At Auto Shanghai 2021, German mega-supplier ZF is unveiling a number of new technologies and design wins. A new ProAI supercomputer, suitable for all applications for automated driving SAE Levels 2 to 5, has gotten several major orders with the start of series production in 2024. One of those is a cost-effective automated valet parking solution for serial production in 2022. The company will produce will Full-Range Radar for an OEM in China. And commercial vehicle customers will see a few new offerings.

 

New ProAI supercomputer

Headlining the announcements was the debut of its next-generation ProAI supercomputer, which is not only smaller and more economical but also more powerful. It has better computational performance, with up to 66% more TOPS (trillion operations per second), and it consumes up to 70% less power at 3 TOPS/W on average.

“It is currently the most flexible, scalable, and powerful automotive-grade supercomputer in the world,” said Dr. Holger Klein, ZF Board Member for the Asia-Pacific region, at Auto Shanghai.

The new ProAI supports ZF’s belief that the intelligence of future vehicles will be controlled by a few extremely powerful central computers that will run the computationally intensive software functions that control critical vehicle domains. These include automated driving, electric mobility, vehicle motion control, and integrated safety. The company believes that high-performance computers and intelligent software functions are the key enablers for the software-defined vehicles of the future.

Designed for the requirements of software-defined vehicles and their new electric/electronic architectures, the AI (artificial intelligence)-capable computer can serve as domain, zone, or central controller. Suitable for any vehicle type and level of automated or autonomous driving from Level 2 to Level 5, it is planned for production in 2024.

“Several significant orders for our high-performance computing platform confirm our vanguard position to drive vehicle intelligence,” added Klein.

Optimized for deep-learning processes, the ProAI board offers a 360° GPU-driven fusion of sensor data including environmental measurement data from radars, LiDARs, cameras, and even audio. The automotive-grade product is resilient and reliable under harsh conditions and offers the latest cybersecurity safeguards. A modular setup means it can be equipped with SoC (system-on-chip) variants from different manufacturers, and it can operate ZF’s own software or those from other developers. Even with passive, air, or liquid cooling, most models will come in the same 24 x 14 x 5 cm (9.4 x 5.5 x 2.0 in) housing.

With the ProAI, ZF can offer OEMs a custom-fitted solution for all their vehicle platforms, software applications, and E/E architectures—through a single type of control unit to increase efficiency and reduce costs. Standardized connectors and the option to link more units together, making it flexible for use and installation in many types of vehicles. It offers the flexibility to achieve computing capability from 20 to 1000 TOPS. An MDI (measurement data interface) can forward collected sensor data unaltered to a central storage system for development and testing purposes, making it much easier for developers to train AI for autonomous driving.

ZF is currently developing the first valet parking system globally that enables driverless parking, relying only on the vehicle’s sensor set and independent of a pre-mapped parking garage infrastructure. At Auto Shanghai 2021, it is demonstrating the vSLAM (visual simultaneous localization and mapping) technology that enables centimeter-level localization and real-time map generation. The sensor set is mainly based on a front camera, a front radar, four surround cameras, and twelve ultrasonic sensors, but is scalable with more advanced sensors as well as connectivity.

“The entire system is developed in China and will have its debut at a Chinese car manufacturer at the end of 2022,” said Renee Wang, President of ZF China and Senior Vice President Operations for the Asia-Pacific region. “We believe that this infrastructure-independent automated valet parking system from ZF will be a cost-effective solution for many global OEMs.”

 

Full-Range Radar for SAIC

In Shanghai, ZF announced a production contract win for its Full-Range Radar technology with SAIC Motor Corp. The supplier says that the high-resolution radar perceives vehicle surroundings in four dimensions, including height, making it similar in capability to optical sensors such as cameras and LiDAR. In combination with these technologies, the new radar can help provide the necessary safety and reliability for semi to highly automated driving including SAE Level 4.

“ZF’s Full-Range Radar represents a significant step forward in sensing technologies,” said Christophe Marnat, Executive Vice President, for ZF’s Electronics and ADAS division. “With its high levels of object and scene recognition and long-range detection, the Full Range Radar is a key to help meet the high perception requirements of L3 and L4 AD at a competitive price level. This sensor is also capable of significantly increasing the potential performance of advanced safety and L2+ driving applications.”

The technology offers high resolution for 4 dimensions, adding elevation (height) to range, velocity, and azimuth (horizontal) capabilities. The addition helps generate an enhanced 3D image of the traffic situation, enriched with speed information, resulting in high-resolution environment sensing. This type of data helps a vehicle on a highway to detect the end of a traffic jam under a bridge at an early stage and brake accordingly. The technology also provides information that helps detect the edge of the road and whether there are free passing areas at the road’s edge.

The new radar offering has a considerably higher resolution than mid-range radars. Instead of the typical 12 channels (three transmitters and four receivers), it has 16 times greater channels, with 4 MMIC (monolithic microwave integrated circuit) chips combined, resulting in 192 available channels.

The resulting high information density enables very detailed object recognition. For example, it receives around 10 data points from a pedestrian instead of just one or two compared to typical automotive radars, allowing more precise information. It also registers the speed of the measured object for each measuring point, for it can resolve the movement of individual limbs, for instance, to potentially recognize the direction in which a pedestrian is walking.

At 350 m (1150 ft), ZF says the new radar’s range is well above the current state-of-the-art. It uses the 77-GHz band and Fast Ramp FMCW (frequency modulated continuous wave modulation shared with other radar sensors from ZF.

ZF says that the Full-Range Radar is an important addition to its sensor suite for automated driving functions. With an aperture angle of +/-60 degrees, it is designed for a wide range of situations, from slow city traffic to driving on country roads and highways.

The company will begin supplying the technology to SAIC in 2022.

 

Commercial vehicle offerings

On the commercial-vehicle front, ZF’s most advanced emergency braking system, OnGuardMAX, had its global launch at the show and its debut on Chinese roads. The technology can identify and react precisely to a broad range of moving and stationary objects including vehicles and pedestrians.

The company also introduced OptiPace, a predictive economical cruise control system for improved vehicle efficiency and reduced CO2 emission in support of global sustainability initiatives. It can anticipate and adapt the vehicle’s most economic speed based on the road topography ahead to help minimize fuel consumption, brake pad wear, and emissions.

The Autonomous Driving Open Platform Technology (ADOPT) offers an efficient and flexible industry model for the development of autonomous driving applications for commercial vehicles. It translates instructions from the autonomous driving AI into real vehicle motion commands by enabling the control of all relevant vehicle actuation systems.