Automated Scheduling for Optimal Parallelization to Reduce the Duration of Vehicle Software Updates

Abstract

The needs-oriented expansion and maintenance of vehicle functions through software updates will increase drastically due to current megatrends such as autonomous driving or connected car services. Both, shorter innovation cycles and increasing software volumes resulting from these megatrends, increase update frequency and duration equally. Consequently, a reduction of update duration will become even more important in the future. Popular literature has shown that in a vehicle-internal network of up to 100 electronic control units (ECUs), the parallelization of individual ECU updates has great potential to reduce the total update duration. However, the methods presented so far do not show an optimal parallel scheduling algorithm that uses the full potential of the vehicle-internal network. In addition, some of the approaches are also limited to certain bus systems or do not have an automated parallelization strategy based on ECU requirements. The approach introduced here uses a high abstraction model for optimal parallel ECU software update scheduling. For this purpose, we developed two scheduling algorithms, one as a mixed-integer linear program and the other as a hybrid algorithm. We verified our approach in a real testing environment with up to 20 ECUs regarding update duration and computation time. Evaluation shows that the reprogramming time can be reduced by up to 77% compared to sequential update processing through optimal parallelization. Furthermore, the model is very easy to adapt to different and new vehicle generations due to its high abstraction.