• Next-gen SDV „hyperconverged“ architectures are significantly increasing the security & safety challenges
• A tour and benchmark of existing market solutions and the new risks
• Insights into solutions for achieving hyperconvergence, safely-

The next trend of SDVs is hyper-convergence, or centralization, of the E/E architecture. This means merging several functions of the vehicle into single ECUs, or even single system-on-chips. Those approaches seem to bring many benefits, from having a cloud replica of the car, being HW-agnostic, to wiring efficiency and time-to-market, but this also comes with a lot of security and safety challenges. This presentation will dig into those specific challenges, and compare different solutions that will be available in the market in the next decade, highlighting their weaknesses and advantages. Finally, let’s see what technological alignments areneeded in the ecosystem for those solutions to be beneficial to everyone, without compromising safety of road users.

• Updated view on technology evolution for the software-defined vehicle
• Upcoming new requirements: x-by-wire, domain fusion designs, native ADAS L3 architectures

The world of E/E architectures is once more at a crossroads. With an increasing demand to push software-driven features into the volume segment, and a fierce price competition it is up to the E/E architecture to help strike the balance between cost-down and innovativeness. In this talk we present a revised, technology-driven view on technology and refactoring needs for emerging and future vehicle architectures.

• Short life cycle consumer techology meets long-lasting embedded computer systems
• How do we desing hardware, software and services that are modern and future-oriented?

• Advanced Automated Driving Functions are adding customer value and provide a premium individual mobility experience
• AI is a key enabling technology to realize ground-breaking automated driving functions
• AI poses challenges in safety-relevant functions – and a major leap in demand for elaborate embedded compute power

Higher levels of automation in road transport mobility are a true moonshot. Very few OEMs have so far successfully obtained road approval for their Level 3 automated driving system, BMW being one of them. AI already now is a major technological driving force behind automated driving and will continue to be. We will highlight current frontiers of AI development in safety relevant systems and explore, how expected evolution of the technology will influence the road ahead for automated driving functions. Growth in sensor data volume and machine learning compute demand is fuelling a race to ever more potent embedded systems. Just brute-forcing hardware will not do.

• Evolvement of SOAFEE.next for the AI-enabled SDV
• Virtual platform reference support
• Foundational heterogenous compute platform including path to Automotive chiplets

In this presentation, we explore critical advancements in automotive technology, focusing on the transition to software-defined vehicles (SDVs). Our discussion covers three key areas:

1. Evolvement of SOAFEE.next for AI-enabled SDV: We delve into the evolution of the SOAFEE initiative, emphasizing its role in enabling AI-driven capabilities within SDVs.
2. Virtual Platform Reference Support: We highlight the importance of robust virtual platforms as a foundation for early development and testing to support shift-left software development on latest Arm Automotive Enhanced technology.
3. Foundational Heterogeneous Compute Platform and Automotive Chiplets: We outline our vision for a foundational heterogeneous compute platform, paving the way for efficient, scalable, and customizable automotive chiplets. These chiplets empower next-generation vehicles with enhanced performance and flexibility.

• Overview of aiMotive products and market positioning
• How can an ADAS SW stack be ported to different embedded automotive HW compute platforms
• Why an accelerated functional model of the NPU is needed 
• How can the cloud be used for CI/CD and how can this cloud-based testing be used to validated software functional performance with different embedded NPUs

Development of automated driving software requires testing on embedded hardware platforms using large amounts of test data. Furthermore, multiple hardware platforms may need to be targeted simultaneously having different neural processing units (NPU) and other features. An accelerated testing flow is required that allows developers to test features on multiple platforms while supporting CI/CD. We present a multi-stage testing flow that uses software-in-loop (SiL) testing and takes advantage of massive cloud-based acceleration resources running multiple virtual platforms with large amount of test data. We show that accelerated functional models of the NPU are invaluable for enabling SiL and reducing costs.