16nm heterogeneous multi-processing SoCs target ADAS apps

Xilinx Inc. is set to unveil its 16nm heterogeneous multi-processing SoC solutions for advanced driver assistance systems (ADAS) at CAR-ELE Japan. The company's system architects and ecosystem partners will reveal how Xilinx's All Programmable technology allows smarter, connected and differentiated ADAS applications with high levels of performance, power and integration.

The Zynq UltraScale+ MPSoC Multi-Application Execution features the automotive industry's first 16nm MPSoC that started shipping in October 2015. It showcases multiple Linux-based applications on the quad core ARM Cortex-A53 devices, a bare metal application on the dual ARM Cortex-R5 devices, and an OpenGL ES2.0 application on the Mali-400 GPU all running in parallel. These high performance cores, along with the high speed FPGA fabric and wide interconnects for bandwidth, as well as the flexible interfaces all on a single chip is the next generation in ADAS camera and fusion processing silicon.

The 5-Camera ADAS development platform showcases advanced multi-camera, multi-mode processing using five cameras, 4x camera for surround view mode and 1x forward looking camera. This will demonstrate 2D and 3D surround views, automated stitching and calibration of multiple cameras using the logiOWL IP, and Multi-Object Classification on the front camera.

The Real-time Roadway Vehicle/Motorbike Pedestrian Detection displays actual roadway video being processed by the Zynq-7000 All Programmable SoC for parallel real-time vehicle, motorbike/bicycle and pedestrian classification and tracking, as well as graphic overlay generation.

The Advanced ADAS development video recording platform features the logiRECORDER from Xylon that can record up to six synchronised video streams with CAN and GPS metadata and allows developers to play data back through the logiADAK for algorithm refinement. This will be demonstrated with the PC-based logiBUILDER GUI connected via Ethernet to build customised user views.

The ADAS Classification Training (logiSTK) showcases an advanced annotation tool to classify roadway objects of interest including pedestrians, vehicles, motorbikes and bicycles. This tool also allows for customised training of classifiers and is compatible with logiRECORDER videos.