Infineon sees itself on the safe side with its Aurix safety microcontroller family

March 16, 2017 // By Christoph Hammerschmidt
In the ongoing discussion about the right computing architecture for self-driving vehicles, functional safety is emerging as the paramount aspect. For this reason, Infineon regards itself as one of the winners of the tech trend towards the robot car. At Embedded World in Nuremberg, Infineon’s Senior Director Microcontrollers, Thomas Boehm explained the company’s strategy for its Aurix family of microcontrollers.

Currently, the second generation of the Aurix family is in production. Aurix microcontrollers sales so far have crossed the 3 million unit line and is used in advanced series vehicles like BMW’s 7 series and Daimler’s Mercedes E-class since 2016. In both cases the Aurix is involved in data fusion, one of the most demanding computing tasks in Advanced Driver Assistance Systems and self-driving cars for the high safety requirements to these algorithms. In such environments, the computer handling the driving decision must be designed such that it even can survive a power failure, explained Boehm. The Aurix microcontrollers therefore are typically located in ECUs that control power train, chassis and safety applications with time-critical requirements.

Currently, automotive OEMs are mulling over the best layout and partition the real-time date associated to autonomous driving. Will the computers in future car generations be more centralized than today’s far-flung electronics landscape with dozens of independent ECUs? Probably yes, Boehm says. However, the brain in this central computer won’t be implemented in a single microprocessor, not even one with many heterogeneous cores, Boehm believes. Instead, Aurix microcontrollers will be present even in the next generation of car computers –“and they will continue to be responsible for safety topics – not as a piece of IP in a highly integrated SoC but as a separate microcontroller,” Boehm predicts.

Beyond the question about centralizing or decentralizing the computing resources in future vehicles, there are a number of challenges for ECU designers: The high amount of data to be shuffled around between sensors, actuators, antennas and subsystems need to be handled. Therefore, switched Ethernet will be established as data highway in the cars. Software will be another challenge – the complex tasks and the degree at which tasks will be implemented in software instead of hardware will require the reuse of legacy code as well as safety concepts in software testing. The transition from fail-safe to