At the 2026 AI-Defined Vehicle Forum held on September 17, TrustMotion's chief system architect Fu Shen warned that simply moving software to a central computing platform does not solve the underlying challenges of software-defined vehicles (SDVs). "Centralizing software doesn't mean the problem is solved; many real challenges are just beginning," he said, according to D1EV.
From domain controllers to central computing
Fu explained that in traditional domain-controlled architectures, software and hardware remain tightly coupled. When automakers maintain multiple vehicle platforms, switching between architectures often requires redeploying software functions and reintegrating hardware and software, which "hasn't truly solved the problems SDV aims to address."
A more thorough approach, he argued, is to concentrate more software and control logic onto a central computing platform while making zone controllers "thinner," limiting them to relatively stable functions such as I/O expansion, regional gateway, and power distribution. This not only improves hardware and software reuse but also allows over-the-air (OTA) updates to focus more on the central platform.
However, centralization brings new systemic issues. When control logic once distributed across multiple ECUs is consolidated, the control link between the central controller and actuators becomes longer, making it harder to maintain millisecond-level real-time performance. Moreover, when hundreds or even thousands of applications run on the same computing platform, prioritizing tasks, avoiding resource contention, and ensuring that software of different safety levels do not interfere with each other can no longer be solved by a single chip or operating system.
Predictability as the key
"What really needs to be solved is whether the temporal behavior of all tasks can be predicted," Fu said. Instead of optimizing individual tasks or links late in development, he advocated considering task latency, resource requirements, and end-to-end links during the system design phase, using system-level scheduling to determine whether the architecture can meet requirements.
TrustMotion, formerly TTTech Auto, was renamed on April 30, 2026, after being acquired by NXP in June 2025. The company is pursuing this technical path with its MotionWise platform, which uses time-aware and time-triggered mechanisms to plan and schedule tasks uniformly at the design stage, improving determinism in complex central computing architectures. According to public information, MotionWise can handle time-triggered, event-driven, and data-driven tasks, and covers multi-SoC, multi-core, and distributed computing environments through global scheduling.
Fu believes that as automotive architectures shift from distributed to centralized, the competitive focus is also changing: "Putting software into a central computing platform is only the first step; how to make that software run stably and predictably is the real problem centralization needs to solve."
