
NXP, Kalray defy Nvidia with automotive AI computing platform
NXP and the French processor startup company Kalray announced at the Consumer Electronics Show (CES) in Las Vegas that they have entered into a strategic partnership to develop a control computer for autonomous vehicles. Kalray is regarded as a pioneer for processors for the application of artificial intelligence and considers itself well prepared for the particularly demanding control tasks that require maximum functional safety.
Machine learning für safety-critical functions? Kalray says yes
Together, the duo wants to combine the scalable, functionally safe NXP product portfolio of silicon for driver assistance systems (ADAS) with Kalray’s MPPA processors (Massively Parallel Processor Array). The future platform is considered relevant for industry because it meets the high demands for performance, safety and commercial factors especially for Level 2 and 3 in the short term, while at the same time keeping an eye on future Level 4/5 approvals. The partners have also set themselves the goal of eliminating the safety deficiencies of today’s autopilots and autonomous test offerings.
At CES, the partners will present the platform for the first time. It uses a prototype of Apollo, the open software solution for autonomous driving of Baidu. Also at CES, the German automotive supplier ZF Friedrichshafen presented a computer platform for the same field of application; it uses the powerful Xavier chips from Nvidia and the multiprocessor arrangement Ultrascale+ from Xilinx.
ZF demos automotive supercomputer, autonomous minibus
In terms of safety, the ecosystem around autonomous driving is confronted with numerous technical challenges and critical questions and concerns. Recent studies show that while consumers are excited about an autonomous future, they have reservations about whether self-driving vehicles will ever really be safe. Spectacular accidents with prototypes and test vehicles have also raised additional doubts. Headlines were made, for example, by an accident in which a self-driving car from Uber did not recognise a cyclist crossing the road and fatally injured her. In order to overcome the technical hurdles and the lack of customer confidence, fail-safe and redundant vehicle systems are required so that the central computer can protect the occupants with a complex, comprehensively tested safety approach. With more than 25 years of experience in the field of functional safety, NXP is well equipped for these tasks.
NXP and Kalray are now joining forces to jointly develop a central computing platform based on the highest security standards such as ISO 26262. NXP provides the host processor – its S32 processor – with safety-critical ASIL D and ASIL B capabilities to cover the computing requirements of a central computer and also to address the path planning task. Kalray introduces its premium MPPA high performance processors that enable safe machine learning as the vehicle captures its environment.
The first step in the partnership will be the integration of the Kalray MPPA processors into the NXP BlueBox, an embedded platform for autonomous driving. Special features will be low power consumption and high safety, arm technology will be used, open standards will be supported.
Related articles:
NXP enables AI-based ADAS platform; rolls V2X chip
Continental develops auto-driving computing platform with Nvidia
Daimler tests robo taxis – on a platform from Bosch and Nvidia
