refinedTS: Refined Time Synchronization for Cross-Domain CAN-TSN Communication

The Controller Area Network and Time-Sensitive Networking (CAN-TSN) network integrates low-speed event-triggered CAN and high-speed time-triggered TSN, meeting the safety requirements within the intra-domain networks and enabling high-speed transmission across the inter-domain communication. However, there is a large time offset (i.e., deviation) of several milliseconds between Electronic Control Units (ECUs) in CAN-TSN networks through our practical measurement using the Spirent C50 tester. The aforementioned phenomenon is due to the asymmetric transmission delay caused by the difference in communication rates between cross-domain CAN-TSN communication. It is urgent to synchronize time for cross-domain communication to minimize this time offset and meet the automotive real-time communication requirement. Unfortunately, mainstream time synchronization techniques exhibit specific limitations in cross-domain communication. CAN Time Synchronization (CanTSyn) is not compatible with the CAN-TSN networks, because CAN and TSN have different communication protocol stacks. The generic precision time protocol (gPTP) needs to be used multiple times (at least twice) to achieve cross-domain time synchronization, which will result in a cumulative time offset exceeding <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$10~\mu $</tex-math> </inline-formula>s. In this work, we propose a cross-domain refined Time Synchronization (refinedTS) for synchronizing CAN-TSN networks. refinedTS refines the number of packet exchanges from 8 (i.e., twice using gPTP) to 6 (i.e., once using refinedTS) to reduce the cumulative time offset in cross-domain communication. refinedTS further refines the definition of delay into transmission delay and dwell time, using the dwell time as compensation to obtain a more accurate time offset. Automotive-grade experiments of refinedTS show that: 1) the offsets are always within <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$1~\mu $</tex-math> </inline-formula>s in different scenarios; 2) the offsets are less than that of CanTSyn and gPTP; and 3) the incurs computation and memory overhead are 1.3% <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$\sim ~6.6$</tex-math> </inline-formula>% and 0.1% <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$\sim ~0.5$</tex-math> </inline-formula>%, respectively.