Continental relies on Xilinx SoCs for Next-Gen Radar systems

By Christoph Hammerschmidt

With the ARS540, Continental is presenting a high-resolution long-range radar sensor for the next generation of vehicles. The sensor not only detects the distance, relative speed and azimuth of objects, but also their height above road level – which is why Continental calls it a “4D” radar. Previous radar sensors have only detected objects in a very narrow, low height range above the road; this could lead to dangerous situations because, for example, truck trailers at right angles to the road were not registered. The new sensor also detects potholes and kerbs as well as tunnel entrances etc. 

The ARS450 has a sophisticated antenna array with digital beamforming. With this, as well as with 28 physical and 192 virtual antenna channels, the device is able to generate high-resolution radar images. The amount of data generated is considerably higher than with previous radar sensors, so Continental had to look for a more powerful processor platform. Xilinx’s Zynq UltraScale+ was chosen because it delivers 20 times the processing power of the previous platform for raw data and 10 times the performance for object tracking, said Willard Tu, Senior Director of Xilinx’s automotive division.

The high-performance processor platform is one of the reasons that the radar sensor also provides better quality data. It supports the prediction of the movement of objects (vehicles, pedestrians etc) in highly complex traffic scenarios by means of multi-hypothesis tracking. It classifies the radar echoes according to the type of traffic participant – i.e. it distinguishes between trucks, cars, bicycles and pedestrians. The same applies to echoes that can be assigned to traffic infrastructure. This allows the implementation of ADAS platforms that not only automatically adjust speed and brake, but can also take over the steering to a certain extent automatically. Additional sensors are required for the implementation of platforms for higher levels of autonomy – where the driver can also let the car drive itself at times and only has a monitoring function. According to the current state of the art, these are mainly lidar sensors.

Higher resolution, higher data rates: Continental’s  ARS540 “4D” radar sensor

“The Xilinx Zynq UltraScale+ MPSoC platform offers the high performance and advanced DSP capabilities we need to implement the ARS540, combined with adaptability and an industry-leading selection of network interfaces that can process a wide range of antenna data at extremely high overall transmission rates,” said Norbert Hammerschmidt, Head of Radar Program Management at Continental. Neither Continental nor Xilinx wanted to disclose which version of the Zynq UltraScale+ family, currently consisting of nine members, would be used in the new radar. However, Tu indicated that, in addition to the performance of the integrated processors (ARM Cortex-A53 and Cortex-R5 plus dedicated geometry and pixel processors), the generous availability of programmable logic and memory circuits played a role.

With the AR540, Continental will probably initially serve customers with premium class vehicles; in combination with additional sensor technology, it could also be used in robotic taxis. “4D Imaging Radar offers longer ranges and fields of view with better perception and is an important sensor technology for developing Level 2 to Level 5 systems to create a safer driving environment,” comments Cédric Malaquin, analyst at technology consultancy Yole Dévelopement. Malaquin estimates the market for such sensor platforms at $550 million by 2025, with an average annual growth rate (CAGR) of 124%.


Related articles:

Subaru uses Xilinx chips in central driver assistance system

Daimler chooses Xilinx for AI development

Xilinx bolsters MPSoC family for more computing power

Veoneer brings new ADAS generation to the road