veoPipe - Reliable AI in the product development process

In Detail

AI functions are increasingly being integrated into physical products. The AI functions themselves and the resulting products are insufficiently analysed and secured by today's standards, and are difficult to explain and verify (black box issue). However, the analysis and validation, including the explainability and comprehensibility of the product and its functions, are crucial to generating trust and creating broad acceptance of products with AI. This requires an adapted approach in development.

Based on two example products, an adapted product development process will be developed: A system for classifying the condition and safeguarding the performance of a battery, as well as an e-trailer platform including reliable estimation of the condition of the trailer.

For the safeguarding of the physical components and their interfaces with the AI functions, classic quality management methods will be examined and further developed with regard to their suitability and applicability to physical products with AI functions. For the AI itself, methods of AI explainability (xAI) are used in combination with a formal or data-based verification of the system behaviour in order to be able to identify, evaluate and optimise critical system behaviour in a targeted manner.

Integrated into a holistic approach, this results in a stringent product development process. This supports the efficient development of safe products with AI functionalities and strengthens confidence in these products.

Sponsored by

the Federal Ministry for Economic Affairs and Climate Action

and

TÜV Rheinland AG