Software that endangers lives

Sep 4, 2015 | 傳媒報導

From car-sharing, via electric cars to self-driving vehicles – the automotive industry is undergoing the perhaps most massive upheaval in its existence, with software as the most important factor in its center. The problem is that technological development often creates a level of software complexity, which overwhelms the manufacturers.
 

Technological change: An industry in transition

Hardly any other industry is after decades of prosperity in such an upheaval as the automotive industry. Combustion engines dominated the segment for decades and brought it with them that often only functional innovations and mechanical subtleties were spawned in an otherwise well-structured environment. It was about smaller innovations in an established piece of technology, which in its design became more complex, but not in its basic orientation. But the wind has started to turn in the so well-heeled automotive industry for quite some time and thereby unleashed a wave of upheavals with partly different orientation. Under headings such as “mobility” the sector discusses from car-sharing approaches to rebuilding the transport, while the most significant innovation role in the automotive sector belongs to the electric car these days.

In particular, Elon Musk’s electric car manufacturer Tesla, which is in the process of leading the market to a long expected revolution with its Tesla Model S, these days enjoys maximum attention. The electric car from Palo Alto previously impressively showed that electric vehicles are now suitable for everyday use and lifts the discussion about e-mobility to a new level. A level at which Germany has not arrived yet in many ways. Not surprisingly the much-publicized power struggle between VW patriarch Ferdinand Piech and CEO Martin Winterkorn was also about the very sustainability of the automotive industry.

While Tesla’s approach in E-Mobility puts the automotive companies under pressure from one side, Silicon Valley giants like Google with its self-propelled “robot cars” add pressure on the other. Especially on rural roads and highways the computerized vehicles of the search engine giant offer advantages and the first assessment of this ambitious project seems relatively promising. Already in 2020 the internet company wants to sell its first self-driving cars, pursuing a similarly ambitious plan like Daimler, which was more recently allowed to test autonomously driving trucks in Nevada.
 

Numerous blatant product recalls due to software

All these new scenarios – whether it’s car-sharing, electric vehicles or self-propelled robot cars – have one thing in common: software as a key component is at their center. Software has become the backbone of the automotive industry – from production processes to application in the final product. However, although software now plays such a central role, some manufacturers seem overwhelmed by its complexity sometimes. This can be seen in particular on the increase in the number of mass recalls due to software errors, which have occurred in recent times.

For instance, a software problem has forced Ford in July to recall around 433,000 vehicles in North America. In these vehicles a faulty electronic control unit could cause that the motor continues running, even though it was switched off by the driver. A new software was able to fix the loss of control, but it did require that hundreds of thousands of customers brought their vehicle to a repairing service.

In summer the Anglo-Indian automobile manufacturer Jaguar Land Rover started an unscheduled product recall for a total of 65,352 vehicles after a software error caused that the doors on vehicles without ignition could unlock while driving. If you believe the American media, in one case a door even jumped open while driving. In the same month as its American and Anglo-Indian competitors, the Japanese manufacturer Toyota called back a total of around 630,000 hybrid cars in Japan, North America and Europe due to a software-related risk: there was the risk that a software error could have led to an overheating of the system, so that the hybrid system would turn off while driving. An error, which forced the company to appoint holders of affected vehicles to a free software update.

Three highly professional automotive companies, three glaring recalls – and this in just one month. And the list of car recalls due to software errors certainly does not end here:

  • Already in May 2014 Ford was forced to recall 1.4 million vehicles in the US, Canada and Mexico, because a software error could cause problems with the airbag and the door handles
  • In October 2014, German carmaker Audi became victim of a problem and had to recall 850,000 Audi A4 because a software error could affect the airbag in the event of an accident.
  • Nissan called back around 23,000 Micra models by April 2015 because a software error could cause the car to suddenly accelerate without any driver intervention.
  • In May 2015 Volvo had to recall no less than six models after a software bug could cause the electronic parking brake to release when the car is started and a gear is engaged.

The list could probably still continue, as one can note that the number of recalled cars increasingly rises. With 1.9 million recalled vehicles in Germany, last year a new high on automotive recalls was recorded. In the US alone manufacturers had to recall 31 per cent more vehicles 2013 than they have delivered new cars. And these lists won’t contain “silent recalls”, repairs where manufacturers fix problems during a regular visit to the workshop without fanfare.
 

Complexity and time-to-market pressure as causes

But what are the causes of this error rate on the software level if carmakers invest millions in maintenance and quality assurance budgets? Firstly, there is the fundamental challenge that the software systems used in cars have become highly complex. To illustrate: The moon capsule from 1969 still came out with tens of thousands lines of code. A modern car now has around 20 to 100 million lines. So there’s more software code in an Audi A8 than in the Apollo 11. No expert is able to overlook this enormous complexity any more. Add to that the increasing technical complexity of the vehicles themselves, which indeed offers better features, but also increases the risk of errors.

Secondly, manufacturers now follow modular strategies to manufacture different car models with the same components in order to be more flexible, achieve synergies and allow new models to be created in ever shorter intervals and at lower cost. This way, software is used in different models as well. Such a common-part approach is supposed to realize significant savings and make the risk of callbacks controllable, because theoretically the lower amount of car parts can be better controlled. Rather, some of the most recent examples show, however, that expensive recalls can quickly melt the savings of modular strategies, even if the number of callbacks did not grow, but rather the number of vehicles affected.

In sum, software therefore represents a complex and thus difficult to view product to the automotive industry that is used in more and more contexts and must be market ready more quickly than in the past. Considering that advantages in global competition are often achieved by increasing development speed, shifting development processes to suppliers and increasing cost pressure, consequently, the risk of errors and quality fallbacks increase.
 

Avoid million damages caused by software with preventive quality assurance

After all, automakers try to produce their products even faster, more versatile and more cost-effective – development goals, that are not always beneficial to the quality of cars. Mass recalls remain inevitable if something goes wrong. Because the more often a defective part or software is installed, the higher the number of recalled cars. Each million recall arises an immense damage. The auto analyst Koji Endo calculated that a Toyota recall of nearly 6.6 million vehicles, would cost the company at least $600 million plus the often unpredictable damage to its image. If we consider costs of roughly one hundred US dollars per recalled vehicle, it soon becomes clear why automobile companies already put so much money in quality assurance. Any software problem that gets identified before a model leaves the production line saves pure money.

If the number of software-related recalls shall not continue to rise, it is important that carmakers put an even greater emphasis on quality assurance. Namely, on a quality assurance that is not merely designed reactive, but is much more preventively integrated into the development process in order to identify software risks before they escalate to a software error. This is also concluded by Stefan Bratzel, Director of the Center of Automotive Management in Bergisch Gladbach, who told the Spiegel that “the momentum of sales success and higher targets require strong quality management systems and accordingly strong personalities to carry responsibility for it.”

In other words, the maintenance and quality assurance of software should be a topic at highest level. The decision-levels in the automotive industry require representatives with dedicated software understanding and company-wide the creation of quality software solutions should be incentivized accordingly. Besides, creating an early warning system for software risks should be a priority within development. Such a system makes the visualization of technical risks possible – for example, based on Automated data-driven Software Management – and can identify software vulnerabilities in a semi-automated manner and help solving these. In this way, errors and vulnerabilities can be identified and eliminated before they occur, so that expensive and complex product recalls are needed far less frequently.

Pin It on Pinterest