Horizon Robotics releases 2nd generation AI processor
Horizon Robotics unveiled its 2nd generation automotive AI processor, the Journey 2. The company says it is one of the first AEC-Q100-compliant deep-learning compute options available to Tier 1s and OEMs looking to enable the next generation ADAS capabilities and intelligent cockpit experiences with limited power consumption and excellent performance and efficiency.
The processor uses Horizon Robotics' efficient programmable deep learning computation cores—BPU, dual-core ARM Cortex-A53 and dedicated image signal processor. Its dedicated BPU cores reportedly provide a deep-learning performance of 4 TOPS (equivalent computing power relative to GPU) with typical energy consumption at 2 W. It is built on TSMC's stable 28 nm HPC+ process.
Journey 2 can is reportedly able to process 4K video inputs at 30 frames per second and complete the parsing of 23+ semantic categories, 2D+3D detection of hundreds of objects, distance and speed estimation, and other key perception features to meet Euro NCAP 2022 requirement. Horizon Robotics also offers production-ready perception software for ADAS and autonomous driving as an option for turnkey perception solutions based on Journey 2.
Along with Journey 2, Horizon Robotics is launching the Horizon OpenExplorer AI Toolkit, which allows customers to deploy custom deep-learning networks on Journey 2 to leverage its compute performance. The OpenExplorer AI Toolkit includes design examples and best practices to improve time-to-market for customers.
Production vehicles equipped with Journey 2 are set to begin mass production as early as the second quarter of 2020.