- Dec 25, 2013
- 3,899
- 193
- 106
witeken... I think you put this in the wrong place...? Or it is a bit to complicated to apply...?
AT Moderator
Bartman39
Mobileye Overview
Mobileye, an Intel company (Intel’s Automated Driving Group was merged into Mobileye), is the world leader in ADAS. Intel bought Mobileye for $15.3B and the acquisition closed in August 2017.
Financials
Quarter / Revenue ($M) / Operating Income ($M)
2016: 358 / 122
660 employees
Q3'17: 82 / 39
Q4'17: 128 / -12 (resources transfer from IOTG’s ADG)
Q1'18: 151 / 10
2014
2.7M EyeQ shipped.
2015
4.4M EyeQ shipped.
2016
6M EyeQ shipped.
12 design wins.
2017
8.7M EyeQ shipped (24M to date).
30 ADAS design wins with 27 OEMs (2.5x 2016) and ~70 car models.
Design wins for L2+ and L3 with 11 automakers, who collectively make up more than 50% of the worldwide vehicle production (MSS).
2018
15 program launches (2.5x 2016) with 14 OEMs:
In Q1’18, Intel’s autonomous vehicle test fleet started operating in Israel, with other geographies following in the future.
Won a high-volume EyeQ5 design win.
Launch of Mobileye 8 Connect standalone, aftermarket REM system with EyeQ4 and modem.
2019
Program launches:
EyeQ5 series production.
Stetegy
Mobileye has a camera-first approach to AV. Mobileye’s strategy is to build onto its existing ADAS market leadership. They use a combination of in-house EyeQ chips with Atom SoC. First Level 4-5 capable system will be EyeQ5 starting in 2020.
With their REM initiative, they want to crowdsource HD-maps. It has 10cm localization accuracy.
For safety and also for Driving Policy / Planning, they implement their RSS.
EyeQ5 is used for sensing (vision) and planning (fusion and policy), while Atom is used for acting. Radar and Lidar is also used.
Sensing:
Vehicles, pedestrians, cyclists, traffic signs, traffic lights
RSS (Responsibility-Sensitive Safety):
A mathematical model to determine who is responsible for an accident. The goal of autonomous driving should be that self-driving cars never cause accidents and respond properly to mistakes of drivers.
EyeQ
EyeQ3
L3 Production
AT Moderator
Bartman39
Mobileye Overview
Mobileye, an Intel company (Intel’s Automated Driving Group was merged into Mobileye), is the world leader in ADAS. Intel bought Mobileye for $15.3B and the acquisition closed in August 2017.
Financials
Quarter / Revenue ($M) / Operating Income ($M)
2016: 358 / 122
660 employees
Q3'17: 82 / 39
Q4'17: 128 / -12 (resources transfer from IOTG’s ADG)
Q1'18: 151 / 10
2014
2.7M EyeQ shipped.
2015
4.4M EyeQ shipped.
2016
6M EyeQ shipped.
12 design wins.
2017
8.7M EyeQ shipped (24M to date).
30 ADAS design wins with 27 OEMs (2.5x 2016) and ~70 car models.
Design wins for L2+ and L3 with 11 automakers, who collectively make up more than 50% of the worldwide vehicle production (MSS).
6 program launches including:
- GM Super Cruise
- Audi zFas
- Nissan ProPilot
15 program launches (2.5x 2016) with 14 OEMs:
- 4 of which being Chinese
- 4 of which with EyeQ4
- 2 with Trifocal camera config
- All have high-end, full-feature bundles
- 3DVD
- Traffic lights detection and recognition
- Advanced road features: semantic free space, holistic path prediction
- REM: road experience management, i.e. crowdsource mapping, expects to have 2 million on the road by the end of the year
In Q1’18, Intel’s autonomous vehicle test fleet started operating in Israel, with other geographies following in the future.
Won a high-volume EyeQ5 design win.
Launch of Mobileye 8 Connect standalone, aftermarket REM system with EyeQ4 and modem.
2019
Program launches:
- 12 with EyeQ4 (up from 4 in ’18)
EyeQ5 series production.
Stetegy
Mobileye has a camera-first approach to AV. Mobileye’s strategy is to build onto its existing ADAS market leadership. They use a combination of in-house EyeQ chips with Atom SoC. First Level 4-5 capable system will be EyeQ5 starting in 2020.
With their REM initiative, they want to crowdsource HD-maps. It has 10cm localization accuracy.
For safety and also for Driving Policy / Planning, they implement their RSS.
EyeQ5 is used for sensing (vision) and planning (fusion and policy), while Atom is used for acting. Radar and Lidar is also used.
Sensing:
Vehicles, pedestrians, cyclists, traffic signs, traffic lights
- Mature technology as part of ADAS evolution
- Drivable paths though redundancy of both sensing (parsing) and HD-map (REM)
- Parsing for L4 is an open problem
- REM for automating HD-map construction (10kb/km)
RSS (Responsibility-Sensitive Safety):
A mathematical model to determine who is responsible for an accident. The goal of autonomous driving should be that self-driving cars never cause accidents and respond properly to mistakes of drivers.
EyeQ
EyeQ3
- 0.25 TOPS / 3W (40nm)
- In series production since Nov’14
- 1.5 TOPS / 15W (28nm)
- NVIDIA Parker: 1.5 TOPS / 15W
- In series production since Mar’18
- 4 OEMs in ‘18
- 12 OEMs in ‘19+
- 24 TOPS / 10W (7nm)
- NVIDIA Xavier: 30 TOPS / 30W
- Silicon in Aug’18
- In series production from Mar’20
- Design wins: 4 OEMs from ‘20+
- Includes CPU cluster (Atom), two deep learning accelerators (“EyeQ5”), computer vision processors and “multithread accelerator more versatile than a GPU”
- Atom
L3 Production
- Series development 2019+
- Audi, BMW, Fiat-Chrysler, Honda, NIO, Nissan, SAIC
- Strategic partnerships 2020+
- BWM, Fiat-Chrysler, SAIC, NIO
- 3x OEMs in ongoing sourcing decisions
- L4 turnkey solution with Delphi (Aptiv)
Last edited by a moderator:
Facebook
Twitter