Short Term Pain and Long-Term Gains of Ethernet-Based E/E Architectures Adoption in Automotive Sector
Session 1 – Architecture & Networks
February 9, 16:05 p.m. - 16:35 p.m. CET/7:05 a.m. - 7:35 a.m. PT
Short Term Pain: Many of today’s E/E architectures are domains and are moving towards Central Compute platform as well as Zonal edge nodes. This is a big change to E/E Architecture as OEMs are organized primarily into following domains: Powertrain, Chassis, Body, In-Vehicle Infotainment Domain and ADAS. Changing the organization is more a political challenge rather than a technical one. OEMs will feel the short-term pain during the transition as they need to decide what functions are to be performed by all the domains and how much can be moved to a Central Compute platform as well as what is moved or left at the Zonal edge nodes. Deciding what used to be performed by dedicated ECUs as part of the domain, would now be considered as a software functions (features), that can reside as software features within the Central
Compute platform or part in some instances within the zonal edge node.
Long Term Gains: Instead of large central gateways where maintaining timings becomes exponentially hard to do, moving away from large gateways to mini gateways in zonal areas will reduce the complexity of gateways. Within the central computer a platform-based software structure can be created where OEMs can have the flexibility to add and remove software functions. Functionality in compute platforms primarily with Ethernet links, is also important due to the ability to get access to all sensors attached to zonal edge nodes. Ethernet network protocol provides the scalability as bandwidth demands grow. Ethernet is a true scalable network protocol with the added advantage to utilize many of the standard security technology that has been adopted in many industrial sectors today. Creating boundaries within the compute platform while isolating domain features using hypervisors is an added benefit.
Enhancing High-speed Ethernet Transceiver Electromagnetic Immunity Performance with Artificial Intelligence
Dance Wu, Cliff Fung
Session 3 – Safety, Security, and Robustness
February 10, 17:40 p.m. - 18:10 p.m. CET/8:40 a.m. – 9:10 a.m. PT
In the automotive industry, electromagnetic compatibility (EMC) is one of the essential pillars to ensure safety during operation. As Ethernet continues to integrate into the automotive design, it serves as the backbone of critical application. As such, it is necessary for the Ethernet device to maintain data links and transmit all Ethernet frames without any frame drop nor corruption as a result of electromagnetic noise in order to ensure that critical applications maintain uninterrupted operation.
In non-automotive Ethernet, electromagnetic immunity requirement is usually more lenient as there are retransmission protocols in place. However, in automotive Ethernet, such retransmission is not possible due to requirements for low latency. Therefore, Ethernet transceivers need additional design for more resilient operation.
In electromagnetic immunity (EMI), with regard to the act of subjecting system to electromagnetic noise, noise can be partitioned into two type: periodic noise and transient noise. Due to the nature of periodic noise, it is easier for system to detect and to adapt to using existing DSP and Analog design technique. Meanwhile, transient noise is rapid, high amplitude noise, and only occurs for short time duration, making it harder for existing designs to handle.
As automotive Ethernet is accelerating into multi-gig, this transient noise will further infringe into the bandwidth of the transmitted signal. It is more important than ever to develop means of detecting and mitigating such transient noise to further protect the communication channel and improve system safety.
Because of the unpredictability of transient noise occurrence, we need to first be able to predict the existence of such noise. For this purpose, artificial intelligence is an optimal choice to serve as the backbone in the detection and mitigation system. In this presentation, we will discuss how we can leverage its capability to inference current channel condition and apply best mitigation practice through reinforcement learning to enhance electromagnetic immunity performance.