MOBILEVE-迈向自动驾驶时代（英文）-2019.1-44页_4mb

Summary of Document Content

Core Content

The document outlines Mobileye's progress and strategy in the field of autonomous driving and advanced driver assistance systems (ADAS). It highlights the company's design wins, collaborations, technological advancements, and safety initiatives as part of its mission to develop safe, scalable, and economically viable autonomous driving solutions.

Main Points and Key Information

Autonomous Driving Progress

• 30 design wins in 2017 across 27 OEMs and ~70 car models.
• In 2016, Mobileye had 12 design wins.
• The EyeQ® series of chips was a core component of these systems.
  • EyeQ3: 40nm, 0.25 TOPs @ 3W, launched in 2014.
  • EyeQ4H: 28nm, 1.5TOPs @ 15W, launched in 2018, with 12 additional launches starting from 2019.
  • EyeQ5H: 7nm, 2.5 TOPs @ 6W, launched in 2020.
  • Nvidia Xavier: 30TOPs / 30W, used as a benchmark.
• 24M EyeQs were shipped to date, with ~9M EyeQs shipped in 2017.

Key Features and Technologies

• REM (Road Experience Management): A crowdsourcing-based system for HD-map automation and real-time localization.
  • Collects data from vehicles to create RSD (Road Safety Data) at 10kb/km.
  • Supports L2+, L3, and L4 automation.
  • Provides safety guarantees through the RSS (Responsibility-Sensitive Safety) model.
• RSS (Responsibility-Sensitive Safety): A formal model of human judgment that ensures safety and scalability for autonomous vehicles.
  • Defines safe states and proper responses to avoid accidents.
  • Rules include:
    • A hit from behind is not the front car's fault.
    • Unless the front car performed a reckless cut-in.
    • Right-of-way is given, not taken.
    • Be careful of areas with limited visibility.
  • Ensures no "butterfly effect" and separates comfort planning from safety planning.

2017 Program Launches

• 30 design wins in 2017.
• 15 programs to be launched in 2018.
• 24M EyeQs shipped to date.
• 27 OEMs and 14 OEMs in 2018 (4 of which are Chinese).
• New features in 2018:
  • 3DVD: Traffic Lights Detection and Recognition.
  • Semantic Free Space and Holistic Path Prediction.
  • REM integration for L2+ and L3 systems.

Collaborations and Partnerships

• Esri Collaboration: Integration of REM data with Esri's mapping platform for urban safety and public transit.
• REM Deployment Partners:
  • KoMoD Research Project in Dusseldorf, Germany (750 vehicles).
  • Gett in London (500 vehicles).
  • Buggy in New York (2,000 vehicles).
  • Road Safety Authority DGT in Spain (~5,000 vehicles).
  • Guard Insurance in the US (1,000–2,000 tow trucks).
• Strategic Partnerships for L4:
  • BMW, Fiat-Chrysler, SAIC, NIO.
  • 3 OEMs ongoing with sourcing decisions.

REM Applications

• REM in Aftermarket: Mobileye 8 Connect™ supports REM in non-OEM vehicles.
• REM Data Use Cases:
  • Accident hotspots.
  • Infrastructure deficiencies.
  • Parking spaces.
  • Traffic light and sign surveying.
  • Map your city for L4 development.

Strategic Goals

• Level-4/5 Automation: Focus on full autonomy and derivatives for L2, L2+, and L3.
• Economical Scalability:
  • Crowdsourcing for HD-map automation.
  • Scalable, low-power SoC with ATOM CPU.
  • Real-time localization using REM and HD-map.
• Safety Guarantees:
  • Decoupling sensing from planning.
  • RSS model to ensure safety and transparency.
  • Statistical safety as a target for AVs.

EyeQx Family Overview

• EyeQ3, EyeQ4H, and EyeQ5H are part of the EyeQx family, each with increasing performance and efficiency.
• EyeQ5 was introduced in 2018, with series production starting in 2020.
• Intel's role includes:
  • Providing open-compute platforms like open-EQ5.
  • Supporting REM and RSS through industry and regulatory collaborations.

Key Technologies

• REM (Road Experience Management): Enables HD-map automation and real-time localization.
• RSS (Responsibility-Sensitive Safety): A formal safety model to ensure AV safety and scalability.
• EyeQx Family: A series of vision processing units (VPUs) that support ADAS and autonomous driving.

Conclusion

The document emphasizes Mobileye's leadership in autonomous driving technology, highlighting its safety-first approach, scalable solutions, and collaborations with OEMs and industry partners. It also outlines the strategic roadmap for L2+, L3, and L4 automation, supported by REM, RSS, and EyeQx chips. The aftermarket is also a focus, with Mobileye 8 Connect™ and REM deployment to enhance urban safety and infrastructure mapping.