Abstract: With the rapidly increasing number of sensors in today’s automotive designs, modern Advanced Driver Assistance Systems (ADAS) are highly dependent on high-performance multi-core computing platforms and high-speed Ethernet networking. In this paper, we first present an industrial-like interconnection architecture between ADAS clusters and computer modules running AI application software and highlight the unique challenges induced with the simultaneous operation of the Ethernet AVB time-triggered and standard traffic classes. We then demonstrate the advantages of creating a flexible AI development platform with the ability to run simulation programs that abstract the realistic interaction of the AI with the ADAS domain. We provide background information on contemporary automotive cluster controllers as well as the real-time data flow components of typical ADAS ECUs, in order to elucidate the associated complexities. Finally, we propose the adaptable Partitioned Gateway module architecture that integrates an automotive microcontroller with a high-performance Gb Ethernet switch to support various ECU clusters, while minimizing latency and network loading for ADAS sensor data preprocessing.
Keywords: ADAS (Advanced Driver Assistance Systems), Ethernet AVB (Audio Video Bridging), Multi-core computing platforms, AI application software, Interconnection architecture, Simulation programs, Automotive cluster controllers, Real-time data flow, Partitioned Gateway module, Gb Ethernet switch
|
DOI:
10.17148/IJARCCE.2023.12826
