Edge Intelligence Redefines What SDVs Can Do

Software-defined vehicles (SDVs) are increasing the need to process intelligence closer to the systems that are generating data. Modern vehicles produce large amounts of data from sensors, control units, and diagnostic systems, making it inefficient and costly to send all the raw data to the cloud for processing.

This is one of the main reasons why edge AI is becoming more relevant in SDVs. Rather than transmitting data to the cloud for analysis, the vehicle can preprocess data locally. This reduces the bandwidth, storage, and cost required to extract insight. Local execution also supports privacy because raw data can remain inside the vehicle.

Latency is another key driver. Some vehicle functions and diagnostics benefit from faster local analysis-for example, when the system needs to detect abnormal behavior, trigger targeted data collection, or support predictive maintenance.

John Heinlein, chief marketing officer at Sonatus Inc., also points to algorithmic privacy as an advantage. "Encapsulating models in the vehicle may help suppliers protect their software know-how while still allowing their algorithms to operate on vehicle data," he said.

The move to edge AI also depends on the vehicle's electrical and electronic (E/E) architecture.

Centralized and zonal E/E architecture for edge AI

The legacy automotive architectures are characterized by a highly distributed network of isolated, application-specific electronic control units (ECUs). These ECUs are constrained by fixed hardware, coupled software, and slow, rigid communication protocols with no computational headroom for machine-learning inference.

To overcome these limitations, automotive manufacturers move to a domain, centralized, and zonal architecture. In a domain-based topology, related vehicle functions, such as powertrain, chassis, body, and infotainment, are grouped and managed by domain controllers.

Zonal architectures further abstract the hardware by organizing the vehicle geographically. Sensors and actuators are connected to local zonal controllers based on their physical location.

"Hardware centralization is the first step; the software-defined vehicle is realized when AI can run at the edge and when hardware continuously learns and adapts," Jeff Chou, CEO and co-founder of Sonatus, said in a statement.

This is important for edge AI because AI models need more than a processing unit; they need access to relevant vehicle signals, a place to execute safely, a way to manage resource limits, and a lifecycle path for updates after production. Centralized and zonal architectures make it easier by consolidating compute and creating more consistent access to vehicle data.

Centralization by itself does not automatically create edge AI. The more important shift is that newer E/E architectures give OEMs a better environment for reusable software layers.

Instead of integrating every analytics model separately with each ECU and vehicle variant, automakers can start building a common infrastructure for data access, model execution, policy control, and update management.

Heinlein said centralized or zonal architecture is not "holy and perfect" by itself.

Intelligence beyond autonomous driving

The operational reality of the SDV extends beyond autonomous driving to encompass edge AI workloads. It includes diagnostics, analytics, and subsystem optimization.

According to the 2026 Omdia SDV survey, the industry is undergoing a significant recalibration to prioritize applications that deliver a measurable return on investment and immediate operational efficiency.

As vehicles become more complex, OEMs need ways to detect faults early, analyze subsystem behavior, improve software post-production, and reduce the time and cost of diagnostics. For example, models that run closer to vehicle systems can detect abnormal behavior, derive new signals from existing sensor data, trigger targeted data collection, support predictive maintenance, and help engineers understand how vehicles behave in real-world use.

According to the Omdia report, leading AI use cases include smart diagnostics and predictive maintenance (34%), driver-safety monitoring (26%), continuous software improvement (26%), and virtual sensors or derived signals (21%).

Heinlein explained that vehicles need better diagnostics by analyzing vehicle data in advance. Edge AI can identify what is happening in the vehicle and provide indications of problems before they become visible to the driver or service technician.

Heinlein described the vehicle as becoming a "black box" for many subsystem makers after production. If the OEM allows it, edge AI and targeted data collection can enable suppliers to monitor, tune, and optimize shipped subsystems, helping them understand how components perform under real-world operating conditions.

Partnership for motion control systems

Schaeffler AG and Sonatus recently announced a partnership that aims to bring edge AI into motion control solutions for SDVs.

Schaeffler provides cross-domain control units and system integration expertise across powertrain, energy, chassis, and body domains. Sonatus provides cloud-native middleware infrastructure, including data collection and model management tools.

The partnership integrates Sonatus Collector AI (for targeted data collection) and Sonatus AI Director (for executing and managing AI models) directly into Schaeffler's embedded control units. "Together with Schaeffler, we are turning static control units into dynamic, intelligence-driven systems," Chou said.

The partnership points to a model where AI-enabled analytics can operate closer to the vehicle systems that generate the most useful operational data. This is critical because motion control domains operate under changing loads, temperatures, road conditions, driver behaviors, and component aging profiles.

The joint solution can help automakers run and continuously improve functions such as steering, braking, and energy management directly on the control unit, with new features and optimizations deployed over the vehicle lifecycle without hardware changes.

Heinlein said that AI can operate alongside safety-critical systems, monitoring them and providing feedback for future upgrades, changes, or tuning without compromising the safety certification of the underlying control path.

"Our central control units are equipped with a pre-integrated software infrastructure that includes solutions such as Sonatus products," said Rodrigo Peres, SVP of business unit vehicle and battery controls at Schaeffler, in a statement. "This simplifies integration for OEMs and helps them accelerate the centralization of their software architecture."

A model that interacts with vehicle subsystems often requires custom integration for the vehicle and its signals, compute resources, and subsystem context, Heinlein said. Once the underlying data connectivity and execution management layer is in place, each new model can reuse it rather than requiring a fresh, custom integration.

The partnership shows that the value of edge AI in SDVs depends on whether OEMs can collect sensor data accurately, process it onboard, deploy models efficiently, and leverage vehicle behavior to improve systems over time.

The post Edge Intelligence Redefines What SDVs Can Do appeared first on Embedded.