BMW uses autonomous cars for boring, repetitive tests

DistinctivelyCanuck

Ars Tribunus Militum
2,698
Subscriptor
I assume we can all guess what driving pattern Clarkson, Hamster, and Captain Slow would have had the cars drive instead of a figure 8 ?

Joking aside, this is a lovely set of test instrumentation. As a test engineer, I'd love to see more details how they did it, and I also know that for competitive reasons, I won't :( (I don't work in automotive.. but I know the drill... )
 
Upvote
71 (71 / 0)
Post content hidden for low score. Show…

AdrianS

Ars Praefectus
3,793
Subscriptor
how were they doing these before autonomous cars ?
RTFA!
They used to fit additional hardware to turn the steering wheel and push the pedals.

It's mentioned in the article:

"Before the current drive-by-wire system, BMW engineers would have to install mechanical robots in the vehicles that steered, shifted gears, and braked. It was a tedious setup that took days."
 
Upvote
67 (67 / 0)
This is a beautiful use-case for our actually current level of self-driving technology. Not even cutting edge stuff here, solid tech that was worked out almost a decade ago. I'm glad to see it's being used this way.

how were they doing these before autonomous cars ?
With human drivers doing the best they could and dealing with the fact that humans can't get precisely the same performance in each test cycle, so the data was inevitably going to be noisy. (The data here will still be noisy, because data always is, but eliminating driving fatigue from the mix should reduce the noise substantially and help quite a bit).
 
Upvote
43 (45 / -2)

dj__jg

Ars Tribunus Militum
2,685
Upvote
39 (39 / 0)

Arstotzka

Ars Scholae Palatinae
1,233
Subscriptor++
That's a surprisingly beefy braking failsafe; if the brake-by-wire fails, is there a backup physical linkage the brake pedal actuates that lacks boost or something?
I don't believe brake-by-wire is legal. Drive-by-wire is a bit of a misnomer; it refers just to the throttle. With electric power steering, computers can control the front wheels. Brakes are mechanical, where pressing on the pedal moves a piston to "compress" the brake fluid. (Brake system pressure can be in excess of 2000psi!) This squeezes the brake pads against the rotor, slowing the car down.

The failure they are concerned about, and why you need a beefy robot, is if there is an issue with power-assisted brakes. Every modern car uses engine power to assist the human's braking capability. If you lose the brake booster, it takes a tremendous amount of effort to get even a modest braking result. The robot has to be able to stop the car in the rare, but potential, case of power-assisted brake failure. And that's why it is so beefy.

EDIT: to clarify, full brake-by-wire is not legal. There is always a mechanical backup for steering and brakes, unlike the throttle. Re-reading this after a few comments I realize I wasn't being clear here.
 
Last edited:
Upvote
18 (23 / -5)

azazel1024

Ars Legatus Legionis
15,085
Subscriptor
I assume we can all guess what driving pattern Clarkson, Hamster, and Captain Slow would have had the cars drive instead of a figure 8 ?

Joking aside, this is a lovely set of test instrumentation. As a test engineer, I'd love to see more details how they did it, and I also know that for competitive reasons, I won't :( (I don't work in automotive.. but I know the drill... )
Ditto. Hadn't occurred to me, but this should be great for other testing. QA/reliability testing. Also things like EPA testing. Of course, that could also lead to EPA gaming...

Want 120k miles of wear study? You could get that in 3 months of around the clock testing with stops just for fuel and maintenance. Throw 10 vehicles out there to get a little bit better "fleet" data. Want to mimic real driving more. You could still do that with a number of vehicles doing stop and go, cruising, low speed, etc. testing. Even with simulated brakes to hot/cold cycle things and just run it around the clock. It would slow things down a lot, but a lot better than throwing test drivers at something that boring, at likely way lower cost in the end.
 
Upvote
9 (9 / 0)

Turbofrog

Ars Tribunus Militum
2,467
The vehicle followed the same path I saw when I arrived. A thick layer of rubber was building up on the asphalt. The BMW had traveled this virtual road all day, and it still had hours to go.
Legitimate question here: if the vehicles are following the exact same path, will that thick layer of tire rubber not start to affect the real world relevance of the data, depending on the testing conditions you're trying to mimic? I'm not a drag racer, but I understand that the starting line has a layer of rubber that is critical to provide traction for obscenely powerful cars and that on an unprepared track, the initial rollout will be significantly slower.

I guess I just sort of assumed that if BMW has such extreme granular control, and such a large area for testing, they would just periodically program in offsets of the whole route - of, say, 1.5 tire widths - in order to ensure that the vehicle is driving on pavement that is more representative of normal roads.
 
Upvote
-10 (5 / -15)

ShortOrder

Ars Scholae Palatinae
1,168
Legitimate question here: if the vehicles are following the exact same path, will that thick layer of tire rubber not start to affect the real world relevance of the data, depending on the testing conditions you're trying to mimic? I'm not a drag racer, but I understand that the starting line has a layer of rubber that is critical to provide traction for obscenely powerful cars and that on an unprepared track, the initial rollout will be significantly slower.

I guess I just sort of assumed that if BMW has such extreme granular control, and such a large area for testing, they would just periodically program in offsets of the whole route - of, say, 1.5 tire widths - in order to ensure that the vehicle is driving on pavement that is more representative of normal roads.
So I'm not sure about BMW's facilities but the GM Milford proving ground has 147 miles of roads for testing and an area they call the "Black Lake" that's about 75 acres of pavement
 
Upvote
4 (4 / 0)

luckof13

Wise, Aged Ars Veteran
148
Subscriptor
Want 120k miles of wear study? You could get that in 3 months of around the clock testing with stops just for fuel and maintenance. Throw 10 vehicles out there to get a little bit better "fleet" data. Want to mimic real driving more. You could still do that with a number of vehicles doing stop and go, cruising, low speed, etc. testing. Even with simulated brakes to hot/cold cycle things and just run it around the clock. It would slow things down a lot, but a lot better than throwing test drivers at something that boring, at likely way lower cost in the end.
Ford has been using robotic drivers for it's durability testing for at least a decade now. They get more constant results in less time than human test drivers. It's safer too as human drivers aren't exposed to repeated rough roads and pothole impacts.
https://www.nytimes.com/2013/06/16/...ad-but-a-robot-driver-takes-it-in-stride.html
 
Upvote
14 (14 / 0)

AvianLyric

Smack-Fu Master, in training
80
Subscriptor++
I don't believe brake-by-wire is legal. Drive-by-wire is a bit of a misnomer;
Break-by-wire (zero mechanical linkages) is completely legal and BMW have been producing cars with it for a few years now.

Even full drive-by-wire (zero mechanical linkages for their brakes or steering) are legal. But there’s currently no production cars with it.

But even in a brake-by-wire system, there’s a full force feedback (actuator acting on the brake pedal, so the driver can feel what the brakes are doing). So if the communication with the break-by-wire system fails, and the system falls-back to its standard config, then you need something that can produce equivalent to human braking power. Otherwise you’ll never overcome the force feedback from the brake-by-wire system.

You could probably modify your break-by-wire system to completely disable the force feedback. But why take that risk? You don’t want your emergency stop system failing because someone forgot to disable force feedback.
 
Upvote
4 (6 / -2)

AvianLyric

Smack-Fu Master, in training
80
Subscriptor++
Legitimate question here: if the vehicles are following the exact same path, will that thick layer of tire rubber not start to affect the real world relevance of the data, depending on the testing conditions you're trying to mimic?
Covered earlier in the article. The cars offset each run to avoid this exact problem. The later statement in the article is taking some artistic license with the actual behaviour, and just makes the article confusing. While I appreciate the little artistic flourish and embellishment in the article, I personally think the article would be better without this specific embellishment.
 
Upvote
13 (13 / 0)

Pluvia Arenae

Ars Tribunus Militum
2,844
Subscriptor++
Legitimate question here: if the vehicles are following the exact same path, will that thick layer of tire rubber not start to affect the real world relevance of the data, depending on the testing conditions you're trying to mimic? I'm not a drag racer, but I understand that the starting line has a layer of rubber that is critical to provide traction for obscenely powerful cars and that on an unprepared track, the initial rollout will be significantly slower.

I guess I just sort of assumed that if BMW has such extreme granular control, and such a large area for testing, they would just periodically program in offsets of the whole route - of, say, 1.5 tire widths - in order to ensure that the vehicle is driving on pavement that is more representative of normal roads.
Looks like you missed this sentence:
To make sure the brake tests are consistent, the vehicles move slightly over between tests so as to not be hindered or helped by any rubber left on the road.
 
Upvote
15 (15 / 0)
All driving is a boring, repetitive test, though the stakes are a bit higher for flesh and blood. The 42,795 Americans who got an F on such tests in 2022 died. We have no statistics for grades C or D, as the maimed, crippled, and traumatized go uncounted, and the sickness and death that indirectly results from paving the planet is impossible to calculate.

Teleologically speaking, death is the primary function of our most common form of transportation. We intend for it to serve many purposes, but the most important thing it achieves is to kill people by bending, breaking, battering, and crushing them. If your a slow reader, the highway system has killed one or two Americans since you began this mild rant.

The automobile ranks very high among humanity's many species-wide suicide attempts. Henry Ford was an anti-humanist monster who should no more be celebrated than Christopher Columbus.

Perhaps autonomous cars will improve the situation; perhaps not. The only thing we can say with any surety is that we will remain indifferent to the suffering caused by the meat grinding death machine that is the automobile.
 
Upvote
-17 (4 / -21)

Pluvia Arenae

Ars Tribunus Militum
2,844
Subscriptor++
All driving is a boring, repetitive test, though the stakes are a bit higher for flesh and blood. The 42,795 Americans who got an F on such tests in 2022 died. We have no statistics for grades C or D, as the maimed, crippled, and traumatized go uncounted, and the sickness and death that indirectly results from paving the planet is impossible to calculate.

Teleologically speaking, death is the primary function of our most common form of transportation. We intend for it to serve many purposes, but the most important thing it achieves is to kill people by bending, breaking, battering, and crushing them. If your a slow reader, the highway system has killed one or two Americans since you began this mild rant.

The automobile ranks very high among humanity's many species-wide suicide attempts. Henry Ford was an anti-humanist monster who should no more be celebrated than Christopher Columbus.

Perhaps autonomous cars will improve the situation; perhaps not. The only thing we can say with any surety is that we will remain indifferent to the suffering caused by the meat grinding death machine that is the automobile.
I'm very sorry that you apparently lost someone, but maybe get therapy instead of doing whatever that was?

Edit: And maybe learn the actual meaning of "teleological."
 
Last edited:
Upvote
9 (9 / 0)
I don't believe brake-by-wire is legal. Drive-by-wire is a bit of a misnomer; it refers just to the throttle. With electric power steering, computers can control the front wheels. Brakes are mechanical, where pressing on the pedal moves a piston to "compress" the brake fluid. (Brake system pressure can be in excess of 2000psi!) This squeezes the brake pads against the rotor, slowing the car down.

The failure they are concerned about, and why you need a beefy robot, is if there is an issue with power-assisted brakes. Every modern car uses engine power to assist the human's braking capability. If you lose the brake booster, it takes a tremendous amount of effort to get even a modest braking result. The robot has to be able to stop the car in the rare, but potential, case of power-assisted brake failure. And that's why it is so beefy.
EV’s, and many modern hybrids, have an electrical brake booster. I believe Bosch calls theirs an iBooster (because it was invented in the i-age).

It works similar, but can also perform brake by wire by actuating the master cylinder on its own. The way it’s implemented in my Tesla the pedal also moves when this happens as they’re linked together, but I suppose it could be possible to have a one-way linkage, but not sure if there’s any point to that.

So yes, there is brake by wire, but there’s also a physical backup, but if there’s no assistance you need to push hard (just like if your engine dies/stops in an vacuum assisted ICE vehicle.
 
Upvote
4 (4 / 0)

azazel1024

Ars Legatus Legionis
15,085
Subscriptor
Ford has been using robotic drivers for it's durability testing for at least a decade now. They get more constant results in less time than human test drivers. It's safer too as human drivers aren't exposed to repeated rough roads and pothole impacts.
https://www.nytimes.com/2013/06/16/...ad-but-a-robot-driver-takes-it-in-stride.html
Neat! Thank you for sharing. This is why I love Ars.

Also, now I am sad that the auto companies are, in fact still, smarter than I am :)
 
Upvote
1 (1 / 0)

azazel1024

Ars Legatus Legionis
15,085
Subscriptor
EV’s, and many modern hybrids, have an electrical brake booster. I believe Bosch calls theirs an iBooster (because it was invented in the i-age).

It works similar, but can also perform brake by wire by actuating the master cylinder on its own. The way it’s implemented in my Tesla the pedal also moves when this happens as they’re linked together, but I suppose it could be possible to have a one-way linkage, but not sure if there’s any point to that.

So yes, there is brake by wire, but there’s also a physical backup, but if there’s no assistance you need to push hard (just like if your engine dies/stops in an vacuum assisted ICE vehicle.
You only have to push hard if the booster fails. If the engine stops/stalls, you only have to push hard after exhausting the stored vacuum in the brake booster. Go sit in your car with it off and pump the brakes. After 2-3 pushes it'll start getting harder and probably in 4 or 5 the pedal will feel rock hard.

Just wanted to correct that. If the engine stops or stalls, you don't immediately loose the power assisted brakes. Generally vacuum steering boosters you'll lose right away, but that is also not much of a concern unless traveling at very low speeds, because the level of assist is generally pretty low at anything much above parking lot speeds. It is intentional in brake booster design so that there is a reservoir of vacuum to not have to rely only entirely on the engine (and there is a one-way valve in the booster, so again, if the engine shuts off, you don't lose the vacuum force in the booster).
 
Upvote
4 (4 / 0)
Post content hidden for low score. Show…

wirrbeltier

Wise, Aged Ars Veteran
182
Can Robo-Zambonis be far behind?!
Not quite the same, but autonomous floor (wet) scrubbers do exist. I used to work at a teaching hospital that had a few roughly shopping-cart sized bots slowly scrubbing the long hallways.

The bots were outfitted with a custom paint job to make them look like miniature ambulances, and would beep plaintively at anyone coming too close. Quite cute, and I hope they just freed the cleaning crew from a chore, rather than giving management a pretext to downsize their workforce.
 
Upvote
2 (2 / 0)

Turbofrog

Ars Tribunus Militum
2,467
Looks like you missed this sentence:
Covered earlier in the article. The cars offset each run to avoid this exact problem. The later statement in the article is taking some artistic license with the actual behaviour, and just makes the article confusing. While I appreciate the little artistic flourish and embellishment in the article, I personally think the article would be better without this specific embellishment.
I did miss that sentence. Thanks. I was indeed confused by the artistic flourish, glad that the engineers at BMW came to the same conclusion that I did.
 
Upvote
3 (3 / 0)

Arstotzka

Ars Scholae Palatinae
1,233
Subscriptor++
Break-by-wire (zero mechanical linkages) is completely legal and BMW have been producing cars with it for a few years now.

Even full drive-by-wire (zero mechanical linkages for their brakes or steering) are legal. But there’s currently no production cars with it.

But even in a brake-by-wire system, there’s a full force feedback (actuator acting on the brake pedal, so the driver can feel what the brakes are doing). So if the communication with the break-by-wire system fails, and the system falls-back to its standard config, then you need something that can produce equivalent to human braking power. Otherwise you’ll never overcome the force feedback from the brake-by-wire system.

You could probably modify your break-by-wire system to completely disable the force feedback. But why take that risk? You don’t want your emergency stop system failing because someone forgot to disable force feedback.

Brake-by-wire still requires a backup physical connection. There may not be engine vacuum giving you extra strength, but if there's zero power the brake pedal still has an effect because it will physically actuate the brake master cylinder. Hybrids and EVs have more complicated systems to manage the transition between regenerative and friction braking, but if the electrical system goes completely dead the brake pedal still must engage the friction brakes. (I literally just changed the brakes on a Prius, and to bed in the pads/rotors, I pulled the relays controlling this system. The computer gets very mad, and the brake pedal requires a fair bit of force, but it still works.) Likewise, steer-by-wire also has a failsafe relay that will, in the absence of power or other failure, cause a clutch to engage and make the steering wheel move the wheels.

Electronic throttles, on the other hand, have no such such mechanical fallbacks. If the hall sensor on the accelerator pedal stops working (magnets, how do they work?) the engine will continue to idle but won't accelerate. There's no legal requirement for a backup linkage.

I'm not aware of a legal system in production road-going vehicles that has zero mechanical connection for steering or brakes.
 
Upvote
4 (4 / 0)

Arstotzka

Ars Scholae Palatinae
1,233
Subscriptor++
If it wasn't then every hybrid and most BEVs on sale today would be illegal. BBW is exactly how you get blended regen/friction braking when using the pedal.
EV’s, and many modern hybrids, have an electrical brake booster. I believe Bosch calls theirs an iBooster (because it was invented in the i-age).

It works similar, but can also perform brake by wire by actuating the master cylinder on its own. The way it’s implemented in my Tesla the pedal also moves when this happens as they’re linked together, but I suppose it could be possible to have a one-way linkage, but not sure if there’s any point to that.

So yes, there is brake by wire, but there’s also a physical backup, but if there’s no assistance you need to push hard (just like if your engine dies/stops in an vacuum assisted ICE vehicle.
Edited my original comment; yes, electronic boosters exist to blend friction and regen brake systems. I was (in my head!) comparing drive-by-wire, i.e. electronic throttle with no mechanical backup, to brake-by-wire, which always has a mechanical system failsafe; even if that failsafe requires a fair bit of extra effort to actuate.
 
Upvote
1 (1 / 0)

siliconaddict

Ars Legatus Legionis
13,050
Subscriptor++
BMW uses autonomous cars for boring, repetitive tests

Well I personally consider BMW's car designs to be boring so I guess that works out. Seriously they are risk averse as there is no reason to have a freaking grille that large. And the lack of a Frunk in the i5's case just screams lets just retrofit our line instead of bothering to make anything new from the ground up. I considered a BMW EV but nothing about their designs get me excited.
 
Upvote
0 (1 / -1)

NetMage

Ars Tribunus Angusticlavius
9,974
Electronic throttles, on the other hand, have no such such mechanical fallbacks. If the hall sensor on the accelerator pedal stops working (magnets, how do they work?) the engine will continue to idle but won't accelerate. There's no legal requirement for a backup linkage.
As far as I am aware, all electronic accelerator pedals have dual sensors setup in a reverse configuration to provide redundancy.
I'm not aware of a legal system in production road-going vehicles that has zero mechanical connection for steering or brakes.
Toyota is planning to have it on one of their BEVs soon, though I supposed the brakes might have a mechanical fallback, the steering will not.
 
Upvote
4 (4 / 0)