As vehicle intelligence continues to scale, one thing is becoming increasingly clear: the traditional ECU-centric architecture is no longer fit for purpose. At AutoSens USA 2026, a series of sessions will explore this shift in depth, examining how OEMs and suppliers are rethinking vehicle architectures to support growing sensor demands, real-time perception, and increasingly software-driven functionality, without compromising on safety or cost.
Written by:
Jenny Campbell B. Rust
Head of Content
The Great ECU Collapse: Why Centralisation Is Gaining Ground
Centralisation is not emerging as a trend by choice, but by necessity. As sensor counts increase, data rates climb, and perception stacks become more complex, traditional distributed ECU architectures are being pushed beyond their limits. What once offered modularity now introduces latency, integration overhead, and inefficiencies that are difficult to sustain at scale.
At its core, the shift toward centralised and zonal architectures is about managing complexity at a system level. Consolidating sensing and compute reduces data movement, simplifies software integration, and enables more deterministic performance, which is critical as vehicles move toward real-time, AI-driven perception. It also supports more efficient update pathways and long-term scalability, aligning with the demands of software-defined vehicles.
There are also clear economic and lifecycle drivers. Fewer ECUs mean reduced hardware costs, simplified wiring, and more manageable update pathways. Centralised architectures are inherently better suited to over-the-air updates and continuous feature deployment, while also creating a foundation for future scalability through software-led upgrades rather than wholesale hardware redesign.
However, this shift introduces new challenges. Consolidating compute raises important questions around redundancy, fault isolation, and safety certification. A failure in a centralised system has broader implications, requiring more sophisticated approaches to fail-operational design and system-level validation. This is also driving an increased focus on predictive maintenance and real-time system monitoring, where link integrity and component health are continuously assessed to detect degradation before it leads to system-level failure. At the same time, bandwidth constraints and thermal management become more pronounced as more functionality is concentrated into fewer nodes.
This tension between simplification and risk concentration sits at the heart of the architectural transition. Centralisation offers a clear path to scalability and efficiency, but only if it can be implemented without compromising safety or resilience.
These trade-offs will be explored in “Architecture Shifts: Collapsing ECUs for Scalable Sensing & Compute,” with perspectives from Stellantis, Bosch, NXP, and Analog Devices.
The Minimalist Vehicle: Doing More with Less
Closely linked to architectural consolidation is the theme of the “Minimalist Vehicle”, a concept that will explore how to reduce latency, compute load, and overall system cost without sacrificing performance.
Rather than simply adding more hardware to meet new demands, the sessions on this topic will examine how intelligent system design can unlock efficiency gains. This includes rethinking where compute should sit within the vehicle, reducing unnecessary data movement, and adopting more event-driven approaches that activate processing only when needed.
Always-On Awareness: Rethinking Surveillance
Another emerging theme across the agenda is the growing role of surveillance within vehicle architectures, particularly as sensing capabilities extend beyond active driving scenarios into always-on awareness. Contributions from Lattice Semiconductor, OMNIVISION, and Analog Devices highlight how vehicles are increasingly expected to observe, interpret, and respond to their environment even when not in motion.
This introduces a new layer of architectural and ethical complexity. Systems must balance low-power, event-triggered operation with continuous readiness, while also ensuring data integrity and system health over time. At the same time, the shift toward persistent sensing raises important questions around data ownership, privacy, and the boundaries between safety, security, and surveillance. As vehicles become more intelligent and connected, these capabilities will not sit in isolation, but as part of a broader architectural shift toward vehicles that are not only reactive, but observant by design.
Networking, Data, and the Multi-Gigabit Challenge
As architectures evolve, so too must the underlying vehicle networks. Sessions across the agenda, including a contribution from General Motors, will explore how technologies like 10BASE-T1S are being optimised within software-defined architectures, reflecting the growing importance of low-latency, scalable communication across increasingly distributed systems.
With zonal architectures distributing sensors and actuators across the vehicle, networking becomes a critical enabler of real-time intelligence. It is no longer just about moving data, but about doing so deterministically, reliably, and at scale.
At the same time, the push toward multi-gigabit data rates introduce new challenges at the physical layer. Sessions covering automotive Ethernet, MIPI A-PHY, ASA Motion Link, and GMSL will examine the realities of signal integrity, conformance, and validation, areas that are becoming increasingly complex, yet foundational to reliable perception and sensor fusion. Taken together, these networking and data challenges point to a broader architectural question, one that extends beyond connectivity alone. The fundamental question of whether these systems are truly future-proof begins to loom large, as a legacy-heavy industry, shaped by complex supply chains and long-term development cycles, works to keep pace with rapid advancements in AI.
Software-Defined… for How Long?
Much of the industry has aligned around the concept of the SDV, but a growing question sits beneath the surface: how long will this paradigm remain the end goal? Designing platforms that can handle continuous over-the-air updates, modular software stacks, and increasing system interdependencies is already a significant challenge. But at the same time, some market disrupters, agile, full-stack integrated players such as BYD and NIO are beginning to move beyond SDV toward AI-defined vehicles, where behaviour is increasingly shaped by data-driven models rather than explicitly coded logic.
This creates a tension across the industry. Traditional automotive manufacturers, often constrained by legacy platforms and organisational structures, are working to establish robust SDV foundations. Meanwhile, newer entrants are iterating faster, integrating hardware and software more tightly, and redefining what “vehicle intelligence” looks like in practice.
AutoSens USA 2026 will provide a space to explore this inflection point. Is SDV a destination, or simply a stepping stone?
From Architecture to Advantage
Throughout my research, a clear narrative has emerged, helping me shape the overarching direction of the agenda: architecture is no longer just a technical consideration, it is a strategic differentiator.
The sessions at AutoSens USA 2026 will explore how decisions made today around compute placement, network design, and safety integration will directly shape vehicle performance, cost structures, and the ability to deliver new functionality over time. More than that, they will determine how well companies can adapt to what comes next, whether that is the continued evolution of the software-defined vehicle, or a faster-than-expected shift toward AI-defined systems.
The move away from distributed ECUs is already underway. The bigger question now is what replaces them, and how future-ready those decisions will prove to be.
Interested in In-Cabin monitoring technology?
With a pass to AutoSens USA, you’ll also get full access to our co-located sister event, InCabin. Take a look at the full agenda for InCabin here >>
Explore more on the AutoSens YouTube Channel⬇
