Qualcomm revealed an autonomous driving platform at CES, branded Snapdragon Ride.

It consists of a family SoCs, a processing accelerator – both designed for ASIL-D level functional safety – and an autonomous stack.

“The combination of Snapdragon Ride SoCs, accelerator and autonomous stack offers automakers a scalable solution designed to support three industry segments of autonomous systems, namely L1/L2 active safety ADAS for vehicles that include automatic emergency braking, traffic sign recognition and lane keeping assist functions; L2+ convenience ADAS for vehicles featuring automated highway driving, self-parking and urban driving in stop-and-go traffic; and L4/L5 fully autonomous driving for autonomous urban driving, robo-taxis and robo-logistics,” according to the firm.

There are heterogenous multi-core CPUs, AI and computer vision (CV) engines and a GPU offering: 30TOPS of processing for L1/L2 applications, up to over 700TOPS for L4/L5 driving.

700TOPS is produced for around 130W of dissipation, according to Qualcomm, whicch said: “This platform can therefore result in designs that can be passively or air-cooled, avoiding the need for liquid cooled systems.”

“We’ve spent the last several years researching and developing our new autonomous platform and accompanying driving stack, identifying challenges and gathering insights from data analysis to address the complexities automakers want to solve,” said Qualcomm product manager Nakul Duggal.

The stack has been created for car makers and tier-1s that do not have their own self-driving algorithms, “offering software and applications for complex use cases, such as self-navigating human-like highway driving, as well as choice of modular options like perception, localisation, sensor fusion and behaviour planning,” said Qualcomm. “This software infrastructure for Snapdragon Ride supports customer specific stack components to be co-hosted with the Snapdragon Ride Autonomous Stack components.”

Explaining the ‘accelerator’, Qualcomm said: “Through the autonomous driving accelerator, Qualcomm Technologies brings energy efficient compute capabilities to mainstream vehicles”, adding that AI processing for camera image signal processors, DSPs for sensor signal processing, CPUs for planning and decision making, GPUs visualization and user experience, and dedicated safety and security sub-systems will be split across the SoC and accelerator.

Snapdragon Ride is expected to be available for pre-development to automakers and tier-1 suppliers in the first half of 2020. Qualcomm Technologies anticipates Snapdragon Ride-enabled vehicles to be in production in 2023.

Ride platform partner ecosystem:

• Arm: functional safety
• Synopsys: ISO 26262 certification, AEC Q100 and quality management
• BlackBerry QNX: safety operating system and hypervisor
• Infineon: Aurix microcontrollers
• Elektrobit: EB corbos automotove software for Adaptive AUTOSAR
• ON Semiconductor: sensors