際際滷

際際滷Share a Scribd company logo

TU Automotive Osram Presentation Final

Rajeev Thakur, P.E.
Rajeev Thakur, P.E.

This document discusses infrared sensors for autonomous vehicles. It provides an overview of OSRAM, a leader in LED and laser technologies. It then discusses the sensing challenges for autonomous vehicles, including range, resolution, field of view, and computational challenges. It reviews sensor technologies like LIDAR, infrared cameras, and RADAR, and their applications in autonomous driving. It also explores concepts for low-cost LIDAR solutions using flash LIDAR and scanning LIDAR technologies.

1 of 23
Downloaded 510 times
www.osram-os.com Infrared sensors for ADAS and beyond LIDAR / Infrared camera Rajeev Thakur| 4th October 2016| Novi Light is OSRAM
IR for ADAS | OS IR NA MK | R.Thakur TU Automotive ADAS & Autonomous | 10/04/2016 2 Content Page 1. OSRAM Overview 03 2. Sensing challenges 06 3. LIDAR 12 4. Infrared Camera 19 5. Sensor Fusion 21 6. Collaboration & Competition in the self driving car business 22
IR for ADAS | OS IR NA MK | R.Thakur TU Automotive ADAS & Autonomous | 10/04/2016 3 Global Market Leader in LED & Laser LIDAR Infrared Lasers - AEB Consumer Industry General Lighting Laser front light Xenon front light Laser front light OLED rear light Matrix LED light Automotive Lighting Source: OSRAM, excluding LAMPS 1) at the end of the fiscal year 2) countries where OSRAM had operations at the end of the fiscal year Employees1) : 20,300 Worldwide Presence2) : >120 countries Revenue1): 3,571.9 m
IR for ADAS | OS IR NA MK | R.Thakur TU Automotive ADAS & Autonomous | 10/04/2016 4 Key Automotive Trends Exterior , Interior & IR Safety Design Visualization & Connectivity Comfort & Safety Key Automotive Trends ExteriorInterior 袖AFS High Luminance LEDs Dynamic Lighting Projection HuD Full Digital Cluster LED Applications New LED Development Display Portfolio HuD Portfolio BLU Displays High ResolutionADB/Matrix ProjectionUltra slim HL LIDAR / ADB Gesture Wireless Connectivity Driver Monitoring Night Vision
IR for ADAS | OS IR NA MK | R.Thakur TU Automotive ADAS & Autonomous | 10/04/2016 5 OSRAM Infrared & Laser Automotive Applications Existing Applications / New Applications Rain Light Tunnel Sensors Ambient light sensors for dimming and illumination Dashboard Car radio Displays Immobilizer Steering wheel angle sensor Blue Lasers for Headlamps Driver monitoring Gesture Recognition IRED based Night vision Blind spot detection Lane departure warning Family Entertainment System LIDAR sensing AEB & ADAS Laser HuD
IR for ADAS | OS IR NA MK | R.Thakur TU Automotive ADAS & Autonomous | 10/04/2016 6 Sensing Needs for Vehicle Environment Continue Brake Steer to safety Prepare for crash Steering is best option Cannot avoid crash Braking is best option NO YES Current Sensing Range Upper limit (For Large Objects) RADAR : 50 - 250m Camera : 50 - 70m LIDAR : 50 200m Braking Distance / Minimum Sensing Range (Assumptions : Dry Road, 袖 = 0.7, 1 sec reaction time) @100mph (161kph/44.7m/s) : > 190 meters @74mph (119kph,33 m/s) : > 112 meters @45mph (72.4kph/20.1 m/s) : > 50 meters @25mph (40kph/11.1 m/s) : > 20 meters If the closing speed is less than ~ 45mph , current sensing technology can mitigate collision to large objects under normal daylight dry conditions (distance < 70 meters) Challenge 1 : Sensing Range Is projected vehicle trajectory safe for next XX meters? Calculate Time to Crash
IR for ADAS | OS IR NA MK | R.Thakur TU Automotive ADAS & Autonomous | 10/04/2016 7 Sensing Needs for Vehicle Environment Who decides ? Ignore objects at own risk .. Challenge 2 : Angular Sensing Resolution Standard object list for detection does not exist (ignore / standardize with risk) LIDAR is capable of < 0.5属 resolution at > 100 m (with small form factor) RADAR size for 0.5属 resolution not practical (~ 0.5 m for 76 GHz RADAR) Camera range needs to improve & image quality in lowlight (or infrared) 1 Bosch Multi Purpose Camera (MPC2) , 1280 x 960 pixels, 50属 HFOV, 28属 VFOV 2 Velodyne VLP16 (0.1属 0.4属) 3 RADAR equation What objects should be detected to avoid collision ? Typical Angular Resolution 1 Camera : 25 pixels / 属 2 LIDAR : 0.3属 3 RADAR : 2.6属 (76 GHz, 10 cm aperture) 1.5m 0.25m 0.4m 0.2m 0.1m 0.2m tire piece potholedog Resolution Size (m) 1属 @ 100 m = > 1.7 m 1属 @ 200 m = > 3.4 m 0.1属 @ 200 m = > 0.4 m
IR for ADAS | OS IR NA MK | R.Thakur TU Automotive ADAS & Autonomous | 10/04/2016 8 Sensing Needs for Vehicle Environment Challenge 3 : Field of View Winding Roads Need Wide FOV Traffic Lights & Overhead Signs Need High FOV Up & Down Ramps Need High FOV FOV Field Of View
IR for ADAS | OS IR NA MK | R.Thakur TU Automotive ADAS & Autonomous | 10/04/2016 9 Sensing Needs for Vehicle Environment Challenge 4 : Computational challenges Time Needed = Sensing time + Reaction Time + Safety Margin Sensing Time per Sensor = (Points/Frame x # of Frames in Buffer x compute time/point) Finer resolution => More data points => more time (or faster computation) Redundancy / sensor fusion needed prior to reaction Reaction Time = (Human delay) + latency in steering or braking system Safety Margin : To accommodate environment conditions (road / temperature) , sensing and computational delays and tolerances 1 Can we unify monocular detectors for autonomous driving by using the pixel-wise semantic segmentation of CNNs? : Department of Electronics, UAH. Alcala de 卒 Henares, Madrid, Spain ; IEEE Workshop in June 2016 on Intelligent Vehicles 1 Sensor / Processor / environment / algorithm .. affects computational time and accuracy
IR for ADAS | OS IR NA MK | R.Thakur TU Automotive ADAS & Autonomous | 10/04/2016 10 Sensing Needs for Vehicle Environment Other Challenges Form factor small and compatible to vehicle styling & materials Increasing noise from surrounding vehicle RADAR/LIDAR .. Dealing with satellite signal / GPS loss in real time Harsh environment Snow/rain/dust/dirt/shock and vibrations Power / EMC / ESD / .. Service Cost .. Tremendous innovation currently in sensing field OSRAM working with multiple startups / Tier1 and OEMs
IR for ADAS | OS IR NA MK | R.Thakur TU Automotive ADAS & Autonomous | 10/04/2016 11 RADAR / Camera / LIDAR Comparison Sensor Typical Range Horizontal FOV Vertical FOV 2020 Price Range Comments 24 GHz RADAR 60 m 1 56属 1 ~ 賊 20属 < $100 USA Bandwidth 100 -250 MHz 2 Robust for Rain/snow ; People Detection / Angular Resolution 77 GHz RADAR 200 m 1 18属 1 ~ 賊 5属 < $100 USA Bandwidth 600 MHz 2 Robust for Rain/snow ; People Detection / Angular Resolution Front Mono Camera 50 m 1 36属 1 ~ 賊 14 属 < $100 Versatile Sensor (Applications) Limited depth perception ; affected by rain / fog Needs illumination (Visible/IR) LIDAR (Flash) 75 m 140属 ~ 賊 5属 < $100 Concerns for Rain/Snow; Good reflection off people w/ angular resolution Range & S/N limited by eye safety LIDAR (Scanning) 200 m 360属 ~ 賊 14属 < $500 Concerns for Rain/Snow; Typically higher price for angular resolution Range & S/N limited by eye safety 1 : Vehicle-to-Vehicle Communications: readiness of V2V technology for application DOT HS 812014 ; Table V-7 2 : Millimeter Wave Receiver concepts for 77 GHz automotive radar in silicon Germanium Technology D.Kissenger (SpringerBriefs 2012) False positives Nuisance to consumer Turns feature off (if possible) False negatives did not meet spec / expectations Optimum combination of sensors will be a learning process Sensor fusion can be done at best on common subset in field of view
IR for ADAS | OS IR NA MK | R.Thakur TU Automotive ADAS & Autonomous | 10/04/2016 12 Flash LIDAR Design Overview start stop t Laser Photodiode Array Emitter Lens Receiving Lens Target FOV Working principle : Laser beam spread into field of view and received on photodiode array. Range determined by eye safe laser power , resolution determined by number of photodiode pixels Why use : Mature low cost sensor that can be integrated into headlamp / Tail lamp / behind windshield / .. Range : ~ 30 - 60 meters @ 24 HFOV Resolution : 3 deg or less Wavelength : 905 nm has proven sufficient for short range Laser : OSRAM lasers with peak power 75 120W , with & without drivers , bare die to SMT w/ < 5ns pulse width (2019 SOP) , also with multiple emitters in one SMT package Photodiode : OSRAM PD array concepts of various sizes planned for SOP 2018 Why not as popular as RADAR yet in NAFTA?: 2019 NCAP upgrade will incentivize market , more room for creativity lower cost than RADAR ; market waiting for low cost scanning LIDAR R Distance C speed of light t time between start - stop of pulse HFOV Horizontal Field Of View SOP Start of Production PD Photo Diode
IR for ADAS | OS IR NA MK | R.Thakur TU Automotive ADAS & Autonomous | 10/04/2016 13 LIDAR Head Lamp Integration LeddarTech Concept Leddartech Video link
IR for ADAS | OS IR NA MK | R.Thakur TU Automotive ADAS & Autonomous | 10/04/2016 14 LIDAR Tail Lamp Integration LeddarTech Concept
IR for ADAS | OS IR NA MK | R.Thakur TU Automotive ADAS & Autonomous | 10/04/2016 15 Phantom Intelligence : Guardian Flash LIDAR The Guardian BY PHANTOM INTELLIGENCE Fully customizable 2x8 Pixels (1x16 also available) Field of View 9属x36属 Customizable up to 2属x120属 Range limited to 30 meters (for cost optimization) Connectivity: USB, CAN, GPIO Programmable alarms/triggers Power Consumption less than 3 Watts Laser Output of 70 Watts Eye Safe (Class 1M) Price: ~ 100$ in 10k units volume production Engineering Samples December 2016 AWL Video link YOU CAN AFFORD THE SAFEST JOURNEY !
IR for ADAS | OS IR NA MK | R.Thakur TU Automotive ADAS & Autonomous | 10/04/2016 16 LIDAR Low Cost Concept (Reference Design) 30m range ~1cm accuracy 16 pixel array 24属H x 6属V FOV : 2 x 8 array (3属 x 3属per pixel) Arrangement of pixels and field of view can be customized in future products. Multiple targets in each pixel can be resolved Targets ~1m apart (range) can be separated Differentiating through performance, small size, scalability, and low power consumption No moving (scanning) parts Sun blinding can affect no more than a single pixel Estimated BOM ~ $25 @ High Volume Functional sample Q1 2017 Target SOP 2019 Distance (m) Area (m族) 1 0.003 2 0.011 5 0.069 10 0.274 20 1.097 30 2.469 Field Of View Per Pixel FOV Horizontal Field Of View SOP Start of Production BOM Bill of Material (For Hard Ware)
IR for ADAS | OS IR NA MK | R.Thakur TU Automotive ADAS & Autonomous | 10/04/2016 17 Scanning LIDAR Technologies Mechanical - Velodyne Principle: Matched pair of laser & detectors rotating with a motor at 5 to 20 Hz Range : 200 m (VLP 32A) Resolution : 0.1 - 0.4属 (VLP 16) Vertical FOV: 28属 (VLP 32A) Price Target : < $500 ~ 2020 Pro : Proven technology Con: Mechanical integration / price Principle: Laser scanned with OPA (& received on SPAD array ) Range : > 150 m Resolution : 0.1属 FOV: 120属 (HFOV & VFOV ; S3) Price Target : <$100 ~ 2020 Pro : small size (1 x 1.5 , S3-Qi) Con: OPA scanning is relatively new technology 1 Velodyne.com 2 Quanergy.com 3 Innoluce.com OPA - Quanergy MEMS Innoluce Principle: Laser scanned with 1D MEMS Mirror (& received on APD array ) Range : > 200 m Resolution : < 0.5属 HFOV: 80属 VFOV: 16属 Price Target : <$100 ~ 2020 Pro : MEMS scanning is proven Con: Working demo not shown yet MEMS Micro Electro Mechanical Systems APD Avalanche Photo Diode OPA Optical Phase Array SPAD Single Photon Avalanche Diodes FOV Field Of View
IR for ADAS | OS IR NA MK | R.Thakur TU Automotive ADAS & Autonomous | 10/04/2016 18 MEMS LIDAR Innoluce OSRAM : Concept Proof Principle: 1 25W OSRAM laser scanned with 1 Innoluce MEMS Mirror and received on an APD array Range : ~ 60 m Resolution : 0.1属 Horizontal and 0.2属 Vertical HFOV: 10属 VFOV: 3属 Next Steps : Show progressively improved reference design demos in next few months Targets : >200 m/car ; > 60m/Ped ; 80属 HFOV ; 16属 VFOV ; < 0.5属 Resolution (High power multiple emitter lasers) High resolution concept proof MEMS LIDAR Video Link
IR for ADAS | OS IR NA MK | R.Thakur TU Automotive ADAS & Autonomous | 10/04/2016 19 Infrared Camera - Interior Mature applications transitioning to mainstream Why IR Camera : Works in day & night without visible illumination Moving to Mainstream : Driver monitoring (Drowsy/Distracted) Catching speed : Gesture recognition Mobile to Automotive : Iris recognition Technology frontiers: NIR sensitivity (15 35%), > 2Mp Global shutter , increasing IRED o/p & efficiency Concern/Tradeoffs: Privacy Vs App. value , Redglow (850 940 nm) Future applications : Optimum airbag deployment, Mood lighting .. Driver Monitoring 1 Deltaid.com Gesture Control Iris Recognition 1
IR for ADAS | OS IR NA MK | R.Thakur TU Automotive ADAS & Autonomous | 10/04/2016 20 Infrared Camera - Exterior Forward Camera Surround View Camera Rearview Camera Cars have to be autonomous at night also Cameras need to work with IR also .. Why IR Exterior Camera : Need to see adjacent lanes at night w/o visible light Whats the problem : Visible cameras block IR for better image / use of color information Options : Use mechanical or SW filter to switch between IR & visible spectrums Challenges : Modify camera / Illumination / SW for wider FOV and range Things to watch out for : Laser beam headlamps (Dynamic range of oncoming camera) / Use of matrix beam lighting (adaptive beams ..) IR emitters in headlamp
IR for ADAS | OS IR NA MK | R.Thakur TU Automotive ADAS & Autonomous | 10/04/2016 21 Sensor Fusion Fuse Information of high quality which overlap Objective of Sensor Fusion : Determine environment around vehicle trajectory with enough resolution, confidence and speed - to navigate efficiently. Object_list RADAR Camera LIDAR Sensor Fusion Car@150m Dont See it (Noise) Not_Classified@100 m & low light Evaluate TTC & brake if unresolved ? @50m Person on bicycle Not classified Dont see it (Noise) Brake or ignore ? Potholes & stuff What can be safely ignored ? Object Identification & Classification in range & FOV of interest must be comparable. LIDAR + Camera fusion potentially better (Due to angular resolution) Camera improvements : Range (~ 70 m); speed (30 60 Hz) & Low light sensitivity
IR for ADAS | OS IR NA MK | R.Thakur TU Automotive ADAS & Autonomous | 10/04/2016 22 Collaboration & Competition - Self Driving Cars Why collaborate : Need NHTSA support Regulations / Testing / Infrastructure / .. Combine R&D resources & strengths Be / be with a technology leader to gain market share Why Compete? Branding / Technology Leadership (Intangible $ Value) ADAS technology has shown real market value ($1500 - $3000/car) Prepare for future market changes (Self driving cars occupy significant share) What more could/should be done ? Use Silicon Valley playbook more open source development Example : Provide raw data from all sensors in a drive ; show me object identification / classification & tracking .. (Buy the best solution..) Make sensor requirements and roadmap open Small startups have very creative solutions & fast development Why be more open ? 1 year after a new gadget is shown 3 more appear next year (Benefit/Cost) Will enable faster development of SDC technology for community & save lives !
www.osram-os.com Thank you ! Contact : Rajeev.Thakur@osram-os.com

More Related Content

TU Automotive Osram Presentation Final

  • 1. www.osram-os.com Infrared sensors for ADAS and beyond LIDAR / Infrared camera Rajeev Thakur| 4th October 2016| Novi Light is OSRAM
  • 2. IR for ADAS | OS IR NA MK | R.Thakur TU Automotive ADAS & Autonomous | 10/04/2016 2 Content Page 1. OSRAM Overview 03 2. Sensing challenges 06 3. LIDAR 12 4. Infrared Camera 19 5. Sensor Fusion 21 6. Collaboration & Competition in the self driving car business 22
  • 3. IR for ADAS | OS IR NA MK | R.Thakur TU Automotive ADAS & Autonomous | 10/04/2016 3 Global Market Leader in LED & Laser LIDAR Infrared Lasers - AEB Consumer Industry General Lighting Laser front light Xenon front light Laser front light OLED rear light Matrix LED light Automotive Lighting Source: OSRAM, excluding LAMPS 1) at the end of the fiscal year 2) countries where OSRAM had operations at the end of the fiscal year Employees1) : 20,300 Worldwide Presence2) : >120 countries Revenue1): 3,571.9 m
  • 4. IR for ADAS | OS IR NA MK | R.Thakur TU Automotive ADAS & Autonomous | 10/04/2016 4 Key Automotive Trends Exterior , Interior & IR Safety Design Visualization & Connectivity Comfort & Safety Key Automotive Trends ExteriorInterior 袖AFS High Luminance LEDs Dynamic Lighting Projection HuD Full Digital Cluster LED Applications New LED Development Display Portfolio HuD Portfolio BLU Displays High ResolutionADB/Matrix ProjectionUltra slim HL LIDAR / ADB Gesture Wireless Connectivity Driver Monitoring Night Vision
  • 5. IR for ADAS | OS IR NA MK | R.Thakur TU Automotive ADAS & Autonomous | 10/04/2016 5 OSRAM Infrared & Laser Automotive Applications Existing Applications / New Applications Rain Light Tunnel Sensors Ambient light sensors for dimming and illumination Dashboard Car radio Displays Immobilizer Steering wheel angle sensor Blue Lasers for Headlamps Driver monitoring Gesture Recognition IRED based Night vision Blind spot detection Lane departure warning Family Entertainment System LIDAR sensing AEB & ADAS Laser HuD
  • 6. IR for ADAS | OS IR NA MK | R.Thakur TU Automotive ADAS & Autonomous | 10/04/2016 6 Sensing Needs for Vehicle Environment Continue Brake Steer to safety Prepare for crash Steering is best option Cannot avoid crash Braking is best option NO YES Current Sensing Range Upper limit (For Large Objects) RADAR : 50 - 250m Camera : 50 - 70m LIDAR : 50 200m Braking Distance / Minimum Sensing Range (Assumptions : Dry Road, 袖 = 0.7, 1 sec reaction time) @100mph (161kph/44.7m/s) : > 190 meters @74mph (119kph,33 m/s) : > 112 meters @45mph (72.4kph/20.1 m/s) : > 50 meters @25mph (40kph/11.1 m/s) : > 20 meters If the closing speed is less than ~ 45mph , current sensing technology can mitigate collision to large objects under normal daylight dry conditions (distance < 70 meters) Challenge 1 : Sensing Range Is projected vehicle trajectory safe for next XX meters? Calculate Time to Crash
  • 7. IR for ADAS | OS IR NA MK | R.Thakur TU Automotive ADAS & Autonomous | 10/04/2016 7 Sensing Needs for Vehicle Environment Who decides ? Ignore objects at own risk .. Challenge 2 : Angular Sensing Resolution Standard object list for detection does not exist (ignore / standardize with risk) LIDAR is capable of < 0.5属 resolution at > 100 m (with small form factor) RADAR size for 0.5属 resolution not practical (~ 0.5 m for 76 GHz RADAR) Camera range needs to improve & image quality in lowlight (or infrared) 1 Bosch Multi Purpose Camera (MPC2) , 1280 x 960 pixels, 50属 HFOV, 28属 VFOV 2 Velodyne VLP16 (0.1属 0.4属) 3 RADAR equation What objects should be detected to avoid collision ? Typical Angular Resolution 1 Camera : 25 pixels / 属 2 LIDAR : 0.3属 3 RADAR : 2.6属 (76 GHz, 10 cm aperture) 1.5m 0.25m 0.4m 0.2m 0.1m 0.2m tire piece potholedog Resolution Size (m) 1属 @ 100 m = > 1.7 m 1属 @ 200 m = > 3.4 m 0.1属 @ 200 m = > 0.4 m
  • 8. IR for ADAS | OS IR NA MK | R.Thakur TU Automotive ADAS & Autonomous | 10/04/2016 8 Sensing Needs for Vehicle Environment Challenge 3 : Field of View Winding Roads Need Wide FOV Traffic Lights & Overhead Signs Need High FOV Up & Down Ramps Need High FOV FOV Field Of View
  • 9. IR for ADAS | OS IR NA MK | R.Thakur TU Automotive ADAS & Autonomous | 10/04/2016 9 Sensing Needs for Vehicle Environment Challenge 4 : Computational challenges Time Needed = Sensing time + Reaction Time + Safety Margin Sensing Time per Sensor = (Points/Frame x # of Frames in Buffer x compute time/point) Finer resolution => More data points => more time (or faster computation) Redundancy / sensor fusion needed prior to reaction Reaction Time = (Human delay) + latency in steering or braking system Safety Margin : To accommodate environment conditions (road / temperature) , sensing and computational delays and tolerances 1 Can we unify monocular detectors for autonomous driving by using the pixel-wise semantic segmentation of CNNs? : Department of Electronics, UAH. Alcala de 卒 Henares, Madrid, Spain ; IEEE Workshop in June 2016 on Intelligent Vehicles 1 Sensor / Processor / environment / algorithm .. affects computational time and accuracy
  • 10. IR for ADAS | OS IR NA MK | R.Thakur TU Automotive ADAS & Autonomous | 10/04/2016 10 Sensing Needs for Vehicle Environment Other Challenges Form factor small and compatible to vehicle styling & materials Increasing noise from surrounding vehicle RADAR/LIDAR .. Dealing with satellite signal / GPS loss in real time Harsh environment Snow/rain/dust/dirt/shock and vibrations Power / EMC / ESD / .. Service Cost .. Tremendous innovation currently in sensing field OSRAM working with multiple startups / Tier1 and OEMs
  • 11. IR for ADAS | OS IR NA MK | R.Thakur TU Automotive ADAS & Autonomous | 10/04/2016 11 RADAR / Camera / LIDAR Comparison Sensor Typical Range Horizontal FOV Vertical FOV 2020 Price Range Comments 24 GHz RADAR 60 m 1 56属 1 ~ 賊 20属 < $100 USA Bandwidth 100 -250 MHz 2 Robust for Rain/snow ; People Detection / Angular Resolution 77 GHz RADAR 200 m 1 18属 1 ~ 賊 5属 < $100 USA Bandwidth 600 MHz 2 Robust for Rain/snow ; People Detection / Angular Resolution Front Mono Camera 50 m 1 36属 1 ~ 賊 14 属 < $100 Versatile Sensor (Applications) Limited depth perception ; affected by rain / fog Needs illumination (Visible/IR) LIDAR (Flash) 75 m 140属 ~ 賊 5属 < $100 Concerns for Rain/Snow; Good reflection off people w/ angular resolution Range & S/N limited by eye safety LIDAR (Scanning) 200 m 360属 ~ 賊 14属 < $500 Concerns for Rain/Snow; Typically higher price for angular resolution Range & S/N limited by eye safety 1 : Vehicle-to-Vehicle Communications: readiness of V2V technology for application DOT HS 812014 ; Table V-7 2 : Millimeter Wave Receiver concepts for 77 GHz automotive radar in silicon Germanium Technology D.Kissenger (SpringerBriefs 2012) False positives Nuisance to consumer Turns feature off (if possible) False negatives did not meet spec / expectations Optimum combination of sensors will be a learning process Sensor fusion can be done at best on common subset in field of view
  • 12. IR for ADAS | OS IR NA MK | R.Thakur TU Automotive ADAS & Autonomous | 10/04/2016 12 Flash LIDAR Design Overview start stop t Laser Photodiode Array Emitter Lens Receiving Lens Target FOV Working principle : Laser beam spread into field of view and received on photodiode array. Range determined by eye safe laser power , resolution determined by number of photodiode pixels Why use : Mature low cost sensor that can be integrated into headlamp / Tail lamp / behind windshield / .. Range : ~ 30 - 60 meters @ 24 HFOV Resolution : 3 deg or less Wavelength : 905 nm has proven sufficient for short range Laser : OSRAM lasers with peak power 75 120W , with & without drivers , bare die to SMT w/ < 5ns pulse width (2019 SOP) , also with multiple emitters in one SMT package Photodiode : OSRAM PD array concepts of various sizes planned for SOP 2018 Why not as popular as RADAR yet in NAFTA?: 2019 NCAP upgrade will incentivize market , more room for creativity lower cost than RADAR ; market waiting for low cost scanning LIDAR R Distance C speed of light t time between start - stop of pulse HFOV Horizontal Field Of View SOP Start of Production PD Photo Diode
  • 13. IR for ADAS | OS IR NA MK | R.Thakur TU Automotive ADAS & Autonomous | 10/04/2016 13 LIDAR Head Lamp Integration LeddarTech Concept Leddartech Video link
  • 14. IR for ADAS | OS IR NA MK | R.Thakur TU Automotive ADAS & Autonomous | 10/04/2016 14 LIDAR Tail Lamp Integration LeddarTech Concept
  • 15. IR for ADAS | OS IR NA MK | R.Thakur TU Automotive ADAS & Autonomous | 10/04/2016 15 Phantom Intelligence : Guardian Flash LIDAR The Guardian BY PHANTOM INTELLIGENCE Fully customizable 2x8 Pixels (1x16 also available) Field of View 9属x36属 Customizable up to 2属x120属 Range limited to 30 meters (for cost optimization) Connectivity: USB, CAN, GPIO Programmable alarms/triggers Power Consumption less than 3 Watts Laser Output of 70 Watts Eye Safe (Class 1M) Price: ~ 100$ in 10k units volume production Engineering Samples December 2016 AWL Video link YOU CAN AFFORD THE SAFEST JOURNEY !
  • 16. IR for ADAS | OS IR NA MK | R.Thakur TU Automotive ADAS & Autonomous | 10/04/2016 16 LIDAR Low Cost Concept (Reference Design) 30m range ~1cm accuracy 16 pixel array 24属H x 6属V FOV : 2 x 8 array (3属 x 3属per pixel) Arrangement of pixels and field of view can be customized in future products. Multiple targets in each pixel can be resolved Targets ~1m apart (range) can be separated Differentiating through performance, small size, scalability, and low power consumption No moving (scanning) parts Sun blinding can affect no more than a single pixel Estimated BOM ~ $25 @ High Volume Functional sample Q1 2017 Target SOP 2019 Distance (m) Area (m族) 1 0.003 2 0.011 5 0.069 10 0.274 20 1.097 30 2.469 Field Of View Per Pixel FOV Horizontal Field Of View SOP Start of Production BOM Bill of Material (For Hard Ware)
  • 17. IR for ADAS | OS IR NA MK | R.Thakur TU Automotive ADAS & Autonomous | 10/04/2016 17 Scanning LIDAR Technologies Mechanical - Velodyne Principle: Matched pair of laser & detectors rotating with a motor at 5 to 20 Hz Range : 200 m (VLP 32A) Resolution : 0.1 - 0.4属 (VLP 16) Vertical FOV: 28属 (VLP 32A) Price Target : < $500 ~ 2020 Pro : Proven technology Con: Mechanical integration / price Principle: Laser scanned with OPA (& received on SPAD array ) Range : > 150 m Resolution : 0.1属 FOV: 120属 (HFOV & VFOV ; S3) Price Target : <$100 ~ 2020 Pro : small size (1 x 1.5 , S3-Qi) Con: OPA scanning is relatively new technology 1 Velodyne.com 2 Quanergy.com 3 Innoluce.com OPA - Quanergy MEMS Innoluce Principle: Laser scanned with 1D MEMS Mirror (& received on APD array ) Range : > 200 m Resolution : < 0.5属 HFOV: 80属 VFOV: 16属 Price Target : <$100 ~ 2020 Pro : MEMS scanning is proven Con: Working demo not shown yet MEMS Micro Electro Mechanical Systems APD Avalanche Photo Diode OPA Optical Phase Array SPAD Single Photon Avalanche Diodes FOV Field Of View
  • 18. IR for ADAS | OS IR NA MK | R.Thakur TU Automotive ADAS & Autonomous | 10/04/2016 18 MEMS LIDAR Innoluce OSRAM : Concept Proof Principle: 1 25W OSRAM laser scanned with 1 Innoluce MEMS Mirror and received on an APD array Range : ~ 60 m Resolution : 0.1属 Horizontal and 0.2属 Vertical HFOV: 10属 VFOV: 3属 Next Steps : Show progressively improved reference design demos in next few months Targets : >200 m/car ; > 60m/Ped ; 80属 HFOV ; 16属 VFOV ; < 0.5属 Resolution (High power multiple emitter lasers) High resolution concept proof MEMS LIDAR Video Link
  • 19. IR for ADAS | OS IR NA MK | R.Thakur TU Automotive ADAS & Autonomous | 10/04/2016 19 Infrared Camera - Interior Mature applications transitioning to mainstream Why IR Camera : Works in day & night without visible illumination Moving to Mainstream : Driver monitoring (Drowsy/Distracted) Catching speed : Gesture recognition Mobile to Automotive : Iris recognition Technology frontiers: NIR sensitivity (15 35%), > 2Mp Global shutter , increasing IRED o/p & efficiency Concern/Tradeoffs: Privacy Vs App. value , Redglow (850 940 nm) Future applications : Optimum airbag deployment, Mood lighting .. Driver Monitoring 1 Deltaid.com Gesture Control Iris Recognition 1
  • 20. IR for ADAS | OS IR NA MK | R.Thakur TU Automotive ADAS & Autonomous | 10/04/2016 20 Infrared Camera - Exterior Forward Camera Surround View Camera Rearview Camera Cars have to be autonomous at night also Cameras need to work with IR also .. Why IR Exterior Camera : Need to see adjacent lanes at night w/o visible light Whats the problem : Visible cameras block IR for better image / use of color information Options : Use mechanical or SW filter to switch between IR & visible spectrums Challenges : Modify camera / Illumination / SW for wider FOV and range Things to watch out for : Laser beam headlamps (Dynamic range of oncoming camera) / Use of matrix beam lighting (adaptive beams ..) IR emitters in headlamp
  • 21. IR for ADAS | OS IR NA MK | R.Thakur TU Automotive ADAS & Autonomous | 10/04/2016 21 Sensor Fusion Fuse Information of high quality which overlap Objective of Sensor Fusion : Determine environment around vehicle trajectory with enough resolution, confidence and speed - to navigate efficiently. Object_list RADAR Camera LIDAR Sensor Fusion Car@150m Dont See it (Noise) Not_Classified@100 m & low light Evaluate TTC & brake if unresolved ? @50m Person on bicycle Not classified Dont see it (Noise) Brake or ignore ? Potholes & stuff What can be safely ignored ? Object Identification & Classification in range & FOV of interest must be comparable. LIDAR + Camera fusion potentially better (Due to angular resolution) Camera improvements : Range (~ 70 m); speed (30 60 Hz) & Low light sensitivity
  • 22. IR for ADAS | OS IR NA MK | R.Thakur TU Automotive ADAS & Autonomous | 10/04/2016 22 Collaboration & Competition - Self Driving Cars Why collaborate : Need NHTSA support Regulations / Testing / Infrastructure / .. Combine R&D resources & strengths Be / be with a technology leader to gain market share Why Compete? Branding / Technology Leadership (Intangible $ Value) ADAS technology has shown real market value ($1500 - $3000/car) Prepare for future market changes (Self driving cars occupy significant share) What more could/should be done ? Use Silicon Valley playbook more open source development Example : Provide raw data from all sensors in a drive ; show me object identification / classification & tracking .. (Buy the best solution..) Make sensor requirements and roadmap open Small startups have very creative solutions & fast development Why be more open ? 1 year after a new gadget is shown 3 more appear next year (Benefit/Cost) Will enable faster development of SDC technology for community & save lives !
  • 23. www.osram-os.com Thank you ! Contact : Rajeev.Thakur@osram-os.com