Cars need new architectures!
A modern car is a complex network on wheels. Vehicles now have up to 150 control units. The wiring harness consists of up to 5 km of cables and can therefore weigh more than 50 kg. Up to 25GB of data is exchanged via this huge network - every hour! And the number of lines of code is now up to 150 million.
But we are still a long way from the end: autonomous vehicles are expected to require 500 million lines of code, possibly more. Traditional architectures for cars can no longer cope with this. I will discuss why this is the case and what the alternatives are below.
Vehicle architecture today
The architecture of established vehicles has evolved from the fact that the combustion engine was at the center. Engine technology was previously the most important know-how of the manufacturers. This is why they started to outsource everything else to suppliers as far as possible many decades ago. Initially these were mechanical components, later electronics were added. In the process, all components were Black Box treated.
But over time, these components contained more and more electronics and also control units, which were then connected to a common data bus. Also, or precisely because manufacturers treated them like black boxes, they had more and more unexpected or unwanted interactions.
Is it possible to determine whether a change to the software in one vehicle control unit affects another? 16% of the developers surveyed say it is impossible.
IEEE Spectrum
The fact that decentralized architectures in cars have reached their limits is also confirmed by Continentalwho have developed a new architecture that is already in use in the VW ID.3 is used.
Tesla shows the way
Continental relies on a server-based, function-oriented vehicle architecture based on a central "In-Car Application Server" (ICAS1).
The role model is - how could it be otherwise - Tesla. Tesla recognized early on that in a electric vehicle the software has such a central position that it must not be outsourced. In Model 3, Tesla has a central control unit and is years ahead of the industry with this architecture.
Such an architecture also mitigates the danger posed by the lack of resources. In January, one million fewer vehicles were produced worldwide than planned. In May, the figure was already over four million. The automotive industry is shaken. If one of 150 control units cannot be produced due to a chip shortage, production of the vehicle may be blocked. Tesla is in a much more robust position.
Electric car and AI result in unmanageable complexity
Also Andy Whydell from ZF Friedrichshafen sees the limits of controllabilitythat require new architectures for cars. He sees the main drivers here as Variation diversityDue to the permutations of the various components and control units, many vehicles that roll off the production line are unique.
In some cases, the automotive industry no longer considers the total number of lines of code as a measure of complexity, but rather the number of software developers that an OEM or supplier employs to meet current and future requirements.
Andy WhydellVice President, Product Planning, Vehicle Systems at ZF Group
A first important step is to reduce the number of control units in order to minimize integration problems. However, this means that a lot of control unit software cannot be used in a server architecture without further ado.
Conclusion: The newcomers have it easier
Despite the enormous efforts of the established OEMs, the question is whether companies such as VW, Daimler or BMW will be able to turn the corner. There will be enormous resistance from within their own ranks to hold on to the achievements of the past. To draw a comparison with the smartphone industry: Nokia could have invented the iPhone, but Apple did. Will we see something similar in the automotive industry? Only time will tell.
Image by Tayeb MEZAHDIA from Pixabay
