Marelli Unveils AI-based ECU for engine and vehicle control in motorsport

Marelli's VEC_480 AI-powered ECU delivers 2.5x more computing power and 10x greater bandwidth, enabling real-time AI for motorsport. It will be unveiled at the Professional MotorSport World Expo in Cologne.

Saitama, Japan- Marelli Holdings Co., Ltd. has introduced the VEC_480, an AI-powered Electronic Control Unit designed for engine and vehicle control in motorsport applications. Compatible with all types of vehicle propulsion from traditional to electric supports real-time AI computation onboard. The VEC_480 will be showcased at the Professional MotorSport World Expo in Cologne, Germany, on November 13th and 14th.

This technology sets a new standard for Vehicle Control Units (VCU) in motorsport, delivering enhanced performance, efficiency, reliability, and connectivity. Compared to previous VCUs, the VEC_480 offers 2.5 times higher real-time computing power, a 10x increase in inter-processor bandwidth, and improved RAM memory bandwidth, ensuring greater reliability in critical vehicle operations.

VCUs are high-performance control units that integrate different functions into a single device: engine and chassis control and actuation, data logging and gateway to telemetry and cloud, in-car networking.

Based on Marelli Motorsport’s solid know-how in vehicle control solutions, the VEC_480 is engineered to manage the increasing complexity of neural based algorithms in real time (milliseconds). This is achieved through the adoption of an advanced AI Accelerator (NPU), available with a high computing capacity up to 26 TOPS (Tera Operations per Second). This cutting-edge technology provides greater potential for internal vehicle networking and engine or vehicle management.

The powerful AI accelerator embedded in the device supports real-time AI inference with low latency and high efficiency, paving the way to neural virtual sensors, data inference with Artificial Intelligence, real-time video processing (track detection, object detection, and more), localization and positioning, performance analysis, predictive analysis, voice recognition. The technology is also compatible and supports the top AI frameworks, such as TensorFlow, TensorFlow Lite, Keras, PyTorch, and ONNX.