Tesla to stop killing drivers: Software update beamed to leccy cars Tesla is changing how its "Autopilot" super-cruise-control works in response to the death of one of its customers. The over-the-air software update will be automatically applied to the electric cars this month and will expand the use of radar sensors to decide whether a crash is likely to occur. In a blog post on the Tesla …
My Audi is now more than five years old, and it has a driving assistance system which the manufacturer does not market as an "Autopilot", rather they call it Adaptive Cruise Control. It relies mainly on two Radar antennas mounted at the front of the car beneath the headlights, in addition it leverages a front-looking camera which keeps track of the road markings - the radar adapts to that camera view and looks along the current lane which prevents unnecessary braking when there is a road bend and another car on a neighboring lane.
The system is very reliable and really convenient when being stuck in a stop-and-go traffic jam, also helps greatly during normal traffic situations, for instance when another driver unexpectedly pulls over into my lane. The system reacts properly also to smaller vehicles on the road (eg. bicycles) and no doubt would detect a big truck crossing the lane ahead. However, I still keep watching the street rather than watching Harry Potter videos while driving.
I simply can't understand why Tesla did not do proper market research and did not find out how experienced car makers are helping their customers with advanced car assistance systems. Again, my car is more than five years old. How come that Tesla started to build radar antennas into their cars only as late as 2014, and only now starts to take radar somewhat more serious ?
Can we please refrain from untenable statements ? That particular adaptive cruise control has actually three settings for different driving styles. Even the most "dynamic" one has much more safety margin than those idiot drivers who actually drive up to just a few metres, even at more than 200 km/h - without using any assistant systems, and their biochemical brain either ...
"Audi....1 metre at 80MPH?"
Clarkson coined the term 'The Full Audi' for this.
Other makers need to include a reverse firing radar to detect tailgaters. Then automatically vary their own cruise control speed up and down by several kmh, timed to be out-of-phase with the tailgating Audi. Basic Control Theory concludes that applying such out-of-phase feedback into a high gain system creates an oscillator.
The Audi driver will have to turn it off and go away.
Maybe we should keep technology and the personal behaviour of drivers and their preference for diffferent car brands in different countries a bit separate ?
Seems that in the UK there might be a certain preference by aggressive drivers for car brand A, whereas in Germany those type of drivers were typically more inclined towards car brand B. Gradually the situation improves, you'll now find less aggressive driving and also less drunk driving than in earlier years.
Driving assistant systems seem to help here. The German government is now also mulling over alcohol sensing devices to prevent drunk drivers from starting their car.
We won't mention why Audi moved to twin transceivers plus camera - OK we will, the earlier version with a central transceiver was prone to getting confused by oncoming traffic, corners and other things, plus the vagaries of the interesting range of radar profiles of other traffic such as the stealthy rear end of a mk2 Mondeo saloon which could appear on occasion to have accidentally matched what needed millions for the F117 to achieve.
It also isn't widely mentioned how the camera positioning has evolved over a short period as real life proved certain engineering assumptions wrong.
I have to say that the Tesla statement suggests they don't really properly understand how the sensors behave, or the best way to use them, or what is reliable and even things like their iteration rates seem a bit optimistic when you consider what can happen in 100ms at normal traffic speeds. And this assumes they have used a proper certifiable scheduler to drive it.
You can throw software updates at the thing but their hubristic approach smacks of an Internet startup way of thinking that doesn't match with the sound engineering approach something like this demands.
"My Audi is now more than five years old, and it has a driving assistance system which the manufacturer does not market as an "Autopilot", rather they call it Adaptive Cruise Control."
The VW/Audi/etc. system doesn't do autosteer AFAICS.
Since you are still steering the car you are less likely to be disengaged from the road conditions in which you are driving. (Though I find myself entering a "fugue state" on older Motorways if I am not careful.)
If I recall correctly Mercedes and BMW have "lane keeping" assistance features, so they will have some form of autosteer, does anyone know what the accident statistics for these cars are like?
Possibly the main difference between all the cars with this kind of self-drive ability is the fact that the Musk is so high profile and keeps talking about it?
Tesla's Autopilot team has a Confidence-to-Competence Ratio that exceeds unity.
Volvo has been similar, with their two famous hubris-laced Self-Braking failures. Two videos available on-line. Hubris is standing in front of a prototype. Poster children of attitude based failure. Have they learned? Issued a memo to not stand in front?
Google seems to be more cautious. More sensors, slower speeds, acknowledged as not yet ready for prime time. I can't see much to complain about their Ratio.
That Confidence-to-Competence Ratio is what makes some of these folks dangerous.
An attitude adjustment is necessary.
They've forgotten that "A.I. is hard."
Which is really stupid.
Driving assistance systems like adaptive cruise control are certainly ranked below A.I., actually that's a special case of process control technology - just like the autopilot and anti-collision warning systems found in airliners since decades.
Agreed, blasting such systems into the market without very intensive testing and QA is really stupid. An attitude adjustment is necessary for those companies that don't do it yet, and also for consumers who do believe marketing BS too easily.
Latest Tesla death (wrecked exotics dot com): Tesla driver going 96 mph on a 50 mph road, explodes into smithereens, burns down to a crisp.
Tesla company confirms the most crucially important fact: the autopilot system had been disengaged. No fault to the company.
Cheers!
Do we know how the accident unfolded? How long did it take for the car to drive into the lorry? If it went relatively fast (like one second), the driver might not even have time to react, much less react properly and for the car to actually relinquish control (it doesn't see anything wrong, so there's no emergency to react to at that exact point in time).
If it took several seconds, then the lorry driver should have noticed something odd and blow the horn (unless he was on assisted cruise control, too).
I'm unconvinced that Tesla is at absolutely no fault here.
> Do we know how the accident unfolded?
I read a report of the investigation; the answer is yes, we do, to a certain extent. The long trailer truck was turning across the carriageway in front of the Tesla, so it was a side-on impact, at speed. The truck driver probably mis-estimated the speed of oncoming traffic, if he saw it at all, and the Tesla didn't slow, slicing its roof off on the underside of the truck and continuing on down the road on the other side.
See e.g. reports at Electrek.
There are also allegations that the Tesla driver was watching a movie while the car's 'Autopilot' was engaged.
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2024
