Autonomous Car Chips Beef up Sensor Fusion

The recent announcements from NXP and STMicro accentuate the journey from ADAS to autonomous vehicles.

NXP Semiconductors, who became the market leader in automotive chips after its $12 billion acquisition of Freescale Semiconductor, is now cementing its lead on the next connected-car frontier: autonomous vehicles. At the NXP FTF Tech Forum, previously known as Freescale Technology Forum (TFT), the chipmaker has showcased its new advanced driver assistance systems (ADAS) and autonomous vehicle platform: Bluebox.

It's an open and modular vehicle platform that comprises of two processors: S32V vision processor for safety applications and LS2088A embedded processor for compute-centric tasks like sensor fusion. The S32V vision processor boasts graphics engines, image-processing accelerators and advanced APEX image processing along with high-performance ARM cores.

The LS2088A embedded processor integrates eight 64-bit ARM Cortex-A72 cores running at 2 GHz. Moreover, along with specialized accelerators and communications interfaces, it features DDR4 memory controllers to ensure extremely low latency. NXP claims that these chips meet the ASIL functional safety requirements mandated by the automotive industry.

Bluebox is a full-fledged automotive platform encompassing cameras, radars, and V2X systems.

At the NXP FTF Tech Forum 2016, being held on 17-19 May in Austin, Texas, the firm displayed the BlueBox central computing engine alongside cameras, radars, LiDAR, vision sensing, and onboard vehicle-to- infrastructure (V2X) system. NXP also demonstrated BlueBox in the world's first 3D-printed car "Strati" built by Chandler, Arizona–based Local Motors.

NXP claims that its BlueBox ADAS and self-driving platform is ready for production (PDF), and four large automakers are already testing it. Furthermore, NXP managers are confident that the self-driving platform can serve Level 2, Level 3 and Level 4 autonomous cars. Level 4 can handle all aspects of driving—steering, speed control, and automatic brakes—under specific circumstances such as low-speed traffic jam and merging onto an expressway.

 

ST's Preemptive Strike

Meanwhile, NXP's automotive nemesis STMicroelectronics has unveiled its own compute engine for autonomous vehicles. ST and its longtime ADAS partner Mobileye have announced that their fifth-generation vision system-on- chip (SoC)—EyeQ5—will now support up to 20 external sensors, including camera, radar, and LiDAR.

The new EyeQ5 chipset boasts eight multithreaded CPUs that are coupled with Mobileye's 18 vision processor cores. That's a significant boost from the fourth-generation EyeQ4 vision chipset that featured four CPU cores alongside six vision processors.

The EyeQ5 automotive chipset is 8x times more powerful than its predecessor EyeQ4

ST claims that EyeQ5 supports the highest level of safety in automotive applications—ASIL B (D) specification defined in the ISO 26262 standard. It further builds a security defense by integrating a Hardware Security Module that creates the root of trust from an encrypted storage device.

A closer look at the ST-Mobileye announcement shows that while its EyeQ chips have been strong on the automotive vision front, they are now catching up on the sensor fusion side, a forte of NXP's Bluebox autonomous car engine. On the other hand, Bluebox, a control and sensor fusion platform, is making strides in the computer vision areas.


However, Bluebox has the time-to- market advantage. Automakers are already sampling NXP's product offerings in the Bluebox platform while ST's chipset for autonomous vehicles will be available in 2018.