Automation and auto industry: what level of errors are acceptable?

Two reported cases of GPS malfunctioning in recent weeks rocked the automotive world, raising some sharp questions about the safety and dependability of connected technology. 

On October 1, 2023,  two Kerala-based doctors, Dr. Advaith (29) and Dr. Ajmal Asif (29), along with three other individuals, embarked on a fateful trip in a Honda Civic. The guidance system is alleged to have mistakenly identified a river as a flooded road, leading the unsuspecting travellers astray. Oblivious to the impending danger, the car pressed on, following the straight path indicated on the map. However, tragedy struck as the vehicle gradually succumbed to the river's depths. While three of the occupants managed to escape the sinking car, Dr. Advaith, who was behind the wheel, and Dr. Asif, another passenger, lost their lives. 

The development came just days after a North Carolina family in late September took legal action against technology giant Google. The lawsuit stemmed from the untimely death of Philip Paxson, a medical device salesman and father of two. On September 30, 2022, Paxson's ill-fated journey took him to Snow Creek in Hickory, where his Jeep Gladiator met a watery demise. The lawsuit alleges that Google was fully aware of a collapsed bridge along Paxson's intended route but failed to update its navigation system accordingly. 

When asked to comment on the GPS malfunctioning cases, Jean-François Tarabbia, the Head of Architecture and Networking at Continental AG, responded to Autocar Professional, saying that while he remains unaware of the specific cases mentioned, he highlighted two critical issues that arise in such situations.

The first issue, according to him stems from the incorrect assumption that GPS values are always accurate. He further explained that software failures are typically analysed using a fault tree analysis method to understand the consequences of incorrect values. To ensure reliability, GPS information is usually cross-validated with other data sources and assessed for plausibility. Tarabbia acknowledged the complexity of the matter, stating, "Nevertheless, it's a complex issue because there is a lot of differential information that we combine."

The second issue pertains to the use of artificial intelligence (AI), where the learning process must be appropriately calibrated and risks must be mitigated. Tarabbia expressed confidence in the current state of fusion technology, proclaiming it to be "extremely, extremely good." He asserted that the remaining cases of errors are minimal when compared to the mistakes made by human drivers. To support his argument, Tarabbia referenced tech billionaire Elon Musk, who successfully demonstrated the efficacy of Tesla's autopilot system by meticulously recording all the miles driven using it. By comparing accident rates with and without autopilot, Musk's data showed a significant reduction in accidents when using the automation feature compared to human-driven vehicles.

Tarabbia then delved into the philosophical question asking, "What level of errors from machines can we accept?" It prompts a deeper reflection on the tolerance society should have regarding mistakes made by automated systems.

Increasing India-focused products

Discussing the growing prominence of India's automotive industry in software-driven vehicles, Tarabbia highlighted Continental's integration of its design centre in India into its global network, which has enabled valuable insights into customer needs and emerging technologies across diverse markets. Continental, he claims has formed a dedicated team in India to address local cost constraints and develop suitable technologies, collaborating closely with major OEMs to create practical solutions that balance performance and cost. 

Addressing  the challenges faced by the automotive industry in software-defined vehicles, he noted that the need for increased standardistion of middleware is emerging  as a solution to seamlessly integrate software across different hardware platforms while reducing costs associated with middleware development.

Additionally, Tarabbia emphasised the importance of managing cybersecurity risks as connected vehicles become more prevalent. Optimising the overall vehicle architecture is becoming  crucial to efficiently utilize resources and enhance performance and functionality.

Lastly, Tarabbia highlighted the significance of the validation process, which ensures the proper functioning and compatibility of integrated functions from multiple suppliers, providing assurance and reliability.