Intel Corporation Presents at Deutsche Bank AutoTech Conference, Nov-10-2020 - SEHK:4335
SEHK:4335
Amnon Shashua [Former Senior Vice President] 💬
Amnon Shashua, likely speaking on behalf of Mobileye, a subsidiary of Intel Corporation, discussed several key topics related to the company's strategy and technological advancements in the automotive industry, particularly concerning advanced driver-assistance systems (ADAS) and autonomous driving technologies. Here’s a detailed summary of his statements:
Strategy Overview:
- Three Pillars of Business:
- Driving Assist: Main revenue generator today, ranging from Level 1 to Level 2 systems. Mobileye works with Tier 1 suppliers and carmakers.
- Crowdsourced Data: Technology announced in 2015 that collects data from vehicles equipped with front-facing camera technology. This data is used to build high-definition maps and gain insights about road structures.
- Full-Stack Self-Driving System: Targeting consumer autonomous vehicles, with robo-taxis as an interim step.
Business Achievements:
- Silicon Sales: More than 65 million chips sold, equating to over 65 million cars on the road with Mobileye technology.
- Production Programs: 48 running production programs in the current year, with 29 design wins.
- Five-Star Rated Vehicles: Since 2018, Mobileye technology is present in 70% of these vehicles.
Approach to Driving Assist and Autonomous Driving:
- Camera Subsystem as Anchor: Camera-only subsystem serves as the foundation, with additional modalities like radars and LiDAR added for increased reliability.
- SuperVision Product: Based on two EyeQ5 chips, 11 8-megapixel cameras, and integrated high-definition maps, providing hands-free driving capabilities.
- Geely Partnership: The first OEM partnership for SuperVision, launching in Q3 of the following year in China.
Technology for Map Building:
- Crowd-Sourced Mapping: Every vehicle with a front-facing camera sends sparse data to the cloud, where high-definition maps are automatically generated.
- Coverage and Updates: Currently, 7.5 billion kilometers of roads have been collected globally, with 8 million kilometers covered daily. Approximately 50,000 kilometers of map data are updated weekly, projecting to 1 billion kilometers of roads collected and 1 million kilometers of map data updated daily by 2024.
Formal Safety Layer: RSS
- Responsibility Sensitive Safety (RSS): A formal model of decision-making that defines careful driving judgments. It consists of two components:
- Parametric Assumptions: Specifies assumptions made by drivers when merging into traffic.
- Worst-Case Analysis: Takes the worst-case scenario to ensure safe driving.
- Standardization Efforts: Intel chairs an IEEE workshop on formal definitions of decision-making, and RSS has been standardized in China.
Mobility as a Service:
- Partnerships: Mobileye collaborates with public transit operators and transportation network companies (like Uber and Lyft) to provide autonomous vehicles as a service (VaaS), rides as a service (RaaS), and mobility as a service (MaaS).
- Vertical Integration: Flexibility to break down services into different business models and partnerships, including licensing parts of the stack or managing operations themselves.
Challenges:
- Accuracy: Meeting stringent mean-time-between-failure requirements, targeting 2 to 3 orders of magnitude better than human-level statistics.
- Geographic Scalability: Enabling consumer vehicles to drive anywhere through crowd-sourced REM technology.
- Regulatory Certainty: Addressing legal and liability issues for deploying autonomous vehicles.
Redundancy and Economical Impact:
- Redundant Systems: Separate camera and radar/LiDAR subsystems improve reliability and reduce costs by leveraging the camera subsystem for driving assist evolution.
- Flexibility in Sensor Setup: Imaging radars and front-facing LiDAR provide a cost-effective redundancy approach.
Mapping Approaches:
- Without Maps: Relying solely on online sensing and basic navigation maps.
- High-Definition Maps Using Specialized Vehicles: Not scalable due to high costs and manual labor required.
- Crowd-Sourced High-Definition Maps: Utilizes data from millions of vehicles to build detailed maps automatically, enabling geographic scalability and additional services like infrastructure surveys.
Regulatory Standardization:
- Country-by-Country Approach: Working with regulatory bodies to achieve industry consensus, such as through the IEEE P2846 workgroup chaired by Intel.
Partnerships and Ecosystem:
- Driving Assist Domain: Collaborates with Tier 1 suppliers and OEMs, often acting as a Tier 2 supplier.
- SuperVision and Beyond: Provides a full subsystem directly to carmakers, as seen with the Geely partnership.
- Mobility as a Service: Forms partnerships with public transit operators and transportation network companies to provide end-to-end mobility solutions.
Cooperation and Vertical Integration:
- Cooperative Approach: Prefers partnerships for better business models and capital expenditure.
- Moovit Integration: Relies on Moovit for additional layers of operations and customer-facing applications, enabling a complete mobility service offering.
These insights provide a comprehensive overview of Mobileye’s strategic approach and technological advancements in the autonomous driving sector.