Qualcomm launches its own self-driving car platform with soc and accelerator

Qualcomm has developed its own platform for self-driving cars. Snapdragon Ride is intended to power cars at all levels of autonomous driving. The platform consists of both hardware and software, which automakers can use to make cars autonomously.

Qualcomm showcased Snapdragon Ride at the CES trade show in Las Vegas. The platform consists of three parts. There is a Snapdragon system-on-chip, an accelerator and a software package. The soc is intended to do a large number of calculations itself, while the accelerator can take on various calculations to relieve the soc. Qualcomm says the platform uses artificial intelligence to make those calculations even more efficient, but the company doesn’t provide much further details.

The platform aims to help automakers with specific tasks required for autonomous driving. Qualcomm says it will allow manufacturers to build cars that can operate on all five levels of autonomy, from simple tasks like adaptive cruise control to full autonomy in busy cities. The platform is based on modular multicore chips and a GPU, although precise specifications are not known, and uses passive cooling or air cooling. The first tier of autonomous driving would already require 30 tera operations per second. At tier 5, it’s about 700 tops at 130 watts, Qualcomm says. The company says no additional water cooling or other systems are needed while driving. This reduces costs and simplifies the design.

Another part of the platform is a software package with which the standard aspects of autonomous driving can be built into a car, such as emergency braking, traffic sign recognition, automatic lane change or self-parking. According to Qualcomm, the software has a modular structure so that manufacturers can easily integrate their own systems. They can also add software for ‘complex cases’ while driving, such as models that mimic human behaviour.

Qualcomm works with various partners who can start using the platform from mid-year. By mid-2023, the company expects the first cars using the technology to hit the road.