4 ms·

My understanding is that car brakes can generally exert more braking force than the tire's grip on the road. A long stopping distance may represent a problem wi

by gnode 8y ago

My understanding is that car brakes can generally exert more braking force than the tire's grip on the road. A long stopping distance may represent a problem with the anti-lock braking system (which is likely software controlled, and can be updated over the air).

Animats 8y ago
Interestingly, Waymo's self-driving car cannot be updated remotely.[1] As a security measure, Waymo deliberately did not install a data connection to allow that. Probably a good decision. [1] http://www.documentcloud.org/documents/4458000-Waymo-Driverless-AV-Application.html http://www.documentcloud.org/documents/4458000-Waymo-Driverl...
- gnode 8y ago
  That's probably also to do with Waymo's cars being developed primarily for ride hailing fleets. With a private ownership model, requiring owners of variable technological faculty to manually apply updates would be less practical.
- londons_explore 8y ago
  That was more a PR move than a real security measure. Signing updates and checking signatures is established technology to certify that updates come from the right place and are not tampered with. That and the updates are likely >10 GB, so you wouldn't want to download them over mobile data anyway.
magduf 8y ago
It could, but it could also be a function of the car's weight (batteries are heavy) and the tires it comes equipped with from the factory. Poor tires will yield poor braking performance due to lack of grip.
- dragontamer 8y ago
  But it can't be poor tires, because the Model 3's brakes were proper in the first test (~130 feet stopping distance). The same car with the same tires then had different braking behavior in every test after the first test. That's just odd.
  - londons_explore 8y ago
    Most vehicles detect a sharp press to the brakes, and add extra assistive braking force. Thats from the observation that most humans do not apply enough braking force in an accident (potentially because they have never applied that much force before, and don't know how much is 'enough'). It's possible that the tesla software either isn't applying the extra assistive force, either due to incorrect tuning or a bug.
  - cc439 8y ago
    I can almost guarantee the problem is caused by the brake pads glazing over. The working surface of a brake pad will become work hardened, which drastically lowers their friction coefficient, once pushed past their designed temperature limit. The Model 3 (3,800lbs for the long range model) is a fat pig of a car compared to other, conventionally powered cars in its class (3,300lbs for a BMW 3 Series) but this is countered by Tesla fitting them with relatively enormous brakes (14" rotors up front, 13.5" in the rear, larger rotors have more heat absorbing capacity for the pads). If I had to guess the root cause for this problem, I'd say it is likely down to Tesla's focus on minimizing aerodynamic drag causing a lack of airflow within the wheelwell and across the brakes out the wheel itself. Either that or they chose an absolutely terrible brake pad compound but that just doesn't seem likely in this modern era. Either way, the brakes are getting so hot after a single stop that CR's cooling procedure (a reasonable one I might add) is inadequate to keep the brake pads from drastically overheating. This also explains why braking performance was still compromosed after letting the car cool overnight. The work hardened surface will perform poorly until it is literally ground off from further heavy use of the brakes.
    - dragontamer 8y ago
      > I can almost guarantee the problem is caused by the brake pads glazing over. That's actually a very reasonable explanation. I can't say I'm a car expert, but your logic at least sounds correct.
- gnode 8y ago
  I'm not sure how true this is. As the car's weight increases so too does the grip, as it's a function of down-force multiplied by the roughly constant coefficient of friction of the tires. Conceivably, the car's weight may mean that aerodynamic down-force produces a relatively smaller impact on grip, and air resistance a relatively smaller deceleration. But I'm not sure that's enough to account for the difference.
  - dragontamer 8y ago
    Excellent racing tires have coefficient of friction > 1. That is, a 2-ton car can have more than 2-tons of force applied from the tires. But most consumer tires have coefficient of friction of less than 1. ~.9 for higher end stuff, but it could be as low as ~.5 or so. Which mean 2-tons of car will lead to LESS than 2-tons of acceleration / deceleration / turning force. In effect, "more weight is more advantage" is only true if you have a coefficient of friction greater than 1.
    - londons_explore 8y ago
      The last line of your post isn't clear. Since F=MA, A=F/M. Assuming the braking force (F) is μ (coefficient of friction) * Mass * Gravity, thats F=μMG. Combining those equations you get: A = μMG / M And the M's cancel, giving: A = μG Therefore the braking distance due to road friction alone (and ignoring aerodynamic downforce effects) is independent of the mass of the vehicle.
      - magduf 8y ago
        Isn't this neglecting the forward velocity and momentum? According to theory, the width of a tire has zero effect on tire grip. Race cars and high-performance cars prove this is incorrect: there's a reason they have wide tires (which also cause poorer fuel economy through extra friction). Tires and their interaction with the road are a lot more complex than some simple physics equations.
      - dragontamer 8y ago
        Hmm, that makes sense. Its like the whole "area" doesn't matter as well. I can't help but think that maybe the Newtonian friction model is overly simple sometimes. Surely weight and surface area would have an effect? But perhaps not really. Or at least, not in a way that's very measurable.
dragontamer 8y ago
But braking is a solved problem. Apply the maximum force to the tires, until right before the tires begin to slip. Then, reduce the force to prevent slipping. Stay within the "static coefficient of friction", which is higher than "dynamic coefficient of friction". Alternatively, there's also "let the human do it". (Aka: threshold braking), but most people prefer to have the computer do this task during an emergency. This algorithm has been used in anti-lock brakes for decades at this point. I'm not sure why it would need an OTA update to improve. Once this algorithm is properly implemented, then the only thing left is to buy better tires and use better rubber to get better grip. This CR report suggests that Tesla's braking behavior is suboptimal. There's no reason for a smaller, lighter Model 3 to have a worse braking distance than the freaking F-150 truck.
- gnode 8y ago
  > But braking is a solved problem. That's not to say Tesla haven't decided to re-solve it / implement it themselves anew. The rest of the auto industry has been doing ABS much longer than Tesla has existed. It's also not a prescriptive algorithm. ABS systems don't stay within the static coefficient of friction; they repeatedly retest the boundary between static and dynamic friction (hence the judder). It's a bit like optimising TCP window size for those familiar with it. There are a bunch of parameters you can tune.
  - londons_explore 8y ago
    It also depends what your goal is. Typical ABS systems typically perform far worse on gravel or sand than non-abs systems for example. It turns out on gravel locking the wheels is usually a better solution because then you start to plough deep into the gravel. No ABS system could determine that beforehand, since it isn't to know if the gravel is deep (allowing ploughing deep into it) or shallow (where staying on the boundary of static/dynamic friction works better).

by gnode 8y ago

Animats 8y ago

Interestingly, Waymo's self-driving car cannot be updated remotely.[1] As a security measure, Waymo deliberately did not install a data connection to allow that. Probably a good decision. [1] http://www.documentcloud.org/documents/4458000-Waymo-Driverless-AV-Application.html http://www.documentcloud.org/documents/4458000-Waymo-Driverl...

gnode 8y ago
That's probably also to do with Waymo's cars being developed primarily for ride hailing fleets. With a private ownership model, requiring owners of variable technological faculty to manually apply updates would be less practical.
londons_explore 8y ago
That was more a PR move than a real security measure. Signing updates and checking signatures is established technology to certify that updates come from the right place and are not tampered with. That and the updates are likely >10 GB, so you wouldn't want to download them over mobile data anyway.

magduf 8y ago

It could, but it could also be a function of the car's weight (batteries are heavy) and the tires it comes equipped with from the factory. Poor tires will yield poor braking performance due to lack of grip.

dragontamer 8y ago
But it can't be poor tires, because the Model 3's brakes were proper in the first test (~130 feet stopping distance). The same car with the same tires then had different braking behavior in every test after the first test. That's just odd.
- londons_explore 8y ago
  Most vehicles detect a sharp press to the brakes, and add extra assistive braking force. Thats from the observation that most humans do not apply enough braking force in an accident (potentially because they have never applied that much force before, and don't know how much is 'enough'). It's possible that the tesla software either isn't applying the extra assistive force, either due to incorrect tuning or a bug.
- cc439 8y ago
  I can almost guarantee the problem is caused by the brake pads glazing over. The working surface of a brake pad will become work hardened, which drastically lowers their friction coefficient, once pushed past their designed temperature limit. The Model 3 (3,800lbs for the long range model) is a fat pig of a car compared to other, conventionally powered cars in its class (3,300lbs for a BMW 3 Series) but this is countered by Tesla fitting them with relatively enormous brakes (14" rotors up front, 13.5" in the rear, larger rotors have more heat absorbing capacity for the pads). If I had to guess the root cause for this problem, I'd say it is likely down to Tesla's focus on minimizing aerodynamic drag causing a lack of airflow within the wheelwell and across the brakes out the wheel itself. Either that or they chose an absolutely terrible brake pad compound but that just doesn't seem likely in this modern era. Either way, the brakes are getting so hot after a single stop that CR's cooling procedure (a reasonable one I might add) is inadequate to keep the brake pads from drastically overheating. This also explains why braking performance was still compromosed after letting the car cool overnight. The work hardened surface will perform poorly until it is literally ground off from further heavy use of the brakes.
  - dragontamer 8y ago
    > I can almost guarantee the problem is caused by the brake pads glazing over. That's actually a very reasonable explanation. I can't say I'm a car expert, but your logic at least sounds correct.
gnode 8y ago
I'm not sure how true this is. As the car's weight increases so too does the grip, as it's a function of down-force multiplied by the roughly constant coefficient of friction of the tires. Conceivably, the car's weight may mean that aerodynamic down-force produces a relatively smaller impact on grip, and air resistance a relatively smaller deceleration. But I'm not sure that's enough to account for the difference.
- dragontamer 8y ago
  Excellent racing tires have coefficient of friction > 1. That is, a 2-ton car can have more than 2-tons of force applied from the tires. But most consumer tires have coefficient of friction of less than 1. ~.9 for higher end stuff, but it could be as low as ~.5 or so. Which mean 2-tons of car will lead to LESS than 2-tons of acceleration / deceleration / turning force. In effect, "more weight is more advantage" is only true if you have a coefficient of friction greater than 1.
  - londons_explore 8y ago
    The last line of your post isn't clear. Since F=MA, A=F/M. Assuming the braking force (F) is μ (coefficient of friction) * Mass * Gravity, thats F=μMG. Combining those equations you get: A = μMG / M And the M's cancel, giving: A = μG Therefore the braking distance due to road friction alone (and ignoring aerodynamic downforce effects) is independent of the mass of the vehicle.
    - magduf 8y ago
      Isn't this neglecting the forward velocity and momentum? According to theory, the width of a tire has zero effect on tire grip. Race cars and high-performance cars prove this is incorrect: there's a reason they have wide tires (which also cause poorer fuel economy through extra friction). Tires and their interaction with the road are a lot more complex than some simple physics equations.
    - dragontamer 8y ago
      Hmm, that makes sense. Its like the whole "area" doesn't matter as well. I can't help but think that maybe the Newtonian friction model is overly simple sometimes. Surely weight and surface area would have an effect? But perhaps not really. Or at least, not in a way that's very measurable.

dragontamer 8y ago

But braking is a solved problem. Apply the maximum force to the tires, until right before the tires begin to slip. Then, reduce the force to prevent slipping. Stay within the "static coefficient of friction", which is higher than "dynamic coefficient of friction". Alternatively, there's also "let the human do it". (Aka: threshold braking), but most people prefer to have the computer do this task during an emergency. This algorithm has been used in anti-lock brakes for decades at this point. I'm not sure why it would need an OTA update to improve. Once this algorithm is properly implemented, then the only thing left is to buy better tires and use better rubber to get better grip. This CR report suggests that Tesla's braking behavior is suboptimal. There's no reason for a smaller, lighter Model 3 to have a worse braking distance than the freaking F-150 truck.

gnode 8y ago
> But braking is a solved problem. That's not to say Tesla haven't decided to re-solve it / implement it themselves anew. The rest of the auto industry has been doing ABS much longer than Tesla has existed. It's also not a prescriptive algorithm. ABS systems don't stay within the static coefficient of friction; they repeatedly retest the boundary between static and dynamic friction (hence the judder). It's a bit like optimising TCP window size for those familiar with it. There are a bunch of parameters you can tune.
- londons_explore 8y ago
  It also depends what your goal is. Typical ABS systems typically perform far worse on gravel or sand than non-abs systems for example. It turns out on gravel locking the wheels is usually a better solution because then you start to plough deep into the gravel. No ABS system could determine that beforehand, since it isn't to know if the gravel is deep (allowing ploughing deep into it) or shallow (where staying on the boundary of static/dynamic friction works better).