JERUSALEM -- Mobileye N.V. introduces its 4th generation system-on-chip, the EyeQ4.
Leveraging expertise in designing computer-vision specific cores for over 15 years, the EyeQ4® consists of 14 computing cores out of which 10 are specialized vector accelerators with extremely high utilization for visual processing and understanding. The first design win for EyeQ4® in series production has been secured for a global premium European car manufacturer for production to start in early 2018. The EyeQ4® would be part of a scalable camera system starting from monocular processing for collision avoidance applications, in compliance with EU NCAP, US NHSTA and other regulatory requirements, up to trifocal camera configuration supporting high-end customer functions including semi-autonomous driving. The EyeQ4® would support fusion with radars and scanning-beam lasers in the high-end customer functions.
"Supporting a camera centric approach for autonomous driving is essential as the camera provides the richest source of information at the lowest cost package. To reach affordable high-end functionality for autonomous driving requires a computing infrastructure capable of processing many cameras simultaneously while extracting from each camera high-level meaning such as location of multiple types of objects, lanes and drivable path information," said Prof. Amnon Shashua, cofounder, CTO and Chairman of Mobileye. "The EyeQ4® continues a legacy that began in 2004 with EyeQ1 where we leveraged deep understanding of computer vision processing to come up with highly optimized architectures to support extremely intensive computations at automotive compliant power consumption of 2-3 Watts."
The EyeQ4® will feature four CPU cores with four hardware threads each, coupled with six cores of Mobileye's innovative and well-proven Vector Microcode Processors (VMP) that has been running in the EyeQ2 and EyeQ3 generations. The EyeQ4® will also introduce novel accelerator types – two Multithreaded Processing Cluster (MPC) cores and two Programmable Macro Array (PMA) cores. MPC is more versatile than a GPU or any other OpenCL accelerator, and with higher efficiency than any CPU. PMA sports compute density nearing that of fixed-function hardware accelerators, and unachievable in the classic DSP architecture, without sacrificing programmability. All cores are fully programmable and support different types of algorithms. Using the right core for the right task saves both computational time and energy. This is critical as the EyeQ4® is required to provide "super-computer" capabilities of more than 2.5 teraflops within a low-power (approximately 3W) automotive grade system-on-chip.
The enhanced computational capabilities give EyeQ4®-based ADAS the ability to use cutting-edge computer vision algorithms like Deep Layered Networks and Graphical Models while processing information from 8 cameras simultaneously at 36 frames per second.
The design was done according to the ISO-26262 standard and will provide a safety level of ASIL-B(D). The EyeQ4® will accept multiple camera inputs from a trifocal front-sensing camera configuration, surround-view-systems of four wide field of view cameras, a long range rear-facing camera and information from multiple radars and scanning beam lasers scanners. Taken together, the EyeQ4® will be processing a safety "cocoon" around the vehicle – essential for autonomous driving.
In addition to the EyeQ4® "high" capability version, at an ASP of approximately three times that of an ADAS functionality chip, Mobileye also plans the launch of the EyeQ4® "medium" variant within the same timeframe. The EyeQ4®M will include a subset of EyeQ4®'s computational cores, enabling a select group of functions. In producing a family of EyeQ4® processors Mobileye will ensure that car makers can deliver a scalable hardware solution with full code and pin compatibility, thereby reducing complete validation costs and ensuring that multiple feature bundles can be offered to the end customer at the best possible price.
Engineering samples of EyeQ4® are expected to be available by the fourth quarter of 2015. First test hardware with the full suite of applications including active safety suite of customer functions, environmental modeling (for each of the 8 cameras), path planning for hands-free driving and fusion with sensors, is expected to be available in the second quarter of 2016. Series production is supported for early 2018 start of production.