Tesla Cybertruck uses steer-by-wire with almost no redundancy. Electric power steering was bad enough to cause crashes, this is just ambulance chaser's dream.

>steer-by-wire

This is the norm for most cars with servo steering.

No. Almost all cars except some concepts and prototypes have mechanical link between the steering wheel and the steering rack. Even if the link is augmented by a hydraulic or electronic booster.

If a booster fails you have a chance of overpowering it, it's not that strong. A faulty steer-by-wire will swing you into the oncoming lane with no recourse.

I think the lack of redundancy is the issue. Redundant electrical connections can be very robust otherwise they wouldn't put them in F-35s.

Fighter jets are falling out of the sky quite often. Compare to airbus passanger jets instead.

Partial control failures on passenger jets are not that uncommon.

But no ejection seats to escape a collision.

Anywhere in aviation it's redundant transducers (the thingies that transform rotation into some sort of signal), redundant flight law computers implemented by independent teams on dissimilar hardware, a stupidly robust vote/compare module, I think mostly analog, and usually a last resort direct mode.

When all of that is implemented in cars and weathers for 10-20 years I might consider it.

Otherwise, seeing as Toyota's regen braking code had like 10k global variables, I'm not touching that with at 10 foot pole.

Acceleration/braking can be quite bad but there's normally neutral and ignition switches. With steering there's literally nothing between you cruising along a highway and you splattered on some concrete pillar 200ms later.

That is more than likely generated code. Model based development and code generation (by certified toolchains) are the norm in automotive and aerospace.

Nope, all hand-written apparently.

The F-35 took like 25 years and a staggering amount of billions to develop...

Fly by wire has been a thing since the 80s, F-18 IIRC. At some point we'll have drive by wire, it just makes sense.

This doesn't necessarily follow. Planes became pressurized and got oxygen masks, it's very unlikely that cars ever do. Having steel cables and/or hydraulic lines for the 4-6 dimensions of aircraft controls adds up to some serious extra weight worth fighting for on a plane, while one steel rod might never be more problematic than the heap of problems its absence creates.

As long as the steering wheel is there it might as well be connected to the rack, forever.

EVs can have multiple motors and those motors can be very close to wheels, maybe even inside wheels.

It stops making sense to have a bunch of rods zigzagging everywhere.

Motors have nothing to do with this, you still have to have the suspension struts, and the wheels still have to be mechanically linked by the tie rod and usually the sway bar.

Also inside the wheels makes no sense at all, google unsprung weight.

Nope, ColPAS and ParPAS solutions both normally have the mechanical connection of the steering column between the rack and pinion mechanic and the steering wheel, they mainly differ by the way of introducing the mechanical assist force to the system. (on the steering column, or at the rack and pinion mechanism).

They generally have the following root safety requirements:

- the assist system must not cause "blocking" (unturnable wheels/steering wheel)

- the assist system must not cause unintended steering

- the assist system must be able to restart in less than 20 milliseconds

- the assist system must stay passive after a number of unsuccessful startups/unrecoverable errors (typically 3). In this case the mechanical steering column still provides a way for the driver to control the car.

It has been 10+ years since I last worked such a system, but drive by wire was only on the horizon, not the norm, and as far as I know the authorities only allowed drive by wire solution some 2-3 years ago, with product development lifecycle of around 5 years that means they are still not the norm among new cars I believe.

Back then more redundancy in the electric motors (more gradual failure if one phase goes bust, not going into deadlock in certain angles when any of the 3 phases is malfunctioning) and authenticated communications on the system bus (which was considered trusted earlier) were on the roadmap for the next gen steering systems, which are mandatory requirements before you can get to steer by wire.

Why even put it on the bus? Adding a couple of meters of cable per critical system like steering or even 10 for braking couldn't offset the complexity of bus steering.

On the contrary, adding just "a few meters of cable" per each critical system, and adding separate sensors instead of using the same critical sensors would greatly increase complexity. Also so really underestimate the cabling and sensor needs for modern requirements.

Having redundant bus and redundant sensors usable by multiple subsystems is a better way.

Well, I have heard of power assist using a single simple potentiometer as a steering wheel position sensor in some really cheap cars... I'm not sure how much of an urban legend that is, but nowadays it surely wouldn't cut it for road permission in the EU.

Compared to classic hydraulic lines everywhere dedicated cables seem easy.

Having to account for a DOS attack from a compromised head unit or some such just seems like a nightmare.

On one hand you are right, but only if we can ignore the highly increased risk of electric malfunction from the extensive traditional (non bus based) wiring. This was a common problem in the 1980s when electric gadgets became numerous, but no standard bus was used on many such cars.

To overcome this the plan back when I was working on that product was to use 2 system buses, one isolated internal for the safe critical sensor/control network, and one user facing untrusted bus, with a ECU designed to serve as a firewall basically.

Why even have the firewall? Wouldn't a true airgap be preferable? At the same time, having say steering and brakes on the same bus adds a single point of failure to both steering and brakes, which is not ideal.

Cost and risk management. These systems undergo rigorous safety evaluation, auditing, and testing. This is a manageable risk (very low) given the business requirements and the threat model.

There is zero chance of me trusting that, not that the Cybertruck will ever be allowed on European roads.

https://www.yahoo.com/news/man-alarmed-cybertrucks-steering-...

> with almost no redundancy

Why are you just making stuff up? Cybertruck has lots of redundancy.

Do you have any actual evidence that Cybertruck steer-by-wire design is bad?