Renesas Electronics introduces an upgraded R-Car V3H SoC to enhance deep learning for smart camera applications

Renesas Electronics announced the latest upgrade to the R-Car V3H System-on-Chip (SoC), bringing significantly improved deep learning capabilities to smart camera applications, including driver/passenger monitoring (DMS/OMS), in-vehicle front-facing cameras, round-view systems, and automatic parking up to Level 2+ for most vehicles. The upgraded SoC integrates sensor fusion in the real time domain, up to ASIL Class C metrics, and an architecture optimized for intelligent computer vision. It provides Oems and Tier 1 suppliers with high performance, low power solutions at a competitive system cost and supports the latest NCAP 2020 requirements and migration to the NCAP 2025 3-star technology roadmap.

Based on the SoC released in February 2018 and incorporating the latest recognition technologies, such as integrated IP for convolutional neural networks (CNNS), the upgraded R-Car V3H delivers up to four times the performance of earlier versions in CNN processing, while maintaining low power consumption levels. Achieve overall processing performance up to 7.2 TOPS including all computer vision IP.

The highly integrated SoC supports security metrics up to ASIL Class C in the real time domain, reducing the need to use an external security microcontroller (mcu) to manage sensor fusion and final decision execution. The R-Car V3H has a rich and proven IP that supports sensing stacks, fusion with radar and/or liDAR sensors, and ISPs that can support up to 8-channel megapixel cameras, enabling Oems and Tier-1 suppliers to achieve faster time-to-market with lower material costs.

Naoki Yoshida, vice president of Digital Product Marketing at Renesas Electronic Automotive, said: "As automakers continue to explore new ways to enable smart camera applications, the growing popularity of smart system applications such as driver monitoring, parking assistance, driver recognition and passenger sensing has fueled a higher demand for deep learning capabilities. The upgraded SoC gives R-Car V3H users the flexibility, ease and cost effectiveness to achieve higher computer vision performance and create computer vision for the next generation of front-facing camera systems. "Renesas' open, scalable SoC platform for ADAS and highly autonomous driving applications continues to grow, and we are proud to be able to continue to bring cutting-edge innovation to our customers with this platform."

Key features of the R-Car V3H SoC

● Provides computing performance up to 7.2 TOPS, including 3.7 TOPS for CNN, with optimized performance and power balance

● Support for ASIL Level D development processes to achieve full SoC system functionality

● Supports measurement goals for ASIL Level B (sensor layer, application processor) and ASIL Level C (real time domain) security metrics

● Fully compatible with R-Car V3H SoC hardware and software already in production

● Integrated full range of automotive peripherals including CAN, Ethernet AVB and FlexRay

● A full suite of video processing and image recognition IP for advanced sensing and recognition, including CNN-IP, computer vision engine, image distortion correction IP, stereo vision, classifier and dense optical flow IP

The upgraded R-Car V3H joins Renesas' expanding SoC portfolio in the open, flexible Renesas autonomy platform, which provides Oems and Tier 1 suppliers with complete scalability from entry-level NCAP applications to highly autonomous driving systems. Customers can also combine R-Car V3H SoCs with Renesas' high-performance, low-power RH850 MCUS, integrated power management ics and power transistor devices to provide the critical components needed for increasingly complex functional safety systems.