared me to be honest. Having known a few truck drivers and having a good friend who was permanently disabled by a tractor trailer I know this industry needs to be made safer. I'm not convinced total automation is the solution as well but maybe somewhere in-between is good enough to improve things. Just doing this for pure profit is irresponsible. There is also the issue of if they are going to put any more trucks on the highways and interstates more roads need to be built first. The ratio of big rigs to commuter cars is presently out of control!

> Having known a few truck drivers and having a good friend who was permanently disabled by a tractor trailer I know this industry needs to be made safer.

Having known a few truck drivers myself—including my father—and having a very good friend who was permanently paralyzed by another friend driving a standard car, these things don’t influence each other. The trucking industry is far safer than your average driver[0].

> The ratio of big rigs to commuter cars is presently out of control!

The numbers disagree with you. That ratio is currently about 1:135[1].

[0]: http://www.iihs.org/iihs/topics/t/large-trucks/fatalityfacts...

[1]: a couple quick searches place big rigs at about 2M in 2017 compared to nearly 270M registered passenger vehicles in 2016.

Yes, but I never implied a year by year analysis either: https://www.forbes.com/sites/trucksdotcom/2016/05/23/a-glanc...

I never implied that these were hard facts they were my opinion. What is your opinion? We need more trucks on the road? To me 1:135 is out of control- that's my opinion.

You stated your opinions as if they were facts. You said you know the trucking industry needs to be made safer—that sure sounds like an implied statement of fact, not sharing of opinion. I shared a ton of data that disagrees with what you know—ahem, believe. That should encourage updating your opinions.

Given the facts available on safety records of licensed, commercial truck drivers vs normal drivers—the latter of whom outnumber safer drivers 135:1—it seems sensible to me to form an opinion that the 135 less-safe drivers need to be dealt with before we get too worried about the 1 safer driver. The overwhelming majority of multi-vehicle accidents with big rigs find the passenger car driver to be at fault—we’re talking from 70-90%, based on types of crashes. Non-truck crashes outnumber truck crashes by roughly 3:1 per 100M vehicle miles traveled.[0] This seems to indicate non-truck drivers pose the greatest threat on the roads to public safety.

So sure, form any opinion you like. But maybe be more careful to share them as obvious opinions that aren’t implying they are actually fact-based—by stating you know something is true or declaring the ratio of something is out of control—or someone is likely going to call out those statements as being questionable when compared to the facts of reality. There’s no clear evidentiary basis for arriving at an opinion on a correct ratio of trucks:non-trucks, other than the data we have seems to indicate that fewer passenger cars on the roads is the surest way to increase public safety.

[0]: For a new link summarizing various studies—http://www.trucking.org/ATA%20Docs/News%20and%20Information/...

There are already rules in place for human truck drivers that push the limits of what a human should be called upon to do for many hours at a time. Human truck drivers already have regular accidents due to fatigue. Replacing or augmenting long haul trucking can only be a positive.

What rules are you referring to? Current regulations prohibit drivers from having > 14 hours on the clock, iirc. Ignoring any drivers—or the companies who employ them—who are overtly ignoring these rules and faking their log books to be active longer than 14 hours, there are mandatory 10-hour windows drivers must not be working. That 14 hours encompasses all activity—loading, unloading, weighing, weight redistribution, driving, etc. Driving time itself is capped at 11 hours of an allowed 14-hour workday. Newer trucks even have cameras in them to monitor drivers, as well as other systems that report violations and actively prevent the truck from being used in a way that violates regulations.

> Human truck drivers already have regular accidents due to fatigue.

Fatigue accounts for 13% of truck driver-caused accidents according to DOT[0]. Fatigue is coded twice as often for passenger vehicles as it is for commercial truck drivers.

Moreover, the rate of commercial rigs involved in accidents with passenger vehicles is quite low. The rate of single-vehicle accidents is also lower among commercial trucks. Commercial trucking continues to grow increasingly safer every year since we’ve been keeping track in the 70s[1].

[0]: https://www.fmcsa.dot.gov/safety/research-and-analysis/large...

[1]: http://www.iihs.org/iihs/topics/t/large-trucks/fatalityfacts...

Maybe relevant disclaimer: my father is a truck driver and we regularly talk about this stuff. His experiences have led me to do a bit of research and study on the matter. I don’t work for or on anything trucking-related.

My father is a retired truck driver, it'd be great to curse at each other sometime. :)

My understanding is that the industry (and maybe this has changed or was not good anecdata to begin with) is rife with gaming of the regulations, which in my opinion are already grueling. A human being, no matter how accustomed they are to driving, should not be asked to sit and drive down long mundane stretches of road at a high degree of alertness for 11 hours per day, multiple days per week. I understand that the new time tracking systems will reduce the ability to game the system, but I feel the fact that regulators are calling for these devices and driver awareness monitoring devices should be an indication that maybe we can find a solution that doesn't involve a human.

Your evidence seems to contradict what you are saying. From your second link:

"A total of 3,986 people died in large truck crashes in 2016. Seventeen percent of these deaths were truck occupants, 66 percent were occupants of cars and other passenger vehicles, and 16 percent were pedestrians, bicyclists or motorcyclists. The number of people who died in large truck crashes was 27 percent higher in 2016 than in 2009, when it was the lowest it has been since the collection of fatal crash data began in 1975. The number of truck occupants who died was 47 percent higher than in 2009."

You’re misreading the data presented. Yes, 2016 was worse than 2009, but it was far safer than ‘75 and the decades that followed. That other particular stat on fatalities is only looking at the fatality rates of truck drivers vs passenger car drivers in accidents that are between a passenger vehicle and a commercial rig—and is not a surprising rate considering one ought to expect a truck driver to have a higher likelihood of surviving such a crash compared to a passenger car occupant. When you look at the comparative rates among non-truck accidents and fatalities, truck drivers are far safer. If I correctly recall the data, the rate at which truck drivers are at fault for accidents with passenger cars is also lower than the reverse. When you look at the comparative rates of truck accidents vs passenger car accidents, non-commercial drivers continue to be the most dangerous drivers on the road, and there are vastly more of them putting others at risk.

> The number of truck occupants who died was 47 percent higher (in 2016) than in 2009.

Do you have numbers for dead truck occupants per mile driven? It could be that this is due to truck traffic being lower overall during the 2008/2009 crisis.

That data is in the link I shared. 2009 had more miles driven than 2016.

Can the robot perform a brake check? Can the robot ensure a load is properly secured? The act of driving the truck is only one of the many jobs a trucker does.

They're addressed by established process and routine - 2 things machines do better than us. Can a machine monitor braking or perform the current brake check - absolutely. Could we shift the responsibility of check the load tie-down to the freight facility and monitor ongoing status with sensors. I'd hope so.

You should be asking "can a robot drive for 12+ hrs without fatigue?" or "can we eliminate the restrictive, expensive and often gamed system of keeping drivers within their hours?" This is were long-haul transport could be "disrupted"

>> Can a machine monitor braking or perform the current brake check - absolutely.

I'm not sure you understand what a "brake check" means. It isn't checking the brakes for current functionality. It means a visual inspection of all the brake parts to ensure they aren't going to stop working somewhere literally down the road. It is checking for pins, debris, excessive or unusual wear, or leaks. It would require 3d vision backed up by some serious AI to understand what is going on. And you would probably need some sort of robotic actuator to remove any debris blocking inspection areas.

When you see a truck stopped by the road with the driver walking around the trailer, he is probably doing a legally mandated "brake check". It isn't just pumping the brakes to see they are still there.

It is also checking brakes for functionality before you set out. There is an in-cab procedure too, not just a visual inspection. But I believe that both can be automated. The in-cab brake check is testing for certain pressures in the system under certain circumstances, and could be automated via software-hardware combo. The visual inspection could be done via a series of cameras under a bay the truck rolls over, with software to detect issues with the braking system components visually. Or either of these functions could be performed manually by a human at a waystation before the truck heads out on the (next leg of the) trip.

I don't think people are claiming computers can load or unload a truck.

Computers can check if a brake is working or if the cargo is well balanced better than a human. One could do those today with cheap (on the 100's of dollars) electronics and few lines of code. Nobody does this because it's 100's of dollars more than letting the driver do the same.

I'm saying if you limit the scope to tractor trailers hauling only 20/40 ft ISO shipping containers, then "loading/unloading" can also be done in an automated fashion. I'm saying that if you limit the problem space, and try to approach even just a single use case such as this on a major artery like I-40, I don't see any way you end up in a worse spot than human drivers.

> I don't think people are claiming computers can load or unload a truck.

I think truckers would absolutely love it if this was the case. It would save them so much time waiting on loading/unloading, which decreases their pay.

I don't think they would be the ones to profit from this invention though.

I understand that the job encompasses other duties, my father is a retired truck driver. That said, my answer is "yes, I think they could, just as well, or better than any human".

You could have multiple cameras on the load using computer vision to detect movement. You could have weight sensors sending feedback to the system. I'm sure there are many ways to do this, and any of them would beat a human behind the wheel.

Then why not deploy such systems now? Prove they can handle all the odd jobs better than the human.

Huh? I'm not saying they exist, just that it's not impossible for me to imagine these problems being solved or being minimized to a level that is no worse than current human driver would perform.

dsnuh - I haven't seen anyone argue against "[software+hardware] would beat a human behind the wheel", its just that developing the tech to achieve this feat with the same level of accuracy as the top 10-20% of current CDL drives is extremely expensive and caries a lot of risk. Nobody seems to of cracked the nut on it.

You're right in that there are many ways to do this, but none have come even close to beating a decent human behind the wheel.

I don't see any indication in the article that they are stopping due to technical challenges. They already demonstrated it on the road, so it seems they were well on their way. It appears this project is a victim of politics and legal action.

Why do you think we need to get to the level of top 10%-20% of commercial drivers? How do you come to that cutoff point? If we had automated trucks that could move freight 24/7 with even 5% better than average accident rates (for example) would be a huge win.

I agree for the most part, but it's not quite as simple as just having better accident rates. That is one thing, but there are other considerations as well. When there are accidents, are they predictable ones, or is it completely random? Did it make a decision that we cannot explain which led to the accident? People aren't going to like completely random accidents, even if there are slightly fewer of them. Can we assign fault in the case of an accident, sometimes, always?

I think if you have a predictable accident, that's called negligence, and we have courts to deal with that.

Reply on news.ycombinator.com