Automotive Industry Insights

How does automated driving change the electrical/electronic (E/E) development process?

The standard approach to electrical/electronic development works with automated driving – but only to a limited extent, as the event chains of the driving function are clearly becoming more complex and contain elaborate environmental sensors. Newly established organizations are function-oriented, rather than component-oriented. That means that responsibilities shift, and not just within individual organizations. It also changes the relationship between manufacturers and suppliers, which then, of course, changes development processes and test strategies. At the same time, priorities are shifting as well. Where software algorithms and electronics were previously the sole focus, now AI comes into play. That results in changed process models and the emergence of new roles in data processing. Data train AI, and data are crucial for verifying and validating AI. That requires intensive investments in AI infrastructure in order to develop the right solutions and to support the associated data processes. This encompasses the processing and handling of data, from the acquisition of data logging in vehicles to data enrichment and big data management. Due to the high complexity of systems and their environments the importance of data reaches a new level.

Established methods of model development reach their limits when they encounter data-driven development: AI-learned behavior cannot be modeled directly. A large amount of information must be made available in a systematic way before a sufficiently accurate model can be synthesized.

Thus, AI know-how must be connected with the know-how of electrical/electronic development in the automotive sector. While the introduction of AI is accepted, it cannot lead to weakened competence in development or testing of automotive industry. This is why data-driven development requires a network of partners with an overall view of the process landscape – a partnership that acts independently of domain and thus with a focus on overall development to support the advancement of holistic approaches.

Where previously closed-loop control (prototyping) and validation (closed-loop XIL) were used, data-driven development is enforcing a paradigm shift: data are also being used to stimulate interfaces and subsequent evaluation. This calls for test systems that reproduce recorded and enriched data from test drives or simulations to validate AI algorithms and AI-based ECUs, purely simulatively, based on software or AI algorithms, or directly with the ECU hardware under strict, real-time requirements. Ultimately, testing functional and structural requirements using synthesized and enriched data leads to new challenges in validation. New technologies and methods must be integrated into a holistic process to make this data processing as safe as possible and efficiently connected to all the integration steps along the electrical/electronic development process.

This covers everything from data process to AI training to overall integration and is based on continuously repeated processes that achieve system maturity with increasing amounts of data. Providers of development solutions must, therefore, be involved in the development process much earlier. They will take on the role of development partner to integrate new methods into established processes and to define testing goals and achieve them as efficiently as possible.