Vision MPU enables autonomous cars

2nd March 2015

Siobhan O'Gorman

0 0

A vision MPU, claimed to be the first automotive vision SoC with the requisite reliability, safety and security measures to automate and ‘co-pilot’ a self-aware car, has been introduced by Freescale Semiconductor. Leveraging a host of automotive-grade technologies, the S32V takes the industry beyond the current, convenience-centric ‘assist’ paradigm and toward an era where cars can capture data, process it and actually share control with drivers in critical situations. This capability establishes the essential bridge from the current ‘assist’ era toward the fully autonomous vehicles of tomorrow.

Part of Freescale’s SafeAssure functional safety programme, the MPU is structurally designed to comply with stringent ISO 26262 functional safety standards, and engineered for automotive-grade quality metrics measured in parts per billion. This reliability, together with the performance/power ratios of integrated 2nd gen CogniVue APEX Image Cognition Processing technology, additionally supports the fusion of vision data captured by the S32V with multiple other data streams, including radar, LiDAR and ultrasonic information to enable optimal resolution and image recognition accuracy.

The MPU’s unmatched safety stems from structural, automotive-grade technologies and techniques that assure optimal reliability. Redundant signal paths, software error checking functionality, hardware fault detection and hard partitioning of on-chip processing domains allow the system to safely shut down and perform a controlled reboot without compromising braking or steering.

The S32V also features secure boot, network-grade crypto engines, secure keys, and support for secure hardware extension specifications published by many of the world’s top automakers. With this unparalleled level of security, the MPU helps protect against theft of software algorithms and other IP, while enhancing overall vehicle safety by helping to prevent external attacks and unauthorised access to V2V communications. Discrete encryption domains provide an added layer of protection within the car’s network. If a hacker gains access to one node of the car’s system, they cannot access other nodes; hence attacks are effectively isolated and quarantined.

The S32V vision MPU integrates robust hardware including high-performance CogniVue APEX-642 core image processing technology, as well as four ARM Cortex-A53 cores.

The MPU’s software platform includes Green Hills Software’s INTEGRITY, a safety-certified, RTOS including a powerful set of ISO 26262, ASIL-D certified development tools with highly optimised target solutions. The platform also includes Neusoft Corporation’s advanced, real-time object recognition algorithms to seamlessly detect partial objects, allowing the S32V to interpret and distinguish between road hazards and pedestrian risk.

Full market availability for the S32V is expected in July 2015.

“The next step in the evolution of self-driving cars will rely on automated systems that accurately process information, make decisions and take the requisite actions,” said Luca De Ambroggi, Principal Analyst, Automotive Semiconductors, IHS. “This requires the highest levels of reliability, which are rarely present in consumer-grade silicon. The essential requirement for fail-safe reliability cannot be achieved using consumer-grade silicon without significant costs in system power or form factor.”

“Many automotive vision systems available today are based on consumer-oriented silicon solutions originally designed to enhance gaming graphics or run smartphone apps. But in a new era where cars will serve as trusted co-pilots, utilising consumer-oriented silicon is fundamentally unwise,” said Bob Conrad, Senior Vice President and General Manager, Automotive MCUs, Freescale. “Relying on anything less than automotive-grade silicon to take control of a vehicle and make critical driving decisions is simply not acceptable, not for me, not for my family and not for my customers.”