Motional has selected Ambarella’s CVflow family of AI processors for its driverless vehicle fleet. The Ambarella processors work with Motional’s network of LiDAR, camera, and radar sensors to enable safe vehicle operation in diverse and challenging road conditions, including low-light and high-contrast situations. The announcement comes as Motional continues to ramp up its operations, most recently going fully driverless in Nevada and partnering with Cox Automotive for driverless fleet services.
“We’re at the forefront of commercializing robotaxis, and it’s critical that our vehicle architecture allows us to scale while maintaining the highest possible safety standards,” explained Joaquín Nuño-Whelan, vice president of hardware, Motional. “Ambarella’s processors provide the AI performance, low power consumption, and advanced image processing necessary for the camera perception to perform well under all conditions. This supports the safe operation of our vehicles as we bring driverless technology to consumers worldwide.”
Ambarella’s CVflow system on chips (SoCs) will be part of Motional’s driverless vehicles’ central processing module. The CVflow system will provide image and computer vision processing for cameras in the sensing suite, including the front-facing cameras. When it comes to Motional’s AI algorithms, the CVflow AI engine will facilitate important computer vision tasks, such as object detection, classification, and image segmentation. Last but not least, the SoC’s H.264 encoding will enable the efficient logging of video data from all cameras in the vehicle.
“We are proud to partner with Motional, a leader in driverless vehicle technology,” said Fermi Wang, CEO of Ambarella. “Motional’s autonomous vehicle expertise, including advanced AI algorithms running on our CVflow SoCs, will enable driverless vehicles that combine safety with exceptional performance.”
CVflow Features & Specifications
Ambarella’s CVflow SoC family includes AEC-Q100 qualified processors and processors enabling systems with ISO 26262 ASIL-B (D) functional safety levels. A complete set of tools is provided to accelerate the development and optimization of neural networks, supporting industry-standard training tools including Caffe, ONNX, PyTorch, and TensorFlow.