With 60tops deep learning performance, low power open scalable platformCentral processing ECU for automatic driving.
In December 17, 2020, Tokyo, Japan – TSE:6723, the global semiconductor solution provider, announced today that it has launched an excellent performance, ASIL-D level system on chip (SoC) – R-CarV3U for advanced driving assist system (ADAS) and automatic driving system (AD). R-CarV3U provides breakthrough 60TOPS performance, low power consumption and 96000DMIPS performance for deep learning processing capability, which meets the performance, security and scalability requirements of next generation autopilot ADAS and AD architectures.
The SOC is the latest member of Renesas automation for ADAS and AD. Based on the new R-CarGen4 architecture, the platform can provide scalability from the entry level NCAP application to the advanced automatic driving system. R-carv3u lays the foundation for Renesas’s next generation of high-performance ADAS and ad products.
“We are very pleased to launch a new generation of popular r-carsoc products for the next generation of ADAS and ad vehicles,” said Naoki Yoshida, vice president of digital product marketing division of Renesas automotive. R-carv3u continues to use the development resources such as ADAS and level 2 perception stack of the previous generation products r-carv3m and r-carv3h, and is used together with Renesas automation platform to build a smooth migration path for the realization of level 3 autopilot based on single chip, which can shorten the development cycle and realize safe production. “
Advanced SOC supports industry’s stringent asild standards
The automatic driving system requires functional safety to reach ASILD level, that is, the highest and most stringent vehicle safety level stipulated by ISO26262 road vehicle standards. R-carv3usoc integrates a variety of sophisticated security mechanisms to quickly detect and respond to random hardware failures. Most of the signal processing in SOC can reach asild level, while reducing design complexity and system cost, so that customer products can be quickly put into the market.
Deep learning technology with low power consumption
R-carv3u provides highly flexible DNN (deep neural network) Note 1 and AI machine learning functions. Its flexible architecture can run all the cutting-edge neural networks used for vehicle obstacle detection and classification tasks, provide high performance of 60tops, and realize low power consumption and passive cooling.
R-carv3u also provides a variety of programmable engines, including DSP for radar processing, multithreaded computer vision engine for traditional computer vision algorithms, image signal processing for improving image quality, and other hardware accelerators for key algorithms such as dense optical flow, stereo difference and object classification.
Advanced embedded software platform development, suitable for automatic driving
Renesas’s open and integrated development environment can help customers take advantage of the built-in hardware advantages of r-car platform, as well as low-power features and highly reliable real-time software to promote the rapid launch of solutions based on computer vision and deep learning.
Easy to use debugging and adjusting tools for heterogeneous multi-core hardware can achieve efficient software development. Comprehensive sample applications and online training resources can enable engineers at all levels to start their development work quickly. The compiler and code generator that meet the requirements of functional security and network security can achieve safe and reliable software development.
Users can also combine r-carv3u with Renesas’s high-performance, low-power rh850mcu, integrated power management PMIC and power devices, and match all the key components required by ADAS and adecu, so as to develop the system more efficiently and speed up the launch.
The r-carv3usoc sample is now on the market and is scheduled for mass production in the second quarter of 2023. For more information, please visit: renesas.com/r -car-V3U。
Note 1: deep neural network (DNN) is a kind of deep feedforward artificial neural network, which has been successfully used in visual image analysis, and more and more used in the automotive field, such as road detection or target classification.