The enhanced SoC enables real-time sensor fusion with an architecture optimised for computer vision. This enables Renesas to offer OEMs and Tier1s a solution that combines high computing power with low power consumption and supports the latest requirements for NCAP 2020 and the roadmap for NCAP 2025 (3 Star).
Building on state-of-the-art image recognition technology introduced in February 2018 with the R-Car V3H, which includes integrated IP for CNN (Convolutional Neural Networks), the enhanced R-Car V3H delivers four times the performance in CNN processing compared to the previous version, according to Renesas. It achieves a total processing power of up to 7.2 TOPS.
The highly integrated SoC supports functional safety in the real-time domain up to safety level ASIL C according to ISO 26262, allowing system designers to eliminate the need for an external safety microcontroller (MCU) that governs sensor fusion and final decision-making actions. The R-Car V3H has a range of IPs that allow the execution of perception stacks and sensor fusion with radar and/or lidar. In addition, the integrated ISP supports up to eight MP cameras. This enables OEMs and Tier1s to get to market faster with lower component costs.
The R-Car V3H was developed following an ASIL-D process to avoid systematic failures for the complete SoC. It supports the metrics requirements for ASIL B (sensor layer, application processor) and the functional safety requirements for ASIL C (real-time domain). Both hardware and software are compatible with the R-Car V3H version currently in mass production at Renesas.