NXP and Sonatus Accelerate In-Vehicle Edge AI Deployment

Automotive original equipment manufacturers (OEMs) and suppliers now have a path forward in realizing the full potential of in-vehicle AI. Sonatus AI Director  with integrated NXP eIQ® Auto ML software development environment and NXP’s S32 automotive processing platform comprises a comprehensive edge AI toolchain for local execution of AI workloads designed to overcome limitations in responsiveness, reliability and other performance challenges.

Despite the growing adoption of AI by automakers, innovation has largely been confined to cloud-based implementations, infotainment systems and autonomous driving. While these applications are well-suited to such environments, they do not provide the real-time, low-latency, reliable, safe and secure capabilities required to fully unlock the potential of AI for core vehicle functions. This gap presents a significant untapped opportunity for leveraging in-vehicle AI compute to help reduce latency for faster response times, lowering data transmission costs, protecting proprietary algorithms and intellectual property and enhancing privacy.

Now, NXP and Sonatus are teaming up  to bring OEMs and suppliers a unified platform that turbocharges the development of in-vehicle AI. Sonatus AI Director, with integrated NXP eIQ Auto ML software, utilizes NXP’s S32 automotive processing platform to deploy and execute edge AI models for a wide range of use cases. This combination of hardware and software delivers real-time decision making, adaptive behaviors and autonomy-enabling capabilities to the vehicle edge. Sonatus AI Director is opens a world of possibilities of AI features that can scale across automotive production models.

Accelerated, Scalable AI Deployment for Safety-Critical AI Applications at the Edge

By performing AI compute locally, developers can reduce latency and increase responsiveness, lower data transmission costs, and protect proprietary algorithms and IP.

In addition to NXP eIQ Auto ML software, Sonatus AI Director features S32 processing automotive-grade IP, including high-performance networking accelerators for a comprehensive, end-to-end toolchain to train, validate and deploy AI models by optimizing data flow from vehicle networks to the AI models. This support across the full AI life cycle provides a development environment where OEMs can develop and scale AI workloads across the NXP hardware platform.

For a better experience, download the diagram.

The NXP eIQ Auto ML software development environment provides a uniform, consistent workflow for running ML models on embedded devices. Its unified application programming interface (API) provides support for a variety of automotive-grade inference engines and neural network backends, including GLOW, TensorFlow Lite, Open Neural Network Exchange (ONNX) and Arm® CMSIS-NN. NXP’s eIQ Auto also integrates neural network compilers with a wide range of libraries and compute engines, giving users the flexibility to innovate with advanced ML features. By combining eIQ Auto ML with Sonatus AI Director, OEMs have a safe, reliable and accessible platform for reducing costs, accelerating development and bringing new functionality to customers.

Optimizing ML Models with eIQ Auto and Sonatus AI Director

To evaluate eIQ Auto model optimization, Sonatus AI Director was used to benchmark three engine anomaly detection models on an NXP GoldBox running Ubuntu 22.04. The C++ versions optimized with eIQ Auto ML showed major gains in performance and memory efficiency, with the Glow-compiled model delivering the best results.

Performance gain vs. base Python version

Scalable, Secure AI Execution in Complex Automotive Environments

Executing AI workloads locally requires a processing platform uniquely suited for heterogeneous compute environments, including MCUs, microprocessor units (MPUs) and neural processing units (NPUs), coupled with the safe execution required for scalable, safety-critical AI across different functionalities. Even with the expanding processing capabilities in today’s vehicles, the complexity of multistep machine learning operations (MLOps) workflows has been difficult to overcome. This is where the S32 platform shines. Its built-in accelerators make it possible to optimize data flow from vehicle networks to the Sonatus AI Director platform for edge-based data capture and real-time feedback loops.

The S32 platform makes it possible to effectively allocate CPU resources while preserving access privileges, supporting real-time processing, virtualization and secure execution environments across models. This means AI models can be hosted in a way that protects IP and supports reliable execution without interfering with real-time applications.

As the automotive industry continues to move toward software-defined vehicles, consumers are expecting driver-friendly features that go beyond infotainment and autonomous driving. Sonatus AI Director—powered by NXP’s production-grade eIQ Auto ML software and the S32 automotive platform—will help OEMs break out of current design limitations and enable real-time decisions, adaptive behaviors and autonomous capabilities directly within the vehicle.

Authors

Nikhil Reddy Kothapally

Software Product Manager – Tools, NXP Semiconductors

Nikhil drives software strategy for SDV architectures, shaping modular, customer-centric development tools. With 10+ years in automotive software across OEMs and Tier1s, he leads the development of virtual environments for SDVs powered by cloud-based DevOps and Edge AI integration for High-Performance Computing devices. A certified SAFe Agilist, Nikhil also holds a Product Management credential from the Indian Business School.

Kushal Shah

Automotive Processors Product Marketing, NXP Semiconductors

Kushal leads product marketing for the North America region, driving AI initiatives focused on automotive processors products. With a strong foundation in deep technical roles—most recently in system architecture—he brings a unique blend of strategic vision and hands-on expertise. Passionate about technology and relentlessly focused on customer needs, Kushal consistently aligns innovation with real-world applications.