Opinions On Self-driving Trucks Please!

Im terrified of self driving trucks. i wont have a job, will starve, and will have to rehome my cat.

what oh what will.i do????? we are all gonna DIE!!!!

sorry. the sky is falling the sky is falling chants of these vehicles make me laugh

Im terrified of self driving trucks. i wont have a job, will starve, and will have to rehome my cat.

what oh what will.i do????? we are all gonna DIE!!!!

sorry. the sky is falling the sky is falling chants of these vehicles make me laugh

The only thing that worries me is they will convince everyone they have perfected the software, though we are all likely to be retired before then. There is no such thing as bug free or unhackable software.

Not in my lifetime!

G-Town, "Don’t focus on the symptom, focus on the root problems. " O.K., the fatality rate is a "symptom". The weight of evidence says that the primary cause of the fatality rate is human error - in other words, the "root problem" is human error. Sometimes the truck driver, sometimes some other driver, maybe once in a blue moon someone else made an error that led to a fatality. Do you agree?

Roy, you said:

A major consideration in the development of "driverless" vehicles (trucks, cars, trains, aircraft) is not a technical issue, it is a cultural issue. As a society, we are very slow to accept change of any kind. Back when automatic transmissions were first becoming common place on private automobiles, my grandmother refused to buy a car with an automatic transmission, because "I don't trust it".

We wouldn't have a cultural issue if we knew we could trust technology. Your Grandma was right - automatic transmissions were unreliable. In fact, automatic transmissions have been around for big rigs for decades, but only today are major companies making a major push toward them because they're finally reliable enough that they'll save the company money in the long run. However, they're still more costly and less reliable than a standard. The money they save in fuel and training costs makes up for it though, finally.

So I'm glad you brought up this example. Creating an automatic transmission is about one millionth the complexity of creating software that can safely and reliably navigate modern roadways, and that's not an exaggeration. Now consider this:

The first automatic transmission using hydraulic fluid may have been developed in 1932 by two Brazilian engineers, José Braz Araripe and Fernando Lehly Lemos; subsequently the prototype and plans were sold to General Motors who introduced it in the 1940 Oldsmobile as the "Hydra-Matic" transmission.

So we've had automatic transmissions in production for almost 80 years and they're still more expensive and less reliable than standards, especially in a big rig. 80 years you've had countless engineers and scientists working on these things, and they're still improving them.

Considering how much more infinitely complex the undertaking is, how in the world are they ever going to write software than can navigate today's roadways? It's a pipe dream. In reality they'd only have to make them a little more reliable than humans to be worth doing. But that's saying a lot. The most powerful supercomputer in the world doesn't have 1% of the processing power of the human brain and nervous system. Not only that, but the sensors you would need in order for a computer to take into account all of the things a human can take into account is mind boggling.

Also, consider this.......if a transmission fails you pull off the side of the road and call a tow truck. If an autonomous car fails you probably injure or kill someone. How many errors can you accept in this software? And what if there's a major systematic failure?

But here's the most important question for you. Have you written any software yourself in an attempt to do anything complex? We can't even reliable predict the weather 12 hours out. We still haven't figured out human nutrition. You could put three apple pie recipes in front of Google and Google couldn't tell you which one would taste the best.

Computing power and software isn't within 100 years of being able to drive vehicles on today's roadways reliably. Heck, it still isn't even powering trains and airplanes reliably, and those are 1,000 times less complex than Downtown Chicago traffic at rush hour.

Roy has struck a nerve...

G-Town, "Don’t focus on the symptom, focus on the root problems. " O.K., the fatality rate is a "symptom". The weight of evidence says that the primary cause of the fatality rate is human error - in other words, the "root problem" is human error. Sometimes the truck driver, sometimes some other driver, maybe once in a blue moon someone else made an error that led to a fatality. Do you agree?

Like I said...wait until you are out here.

Without any exception, every boneheaded, moronic move that has required quick thinking and evasive action on my part to avoid an accident was caused by a four wheeler doing something very stupid. No trucker has ever cut me off, blown a stop sign, brake-checked me, tried to drive under my trailer, drove behind me when backing, or fired the one-finger salute at me. Only four-wheelers have done those things.

Truck drivers make mistakes...fact. However vigilant space management, patience, proper pre-trip, and driving with prudence are implied in being a true professional driver and will keep a driver safe 99.99% of the time. All of the current safety technology will not prevent “stupid” or lack of skill and common sense.

Driver training for most non-professional drivers stops at age 16-17. And that very basic training includes nothing about how to drive around big trucks. Nothing! Tell me something Roy; what happens when you tailgate my trailer on the Interstate and I blow one of the outside rear tires? I get tailgated many times every day by 4-wheelers and light trucks...just like NASCAR.

There are many causes of accidents, but very bad and impatient behavior by 4-wheelers is a huge, huge problem and one totally overlooked by the media, politicians and the uninformed public.

And technology, the hand-held type has exacerbated the lack of focus and inattentive driving. It’s chronic.

Not sure where you live Roy, but I’d be happy to host you on my truck for a day-run into North Jersey just so you can see first hand what we actually do out here.

Not sure where you live Roy, but I’d be happy to host you on my truck for a day-run into North Jersey just so you can see first hand what we actually do out here.

You are definitely a better man than I am...

G-Town, I am quite aware of the bad behavior of 4-wheelers, as I have about 3000 hours driving a 26,000 GVWR Box truck, and another roughly 3000 hours pulling loaded goose-neck cattle trailers and flatbed trailers (10,000 to 15,000 lbs) behind woefully underpowered light trucks. No, not as big as what you drive, but slow enough to invoke the events you call bad behavior and I call driver error. But what do you propose to do, short term or long term, to reduce this frequency of collision-inducing behavior? I don't know about Jersey (other than driving on I-95), but around Dallas and Houston it is pretty much a lawless free-for-all, 24/7, and no amount of enforcement or new law seems to change anything. My own view is that the prototype driverless car projects will, over time, yield much technology that can be integrated into cars to prevent certain types of behavior - like running stop signs and red lights. Will the technology be "perfect"? Maybe, maybe not - but right now I'd bet that the technology will be way safer than the teenager playing with their cell phone or the average drunk driver. No, not this year, but the track record of the existing driverless car projects is very encouraging. Over the course of a few decades we have gone from a robot having limited ability to "walk" down an office hallway to cars that can successfully operate autonomously in typical suburban traffic. Now, in cattle country, we see quite a few roll-over accidents that involve big rigs hauling livestock; invariably, the truck going around a curve, and then laying over, without any other vehicles being involved. The other thing that seems to happen all too often are instances of truck going across the median (or off the other way) after blowing a steer tire - occasionally with bad outcomes. What do you propose to do to reduce the frequency of these events?

Roy, you're talking about nothing but the common mistakes or bad decisions people make. What you're not considering is the thousands of perceptions humans make correctly on a continuous basis that would be nearly impossible to program into a computer or detect through a gigantic array of sensors. There are a ton of circumstances that happen all the time which are easy for people to figure out, but would be incredibly difficult for a computer.

For instance, you're driving along in heavy traffic on slick roads and the wind blows a large, loose garbage bag in front of the vehicle. Do you slam on the brakes, potentially causing a huge wreck? Of course not. It's just a bag blowing in the wind. But what are the chances that a computer is going to be able to accurately detect what the heck that is?

Now what if it was Halloween? Is that an empty bag, or a kid in a costume?

Also, think about this. You're driving along on slick roads and a deer walks in front of you. Do you slam on the brakes, or run it over? Chances are the safest thing to do is just run it over. Yes, it's a horrifying decision to make, but if you slam on the brakes you could lose control and take out all kinds of vehicles, or go off the road and roll over and kill yourself. So the safest thing to do is run it over.

Now for a computer to make that same decision, someone is going to have to write a "death algorithm" into the software that knowingly runs down and kills a living creature because it was the safest thing to do in those circumstances. But what if the computer thinks it detects a deer but it turns out to be a kid walking his bicycle across the street and the car purposely runs the kid over and kills him?

There are 10,000 crazy scenarios that happen every day out there that a human can pretty easily interpret but a computer would find nearly impossible to accurately assess. It's just not going to know exactly what it's seeing or what it should do in those circumstances.

Like I said, show me a computer that can effortlessly mow my lawn or pick up dog poop first, or a robot that can make me scrambled eggs. Then we'll talk about the possibility of a computer someday driving a big rig through Downtown Chicago traffic.

Brett - if you have not done so, have a look at: https://en.wikipedia.org/wiki/History_of_self-driving_cars - No, they are not ready for "prime time", but they are getting closer, and various car models can be purchased today that have various components integral to self-driving cars. The technology needed for reliable self driving cars has exactly zero to do with automatic transmission technology. Claiming that "We still haven't figured out human nutrition. You could put three apple pie recipes in front of Google and Google couldn't tell you which one would taste the best." are evidence of deficient software development is way wide of the mark. Human nutrition is largely unknown because of the complete impossibility of ever being able to perform controlled experiments on humans - and has nothing to do with computation. Taste is a question of neurology and biochemistry, not something that can be computed. While weather forecasts are perhaps not as locally precise as you might want, forecasts are much higher resolution in both time and position than they were even 10 years ago; today, the primary limiting factor to more precise weather forecasting is the lack of high spatial resolution of current conditions, rather than compute power as such. (Yes, I have some specific knowledge about weather forecasting, going back to when I worked on the hardware interface control software for the very first Doppler Weather Radar system ("Nexrad"), back in the 1980's). While I appreciate your concern about the vast array of scenarios that must be properly detected by autonomous vehicle control systems, and suitable responses created, the undeniable fact is that there are several prototype vehicles on the road today that perform quite well; see https://en.wikipedia.org/wiki/Waymo . While the reliability of these prototypes is not 100%, the reliability of these systems is rapidly improving as more and more players enter the market, and set up large-scale tests in various locations. Since most of my career has focused on computer simulation, I find it particularly encouraging that (per the referenced essay) Waymo autonomous cars have logged some 5 billion miles in simulated driving. While I do not know the very specific details of this work, I know pretty much at a technical level exactly how such simulation is done, and the capabilities of such simulation to replicate, with many thousands of variations, the type of doomsday scenarios that you outlined. The ability to differentiate between a child on a bicycle, an inflatable likeness of a child, and a deer (say) can be, and no doubt has been, tested hundreds or thousands of times, just as aircraft simulators can provide a pilot with the opportunity to practice, over and over, scenarios that one hopes will never be encountered in actual flight. Interfacing the computer control system of an autonomous vehicle to a “real world” simulator is very mature technology that evolved with the development of computerized avionics in aircraft. On the Space Shuttle project 40 years ago we were able to simulate all of the electronic systems of the vehicle with sufficiently high fidelity that the on-board Flight Control Computers could not distinguish between simulated and actual flight environments; before the first Shuttle Mission those Flight Control Computers had “flown” thousands of missions. (I am eager to work with the truck simulators that I understand are a part of many driver training programs; it will be interesting to me to compare the fidelity of the truck simulators to commercial and military flight trainers.)