Qualcomm Targets Autonomous Vehicles With Snapdragon Ride Platform
As expected, a key focus for CES 2020 is automotive. While the forecasts for autonomous vehicles (AVs) have been curtailed, the industry continues to rapidly drive (pun intended) toward full autonomy. Qualcomm, the leader in mobile technology, unleashed a flurry of automotive announcements at CES, including a new autonomous vehicle control platform dubbed the Snapdragon Ride Platform. While Qualcomm already generates more than US$600m annually from automotive communications, telematics, and infotainment systems, this is the company’s first foray into the command and control systems for automotive. Qualcomm first publicly indicated that a platform was under development at CES 2019, but this is the first release of information on the platform.
The Ride platform consists of application processors (there appear to be at least two variants) and AI accelerators designed to scale from SAE level 1 for Advanced Driver-Assistance Systems (ADAS) to SAE level 5 for full autonomy. The Ride platform leverages a domain-based architecture which allows for automakers to aggregate multiple vehicular domains into one centralized platform and also high-speed connectivity using networks based on Ethernet, PCIe or CAN-FD for communicating between all major systems. The platform also includes a complete software stack supporting multiple operating systems, including Linux and QNX, and applications designed and developed by Qualcomm for self-navigating highway pilot applications utilizing perception, localization, sensor fusion and behaviour planning. Customers, including OEMs and Tier 1 automotive electronics suppliers, will be able to customize and add additional applications. The platform will be available to lead customers in the first half of 2020 and production vehicles on the road in 2023.
According to the company, the Ride platform can scale from 30 TOPS to 700+ TOPS with dual applications processors and AI accelerators. At 700 TOPS, the total system power is only 130W. According to Qualcomm, most configurations should only require passive cooling. In addition, the chipsets, not just the AI core, will achieve 5.4 TOPS/W, double the performance efficiency of competing solutions. These figures are likely to improve every year with each new generation. In addition, the SoCs and AI accelerators are designed per ISO26262 standards for ASIL-D highly reliable safety-critical systems. More details on the applications processors and AI accelerators are expected in the coming months.
Qualcomm is using a fleet of vehicles for software development and training, as well as cloud-based training using a digital twin. I was given a ride in one of the vehicles at CES. The vehicle uses eight cameras and six radar sensors. The Ride platform does support LiDAR as well, but it was not implemented on the current demo vehicle. The demonstration showed the vehicle’s ability to drive in a highway environment, in which it merged, navigated, changed lanes, and even avoided an aggressive driver in a Chevy Camaro in front of us that almost spun out. The company is tested vehicles in multiple environments, including San Diego and Philadelphia, and with different driving scenarios.
For ecosystem development, Qualcomm is providing software-in-loop and hardware-in-loop test infrastructure as part of Snapdragon Ride that allows customers to validate their software implementations on the cloud, independent of actual hardware availability. Qualcomm is also working with vendors to qualify sensors for the Ride platform.
Qualcomm joins a long list of semiconductor companies developing solutions for autonomous vehicle platforms, including Intel, NXP, NVIDIA, Renesas, Samsung, plus startups and OEMs like Tesla that are developing custom AI automotive chips. Qualcomm, however, believes that it is well-positioned to compete in this segment because of its expertise in developing high-performance/low-power solutions combined with the company’s long experience in automotive and its ability to develop advanced solutions on an annual cadence. Qualcomm is the leader in modem solutions for telematics systems and for 5G mobile devices. In addition, Qualcomm products are being designed into multiple future infotainment platforms with 19 automotive OEMs. The company has been focused on automotive for many years and interest in the automotive segment was a key reason the company sought to acquire NXP, a deal that was derailed by the Chinese government. Qualcomm claims that the Ride platform has been in development for five years, but some of the technical elements have been in development for 10 years.
While details of the applications processors and AI accelerator were not shared at this time, the announcement included quotes from partners that shed a little more light on the platform, such as CPU and MCU cores from Arm, Infineon (Aurix), and Synopsys (ARC); a memory system from Synopsys (STAR); OS support from Blackberry QNX (QNX), and sensor support from ON Semiconductor.
It appears that GM is one of the OEMs that may be first to use the Snapdragon Ride platform as the two companies announced in a separate press release that they are continuing their partnership to “power digital cockpit, telematics and advanced driver-assistance systems (ADAS).”
Further highlighting the company’s focus on automotive, Qualcomm also made several other announcements. The 2020 Land Rover Defender is using the Snapdragon-powered PIVI Pro infotainment system along with dual LTE connectivity. Additionally, Qualcomm announced a C-V2X reference platform, Car-to-Cloud services, and new Wi-Fi/BT solutions for automotive.
Qualcomm is not the first semiconductor company to enter the autonomous vehicle segment, but the company has valuable intellectual property in connectivity and processing. In addition, the company does not have to rely on the segment for immediate revenue, as indicated in my previous automotive article. However, the company will have to demonstrate rapid progress, especially around training and verification, to compete in this segment.