What's a reasonable expectation on how finished the software is on a new EV?

Massive issues with VW ID software across the range. Our own MG implementation is not exactly stellar.

One of the problems I always think is car manufacturers' insistence on developing their own systems and unique hardware which always seems to turn out a cludge. There are electronics companies out there who've been doing this stuff for decades and could do so much better.
Media? Sony (e.g.)
SatNav? TomTom
Integration? Apple or Google
 
You get the same with mobile android phones. Each manufacturer insists on installing their own version of the operating system together with a load of useless apps. That's why I usually buy a Moto G phone or a Google Pixel because you don't get any unwanted bloatware etc.
 
1. There is no such thing as bug free software.
2. There is no such thing as a totally compatible standards.
3. Software is not free. It is probably around 80-90% of a product development cost. That has to be recouped somehow!
4. Marketing keep making the car functionality more and more complex by promising new features that the software invariably was not originally designed to do. This increases the bugs.
5. An EV is probably the most complex consumer device that has been produced. It has to integrate so many things and the software is stretched and stretched. It then starts to break....
6. Software integration can be more complex and time consuming than building your own. Add the licence fees and financially it may not be worth it.
7. Integrating other software components may restrict what can be done with other parts of the system.
8. Marketing and customers insisting on Tick Box functions that ultimately no-one uses!
 
Learning the word "No" and how to use it would be a big step. All cars do the job they are intended for pretty well. Most are capable of going far faster than speed limits allow. How do you differentiate yourself?

You need to market your car as being better.... Faster, safer, more gizmos saving you time exercise etc. You need the ability for your car to create its own disco lighting and have a horn that plays the latest Taylor Swift song instead of just going beep. Easy to do! Meanwhile engineers now have to replace a cheap reliable horn unit with an mp3 player with a power amp and speaker and then link that into the infotainment system so that it can mute what you are listening to so that Taylor can spurt something about never ever together. That requires new expensive hardware and suddenly the infotainment starts to break down with the software overload so that what was basic but reliable now falls over. Taylor Swift starts interrupting at random. That gets fixed which then increases the power consumption because the unit can't use the powerdown sleep mode as it has to be ready to blast Taylor Swift at a moments notice so suddenly charging starts to fail because the voltage drop of 0.1v stops the CCS comms which then causes problems as the only solution is to fit different hardware because the software fix doesn't fit in the memory space because the Taylor Swift mp3 takes too much space and can't be shrunk because the licence says it must be the highest quality so that if a fan hears it, they are not disappointed ....

All this because marketing didn't say No.

I think that sums it up!
 
Software glitches will always occur in complex functions. However, some of the problems we have seen, in iSmart for example, should have been detected in the testing phase of development.
I've worked with software for many years and been amazed at some of the bugs which have gone undetected before launch.
 
Considering the questions raised by the video, I would expect the functions promised in the vehicle spec to be present and work.
The members of this forum are well informed but we are only a tiny minority of EV owners. Most owners expect to be able to get in a car and drive it without the need to be a computer expert.
The problem is that there is no IEC or BSI spec for a "Standard Idiot".
But who's the idiot? I don't think it's the average driver.
Should the driving test incorporate questions on the car's software functions?
 
But who's the idiot? I don't think it's the average driver.
Should the driving test incorporate questions on the car's software functions?
The "standard idiot" is an affectionate term within those in the software industry that are developing the user interface and how it operates. It applies to anyone using the interface. The problem is that no matter how hard you might try to think like a user will actually use the interface, there are always users that manage to find a key or a button sequence (often with other car conditions) that breaks the system.

This is why engineers wish there was a BSI (British Standard Idiot) definition that defines all the ways that a user interface can be used. Of course the complexity is such that there isn't but that is it. Common ones that I have come across including using a touch screen while wearing gloves, trying to use fingerprint recognition wearing gloves or sunglasses with facial recognition. These are pretty ridiculous but they happen. No one would do that? Would they?

Entering an engineering mode while stabbing the keyboard is a bit random but to a user that is a major fault. These sequences often require a 16 char password which the user won't guess at random. You can bet that someone will.

What happens if the user presses three buttons at the same time? Do nothing? Act confused? Do one of the operations and hope that it is the one that the user wanted?

Now add to that the ever evolving phone APIs and the subtle changes that now change as a result.

While this is a bit light hearted, it is a real problem and the more complex the interface, the higher the chance that the BSI will find the holes and break the software. Some certainly need fixing but a lot would take so much effort they won't. Some which are caused by 3rd parties can be absolute nightmares.

It might well be worthwhile adding something to the test re basic operation or even adding a quick start guide with the car.
 
With lots of consumer products you have a balance between development costs, time to market and and something customers are willing to accept. In cars with more buttons and fewer screens there were fewer single points of failure. When you have a central display that is mission critical then anything that causes that to fail is an issue.

When they developed the software for the Apollo and space shuttle they probably had quite a big testing budget. The same is probably true of missile guidance and autopilot software development.

However with consumer software we know that OS and app developers do not have a problem releasing a less than perfect product.

I guess the question is how much would more thoroughly tested software add to the price of the final product.
 
Software glitches will always occur in complex functions. However, some of the problems we have seen, in iSmart for example, should have been detected in the testing phase of development.
I've worked with software for many years and been amazed at some of the bugs which have gone undetected before launch.
Yes, the iSmart App is bloody useless! It worked for a week after a very painful unbind process followed by binding which does not work unless the GPS date option is enabled. Great cars, but much work to do via s/w
 
Stumbled upon this the other day, maybe the MG developers have adopted this technique and are just in meetings all day

 
Support us by becoming a Premium Member

Latest MG EVs video

MG Hybrid+ EVs OVER-REVVING & more owner feedback
Subscribe to our YouTube channel
Back
Top Bottom