Hacker News Reader

Comment by poulpy123 | Hacker News Reader

what I don't get is why the effect of a crash on other isn't taken into account in the evaluation of car safety

The impact to pedestrians is taken into account. The impact on other cars isn't, because I don't see how you'd do that objectively - do you just crash it against an identical model? A selection of several other cars from different categories/brands? How would you determine those? Do you just test head-on collisions, or the whole gauntlet of impacts? Wouldn't that make any test require dozens and dozens of vehicles, and ultimately tell you very little because the variety of vehicles on road is so huge that your crash results against a Honda Jazz might mean nothing in a crash against a Civic or anything else.

Well I think one can start with a bit of an average car. I mean humans vary a lot too. Both in size and resilience. If there’d be some wireframe made that can represent a bit of the car rigidity of an average 2008 car for example we can definitely see how this monolith would vaporize any car, versus the Jazz that’d mostly dent itself in a crash. And include that in the rating as well. Current SUV trends are abysmal for others with their 800bhp and near 3000kg kerb weight. It’s I think a solid way to stop this race to the bottom.

it would be possible to test the effect of the car on a (or several) standardized objects, and compare the car resistance to these standardized objects

The issue with that is that manufacturers would obviously work to make sure their cars score super well against those specific objects. Would you be able to make sure it translates well to collisions against real life cars in real life situations? I'm not sure - usually when you start judging manufacturers on a certain metric they make sure to do well on that one metric at the cost of everything else.

Avoiding Goodhart's law is a general problem, and in this case it's the regulator's problem.

No crash test is perfect, but some are useful. Cars don't change that much from year to year and manufacturers tend to copy each other quite a bit. Developing a suite of tests against a selection of typical vehicles seems very doable.

Existing crash tests don't perfectly cover every scenario, and people quibble about the details, but the tests still provide useful information.

This is just an initial response, so probably flawed, but I'd suggest testing against a vehicle which only exists to represent the lowest safety rating (for occupants) that is allowed on the roads of whichever jurisdiction this is.

If the new design results in harm to the passengers, it's unsafe for existing legal road users.

well a strong enough crash will always result in harm to either or both sides. One still has to define the type of crash etc.

I hate these minitanks but I acknowledge it's not simple to test like this.

Make heavier car forbidden on some roads or make them pay proportionally to a ratio of their empty weight vs. weight fully loaded would seem fair. This way trucks are not impacted, but stupidly heavy SUVs are. And apply a factor for EV, due to battery.

> well a strong enough crash will always result in harm to either or both sides. One still has to define the type of crash etc.

Sure, but that's way beyond the scope of a comment on a web forum to enumerate let alone describe in detail. Whatever the current survivability standards are, they'll say speeds and directions etc., so reuse those but with the new car as the other object in the collision.

It is. One of the key differences between the EU and US regulations is (was?) that there is a greater evaluation of what happens to a pedestrian when hit by the car being tested.

I hope this car never makes it into Europe in current form.

Tesla have made statements saying that it won't. As a layman, I don't think the flat front is possible.

Sharp edges are also not allowed in most European countries.

Reply on news.ycombinator.com