Viich

Cameras are extremely reliable - the images are just really hard to process (hence the shortcuts the human brain takes) The problem with just about anything else is that you can't get the range needed easily, along with the fact that using active sensors for a system like this will work for a demo, but the interference created from every car having an active sensor would end up with no useful data for the 'driver'.

If it used the same rear-end avoidance that other cars have, it's pointed down to avoid this problem - this truck/trailer didn't have skirts from what I read, so nothing to give a return, and it doesn't point up to look for a truck like this.

Passive sensors are the only way to go if the technology is going to work mass-market - I can think of a couple schemes to do some separation for radars, etc, but the frequency licensing cost would kill the price point. IR cameras are probably what's needed - go far enough beyond visible light and the computer should have been able to easily pick the truck out from the sky.

OldTryGuy

Seems to me those 18 wheel drivers did quite well driving 80,000lb vehicles, double clutching and using a CB to keep in touch with other truckers good-buddy. One thing for sure, pretty hard for a computer to drive drunk.

Shimagnolo

Those roadways will be the limited access, multi-lane highways, which are already restricted to motor traffic.

ItsJustMe

How many drivers were killed by human error between the time that Tesla rolled out that feature and this event? Even corrected for number of cars on the road, humans are still more dangerous.

Besides, it sounds like a pretty unusual and difficult situation.

Shimagnolo

Read this, and you'll never fly on certain Asian carriers again: Doug Ross @ Journal: The Ominous Facts Regarding Korean Airline Pilots by a Former Instructor

I've read other similar stories from US flight instructors, about the cultural issues in some foreign countries which value socio-economic status over actual skill in choosing airline pilots.

gecho

Implying those are the only two options available. Driving defensively can usually eliminate those kinds of last second "who do I kill" decisions. Yet the media always like to frame it that way.

gecho

That particular test was run by a dealership using a vehicle that did NOT have an auto brake feature. OOPS, more of that human error stuff.

Chris0516

Just goes to prove. Why I don't trust them.

Shimagnolo

Old news from a year ago: History, Travel, Arts, Science, People, Places | Smithsonian

jefnvk

Care to provide an example? Outside of AF296 in 1988, I am not aware of any fatal crashes whose fault have been placed solely on automation, most still boil down to pilot error.

As to the "are they safer?", I think the fact that over the years, there has been one (possibly now two, depending on the results of the Tesla investigation) proven to be at fault in a wreck, out of dozens of wrecks they have been involving humans at fault, should tell us which is safer. Not to mention, that human driver still had full control of the vehicle, and he didn't hit the brakes either.

I love being in control of my vehicle, and actually pay attention and drive when I am behind the wheel, but if it means I don't have to deal with the people who aren't so attentive on the roads, I'll gladly choose us all having autonomous vehicles.

ItsJustMe

"At fault" is an interesting choice here since Tesla has steadfastly insisted that what their cars have is driving assistance and accident avoidance. It's not intended to replace driver vigilance. As things stand right now, the driver is ALWAYS at fault in an accident, because the technology is not meant to be in charge, it's meant to be a last ditch effort if the driver fails in his job. For the Tesla to be at fault, first the driver had to be at fault.

Shimagnolo

I read one article in which the truck driver stated the Tesla changed lanes just before the crash.
If true, since the Tesla does not initiate lane changes, that would mean either:
a) The auto-steering was on, and the driver initiated via the turn signal (as shown in one of his videos).
b) The auto-steering was off, and the driver was doing the steering.

Either way, that would imply the driver was aware, assuming we can believe the truck driver.

KD5NRH

This is also an important point, considering that the description sounds like the trucker left-crossed the Tesla, so it's entirely possible he's looking for any way he can find to spread the blame.

CB HI

Air France Flight 447 Airbus A330 Without the automation, the plane would not have crashed. It is always easy to blame the pilots, avoiding grounding planes.

CB HI

No death in that test.

jefnvk

AF447 was solely due to pilot error after the mechanical issues that started it. Iced pitot tubes gave erroneous airspeed readings, after which the autopilot disengaged and handed control back to the pilots, which then totally botched the job, put the plane in a stall, and crashed it. It was the pilot's incorrect reaction and input that crashed the plane, not automation.

You can maybe make the claim that the information was not presented to the pilots in a clear enough manner, or that they were overwhelmed by the about of alarms going off after incorrect input, but that is not an automation issue.

CB HI

It was the added automation that lead the pilots to put the plan into the stall. Please note that american pilots have refused the same automation, due to the danger.

ThermionicScott 

+1. Using 2015 statistics, there were over 100 traffic fatalities caused by humans that day. Where's the outrage? Where's the "gosh, I don't think I can trust them"?

scott967

You're aware that Hawaiian Air flies the A330 I hope.

I think the general concern is that pilots have come to rely on automation, to the detriment of basic aircraft handling skills. I can see much the same thing happening to ground vehicle drivers.

scott s.
.

CB HI

Yes I am aware. I even fly on Hawaiian Air A330s. Seems you are not aware that the software is different from EU carrier A330s and US carrier A330s. For US, when the pilot releases the stick, the control surface returns to neutral. For EU, the stick moves the control surface and when the stick is released, the control surface remains in its last position until the stick is moved again, to either move the control surface further or the stick is moved the opposite position to move the control surface back towards neutral.

Think of a car that if you turn the steering wheel right for 2 seconds the wheels turn right and when you move the steering wheel to straight up, the wheels stay turned right until you move the steering wheel left to move the wheels left.

scott967

AFAIK, all Airbus sidesticks work what you consider the EU way when in normal law. But this is too far OT so I don't want to get further into it.

scott s.
.

CB HI

Current and prior US Military pilots strongly opposed that stick control system. US carriers do not use it and it was not planned for the proposed KC-45.

northernlights

The Tesla would hit the cyclist because it doesn't recognize them.

The Tesla autopilot feature is designed for freeway use only. It doesn't function in urban driving environments as it cannot see pedestrians and cyclists, can't read street signs, etc. Like cruise control you can only use it on the freeway.

Viich

I'm pretty sure he's talking about the in-development fully driverless cars.

northernlights

That claim might be somewhat misleading. How many times was a collision avoided on autopilot because the human driver was able to retake control of the vehicle before it crashed into something? We know it does happen which is why Tesla requires hands on the wheels at all times when the autopilot is engaged.

But close calls probably aren't recorded so we may never know how often they occur.