From Distributed ECUs to Central Computing and Zonal Architecture

Introduction#

The automotive industry is undergoing a paradigm shift in electrical/electronic (E/E) architecture, moving from distributed electronic control units (ECUs) to central computing with zonal control. This evolution is driven by the need for increased computing power, reduced wiring complexity, and support for advanced features such as autonomous driving and connected services. Distributed ECUs, where each function is handled by a dedicated ECU, have become inadequate due to scalability and weight constraints. The zonal architecture, which groups ECUs by physical location and connects them to a central computer, offers a more efficient and flexible solution.

This guide explores the underlying logic of this transition, focusing on cross-domain fusion, automotive Ethernet topology design, and the adaptation challenges for AUTOSAR Classic Platform (CP) and Adaptive Platform (AP). By understanding these aspects, architects can better navigate the complexities of modern vehicle E/E architectures.

Cross-Domain Fusion and Computing Demands#

Cross-domain fusion, such as merging advanced driver assistance systems (ADAS) and infotainment, is a key driver for central computing. This fusion enables resource sharing, reduces hardware redundancy, and facilitates new user experiences. However, it imposes significant computing demands. For instance, ADAS requires high-performance processing for sensor fusion and AI algorithms, while infotainment demands rich graphics and connectivity. A central computer must deliver scalable compute, often through heterogeneous SoCs with CPUs, GPUs, and NPUs.

To meet these demands, architects must consider power consumption, thermal management, and safety-criticality. Techniques such as hardware virtualization and quality-of-service (QoS) mechanisms ensure that safety-critical functions are not compromised by less critical ones. Additionally, the software architecture must support mixed-criticality execution, which is where AUTOSAR AP and CP come into play.

Automotive Ethernet Communication Topology Design#

With the shift to central computing, in-vehicle networking must evolve to handle high bandwidth and low latency. Automotive Ethernet has emerged as the backbone, replacing legacy buses like CAN and FlexRay. A typical topology uses a central switch connecting multiple zone controllers via multi-gigabit Ethernet links. Each zone controller aggregates data from local sensors and actuators, reducing wiring harness length and weight.

Designing an effective Ethernet topology involves considerations of bandwidth allocation, redundancy, and time-sensitive networking (TSN). TSN standards ensure deterministic communication for safety-critical data, while redundancy mechanisms like IEEE 802.1CB provide fault tolerance. Architects must also plan for future scalability, as data rates are expected to increase with higher levels of automation.

AUTOSAR CP/AP Software Stack Adaptation Challenges#

AUTOSAR Classic Platform (CP) is designed for hard real-time, safety-critical ECUs, while Adaptive Platform (AP) targets high-performance computing with dynamic software updates. In a central computing architecture, both platforms must coexist and interact. CP handles real-time control tasks, such as braking and steering, while AP manages compute-intensive functions like sensor fusion and over-the-air updates.

The main challenge is integrating CP and AP on the same hardware or within the same network. This requires a hypervisor to isolate the two environments and a reliable inter-process communication (IPC) mechanism. Additionally, the software architecture must support service-oriented communication (SOC) for AP and signal-based communication for CP. Tools and methodologies must be updated to handle the complexity of mixed-criticality systems, ensuring safety and security across domains.

Conclusion#

The transition from distributed ECUs to central computing with zonal control is essential for future vehicles. It addresses the limitations of current architectures by providing scalable compute, simplified wiring, and support for cross-domain fusion. However, it also introduces challenges in computing demands, network design, and software integration. By leveraging automotive Ethernet with TSN and adapting AUTOSAR CP and AP through virtualization and careful partitioning, architects can build robust and future-proof E/E architectures. As the industry continues to evolve, collaboration and standardization will be key to overcoming these challenges.