Slide 1: Engines Powering L2+ to L4
CES 2020 Engines Powering L2+ to L4
Slide 2: Mobileye in Numbers
Mobileye in Numbers EyeQ Shipped Over 54M EyeQs shipped to date 46 % CAGR In EyeQ shipping since 2014 47 Running Programs Globally across 26 OEMs In 2019: 33 Design Wins 28M units over life 16 Product Launches Industry first 100° camera with Honda 4 high-end L2+ wins with 4 VW high-volume launch (Go
Slide 3: Mobileye Solution Portfolio
Mobileye Solution Portfolio Covering the Entire Value Chain 2025 2022 L3/4/5 Today L4/ L5 Passenger cars Mobility-as-a-Service Today L2+/ L2++ Consumers Autonomy Full Autonomy L1-L2 ADAS SDS to OEMs Conditional Autonomy Full-Service provider-owning Scalable proposition for Chauffeur mode Driver assi
Slide 4: The ADAS Segment Evolution
Visual perception The ADAS Segment Evolution
Slide 5: L2+ - The Next Leap in ADAS
L2+ - The Next Leap in ADAS The opportunity L2+ global volume expectation (M) Source: Wolfe research, 2019 L2+ common attributes 13 Multi-camera sensing HD maps R CAG Multi-camera front 63% sensing to full surround 3.6 L2+ functionalities range from L2+ - significant added value in comfort, not only
Slide 6: Mobileye Scalable Solution for L2+
Mobileye Scalable Solution for L2+ Camera-based 360° sensing is the enabler for the next leap in ADAS Lean compute 360° cameras sensor suite REM™ HD Maps Driving Policy layer platform Affordability allows mass First in the industry to offer: Entire system running on 2x RSS-based adoption in ADAS Eye
Slide 7: L2+ Business Status
L2+ Business Status More than 70% of the L2+ systems running today are powered by Mobileye’s technology For example: Nissan ProPilot ™ 2.0 VW Travel Assist ™ Cadillac Supercruise ™ BMW KaFAS 4 Additional 12 active programs with L2+ variants and 13 open RFQs
Slide 8: Next Generation ADAS
Next Generation ADAS Unlocking “Vision Zero” with RSS for Humans Drivers ADAS Today ADAS Future Potential AEB, LKA | Emergency driven AEB, LKA, ESC | All in one ESC/ ESP | Prevention driven Prevention driven system Application of brakes Formal Guarantees longitudinally & laterally Vision Zero RSS Je
Slide 9: Under the Hood of Mobileye's Computer Vision
Under the Hood of Mobileye’s Computer Vision
Slide 10: The Motivation Behind Surround CV
Full stack camera only AV The The Motivation goal 10−4MTBF for sensing mistake leading to RSS violation (per hour of driving) Behind Surround CV Why ~ 10−4 Humans probability of injury per hour of driving ~ 10−6 Humans probability of fatality per hour of driving ~10−7 The sensing system desired MTBF
Slide 11: Mobileye's Sensing has Three Demanding Customers
Mobileye’s Sensing has Three Demanding Customers Sensing state for Driving Policy Smart agent for harvesting, ADAS products working under the strict role of localization and dynamic everywhere and at all conditions independency and redundancy. information for REM based map on millions of vehicles
Slide 12: Comprehensive CV Environmental Model
Comprehensive CV Environmental Model Four General Categories Road Semantics Road-side directives (TFL/TSR), on-road directives (text, arrows, stop-line, crosswalk) and their Driving Path (DP) association.. Road Boundaries Any delimiter/ 3D structure/ semantics of the drivable area, both laterally (F
Slide 13: Redundancy in the CV Subsystem
Redundancy in This creates internal redundancy layers for both detection and measurements: the CV Subsystem Appearance Geometry Appearance Geometry In order to satisfy an MTBF based based based based of 10−4 hours of driving of the CV-Sub-system: Multiple independent CV engines overlap in their cove
Slide 14: Object Detection
Object Detection Generated and solidified using 6 different engines Scene segmentation (NSS) VIDAR 3DVD Detection Wheels Full image detection Top View FS
Slide 15: 2D to 3D Process
2D to 3D Process Generated and solidified using 4 different engines VIDAR Visual road model Measurements 2D 3D Range Net Map world model (REM)
Slide 16: Full Image Detection
Full Image Detection Two dedicated 360-stitching engines for completeness and coherency of the unified objects map: Vehicle signature Very close (part-of) vehicle in field of view: face & limits Front right cam Rear right cam Front right cam Rear right cam
Slide 17: Left sector - FID in action
ACTIVE Left sector - FID in action
Slide 18: Inter-cameras tracking
Inter-cameras tracking Object signature network
Slide 19: Range Net
Range Net Metric Physical Range estimation dramatically improve measurement quality using novel methods Range Range net output Traditional classifier output Frame
Slide 20: Pixel-level Scene Segmentation
Redundant to the object-dedicated networks Pixel-level Scene Catches extremely-small visible fragments of road users; Segmentation Used also for detecting “general objects”.
Slide 21: Surround Scene Segmentation with Instance
Surround Scene Segmentation with Instance Front left Front right Rear left Rear right
Slide 22: Road Users – open door
Road Users – open door Uniquely classified , as it is both extremely common, critical, and of no ground intersection
Slide 23: Road Users - VRU
Road Users - VRU Baby strollers and wheel chairs are detected through a dedicated engine on top of the pedestrians detection system
Slide 24: Parallax Net
Parallax Net Parallax Net engine provides accurate structure understanding by assessing residual elevation (flow) from the locally governing road surface (homography).
Slide 25: VIDAR
VIDAR “Visual Lidar”: DNN-based Multi-view Stereo Redundant to the appearance and measurement engines handling “rear protruding” objects – which hover above the object’s ground plane.
Slide 26: VIDAR Input
VIDAR Input Front left Main Front right Rear left Parking left Parking right Rear right
Slide 27: VIDAR Output
VIDAR Output DNN based multi-view stereo
Slide 28: VIDAR Output
VIDAR Output DNN based multi-view stereo
Slide 29: Road Users from VIDAR
Road Users from VIDAR Leveraging Lidar Processing Module for Stereo Camera Sensing – “VIDAR” Dense depth image from VIDAR High-res Pseudo-Lidar Upright obstacle ‘stick’ Object detection extraction
Slide 30: Obstacle Classification
Obstacle Visual perception Classification Obstacle classification e.g., how to differentiate a double parked car from a traffic jam Using cues from the environment • Behavior of other road users • What’s in front of the object • Object location • Opened door • Emergency lights
Slide 31: Road Users Semantics
Road Users Semantics Head/pose orientation Pedestrians posture/gesture. Vehicle light indicators Emergency vehicle/Personnel classification. Emergency vehicle , light indicators Pedestrian understanding
Slide 32: Road Users Semantics
Road Users Semantics Pedestrian Gesture Understanding Come closer You can pass Stop! On the phone
Slide 33: ACTIVE @
ACTIVE 12 @
Slide 34: Slide 34
Slide 35: Slide 35
Slide 36: The full unedited 25min ride is available on YouTube
The full unedited 25min ride is available at Mobileye’s YouTube Channel https://www.youtube.com/watch?v=hCWL0XF_f8Y&t=15s
Slide 37: REM Mapping and Data
REM Mapping and Data
Slide 38: REM Process
REM Process 1. Harvesting 2 3. Aggregation 4 5. Localizing Generating HD Collecting road and Map tile distributed Localizing the car Anonymizing and crowdsourced landmarks through to the car within 10cm accuracy encrypting REM data RoadBook for EyeQ-equipped vehicles in the road book. autonomous dri
Slide 39: REM Volumes
REM Volumes Harvesting agreements with 6 major car makers Harvesting: Over 1M Harvesting vehicles in EU by 2020 Harvesting volumes Over 1M Harvesting vehicles in US by 2021 14M Collecting 6 million km per day from serial production vehicles 12M such as: 10M Volkswagen Golf, Passat, BMW 5 series, 3 s
Slide 40: REM-data Aggregation
REM-data Aggregation RSD Coverage Global Snapshot
Slide 41: REM Milestones
REM Milestones Mapping all of Europe by Q1 2020 Mapping most of the US by end 2020
Slide 42: REM for Autonomous Driving
REM for Autonomous Driving Already operational and is proving to be a true segment game changer For roads above 45 Mph Maps created in a fully automated process TODAY Contains all static, dynamic, and semantic layers to allows fully autonomous drive 45 Mph For roads below 45 Mph Semi-automated proce
Slide 43: REM in China
REM in China Data harvesting agreements in China complying with regulatory constraints Strategic collaboration with SAIC Motor for REM data harvesting Accelerate the AV development for passenger vehicles in China Harvesting data in China as part of a collaboration with NIO on L4 synergy for Robotaxi
Slide 44: The Smart Cities Opportunity
The Smart Cities Opportunity
Slide 45: Mobileye Data Services
Mobileye Data Services Product Portfolio Infrastructure Pavement Condition Dynamic Mobility Asset Inventory Assessment Mapping Automated, AI-powered Automated surveying & Near real-time & historical road asset surveying assessment of road conditions data on movement in the Efficient asset management
Slide 46: Infrastructure Asset Inventory
Infrastructure Asset Inventory
Slide 47: Pavement Conditions Assessment
Pavement Conditions Assessment 5 levels score 0 – Excellent conditions - requires no repair Road Conditions Score – Poor (5)
Slide 48: Slide 48
Slide 49: Pavement Conditions Assessment
Pavement Conditions Assessment Cracks and potholes harvester in action Road Conditions Score – Poor (5)
Slide 50: Pavement Conditions Assessment
Pavement Conditions Assessment Cracks and potholes harvester in action Road Conditions Score – Poor (5)
Slide 51: Pavement Conditions Assessment
Pavement Conditions Assessment Cracks and potholes harvester in action
Slide 52: RSS Driving Policy and Driving Experience
Visual perception RSS Driving Policy and Driving Experience
Slide 53: The Driving Policy Challenge
The Driving Policy Challenge • Do we allow an accident due to a “lapse of judgement” of Driving Policy? • Should the occurrence of “lapse of judgement” be measured statistically? Safety is a technological layer living outside of Machine Learning. It is like “Ethics” in AI - a set of rules. • It all
Slide 54: What is RSS?
The Method What is RSS? A formal model for safety, that provides Defining reasonable boundaries on the mathematical guarantees for the AV to never 01 behavior of other road users cause an accident Within the boundaries specified by RSS, one 02 must always assume the worst-case behavior of other agen
Slide 55: Ego car A is following car B on a single-lane straight road
For Example Ego car A is following car B on a single-lane straight road A B The Goal Efficient policy for A that guarantees not to hit B in the worst-case The Implementation Safe distance for A to not hit B in the worst-case – under a reasonable assumption on V b max brake The Policy Define Dangerou
Slide 56: More Complex Situations
More Complex Situations RSS sets the boundaries of reasonable assumptions for all driving scenarios What is reasonable to assume on B in the scenarios below Multiple Geometry Lateral Maneuvers Occlusions If B can brake at Bmin brake If B can brake at B lat min brake , Assuming the max velocity of B
Slide 57: RSS In Summary
The RSS adheres to 5 principles: In Summary Assuming cooperative behavior on the roadway is the key for drivability and 01 Soundness - full compliance with common sense of human driving “human-like” driving 02 Completeness - covering all driving scenarios by always assuming the worst case under the
Slide 58: Industry Acceptance
Industry Acceptance The RSS is gaining global acceptance as an Automated Vehicle Safety Standard Previously announced adoptions Safety First for Automated Driving IEEE to define a formal model for of RSS: (SaFAD) AV safety with Intel-Mobileye leading the workgroup Companies involved are: BMW, Daimle
Slide 59: Industry Acceptance
Industry Acceptance The RSS is gaining global acceptance as an Automated Vehicle Safety Standard China ITS Industry Alliance (C-ITS) to formally approve an RSS-based standard The standard, “Technical Requirement of Safety Assurance of AV Decision Making”, has been released to public and will take ef
Slide 60: The Path to Becoming an End-to-End Mobility-as-a-Service Provider
The Path to Becoming an End-to-End Mobility-as-a-Service Provider
Slide 61: MaaS Business Status
MaaS Business Status Mobileye is forging driverless MaaS as a near term revenue-generating channel > The JV to bring robotaxi MaaS to Tel- Aviv is officially signed > RATP and Mobileye partnered with the City of Paris to deploy a driverless mobility solution > Deploying and testing in Tel-Aviv durin
Slide 62: Our Self-Driving-System HW Generations
Our Self-Driving-System HW Generations EPM 6 EPM 59 > Deployment in 2023 > Single EQ6H to support E2E EPM 52 functionality > Deployment in Q2 2020 > Additional EQ6H FOP Up to 220 TOPs > Up to 6x EQ5H > In deployment > Additional 2-3 for FOP > Up to 2x EQ5H > E2E support in all aspects- fusion, polic
Slide 63: Main Takeaways
Main Takeaways L2+ a growing new category for ADAS where Surround-CV unlocks considerable value at 01 volume production cost. Realization of (safe) L4 and unlocking the full potential of L2+ requires Surround-CV at a 02 standalone (end-to-end) quality L2+ required HD-map-everywhere at growing use-ca
Slide 64: Thank You!
Thank You!