The next generation of automotive solutions will feature highly autonomous driving and rich infotainment that will demand significant connectivity both externally and internally to the vehicle. Especially, the increasingly high level of automation of current SAE L3-4 prototypes requires unprecedented computing power, at reduced Size, Weight and Power (SWaP), served under tight real-time guarantees. Modern high-performance embedded ECUs, such as the ones powered by NVIDIA or Xilinx accelerators, can jointly work with off-vehicle cloud servers to serve this purpose, however, the key point of predictable data transmission among them is still an open research area. Networking will serve both external edge and cloud communications, as well as internal communication between the various domain or zonal ECUs articulating the architecture. TSN is a key technology enabling networking capabilities for high bandwidth and low latency communication in a mixed criticality environment.
In this presentation, we are going to introduce Time Sensitive Autonomous Architectures (TSAA), a research project for TSN-enabled ADAS architectures. We will present its latest and first iteration, featuring an autonomous luxury sports car controlled by a dual zone ECU architecture built with Xilinx Zynq Ultrascale+ and SoC-e Multiport TSN switches. The software architecture is based on a hypervisor providing strong isolation and virtual access to TSN for the virtual machines, including Linux, Erika Enterprise RTOS and bare-metal environments. In this talk, we will present the key technological innovations of TSAA. Moreover, we will discuss the engineering challenges and the performance evaluation of the TSAA project demonstrator.
TSAA is an initiative driven by the University of Modena and Reggio Emilia, together with its spinoff startup company, Minerva Systems. TSAA is partially supported by the Italian research project SPHERE, funded by the PRIN framework of the Italian Ministry of Research and Education, and is sponsored by SoC-e.