Agenda Introduction Predictive Simulation Simulated Sensors for AV - - PowerPoint PPT Presentation

▶
agenda
SMART_READER_LITE
LIVE PREVIEW

Agenda Introduction Predictive Simulation Simulated Sensors for AV - - PowerPoint PPT Presentation

Aug 14, 2022 195 likes •613 views

P R E D I C T I V E R E N D E R I N G S I M U L AT I O N F O R I N D U S T R I A L C A S E S N ICOLAS D ALMASSO | I NNOVA TION D IRECTOR www.optis-world.com Agenda Introduction Predictive Simulation Simulated Sensors for AV Optis Overview

slide-1
SLIDE 1

www.optis-world.com

P R E D I C T I V E R E N D E R I N G S I M U L AT I O N

F O R I N D U S T R I A L C A S E S

NICOLAS DALMASSO | INNOVA

TION DIRECTOR

slide-2
SLIDE 2

Nicolas Dalmasso | Innovation Director Predictive Simulation for Industrial Applications 26/03/2018 2

Agenda

Introduction ▪ Optis Overview ▪ T

  • wards 0-Physical Prototypes

▪ Product Portfolio Predictive Simulation ▪ Importance of Accurate Engine ▪ Importance of Accurate Input ▪ Importance of Accurate Restitution ▪ Practical Example Simulated Sensors for AV ▪ Physics-based Camera Sensor ▪ Physics-based Lidar Sensor ▪ Physics-based Ultrasonic and Radar Sensors (not covered in this slideset)

slide-3
SLIDE 3

Nicolas Dalmasso | Innovation Director Predictive Simulation for Industrial Applications 26/03/2018 3

OPTIS

OPTIS Overview…

slide-4
SLIDE 4

Nicolas Dalmasso | Innovation Director Predictive Simulation for Industrial Applications 26/03/2018 4

…ALL AROUND THE WORLD

OPTIS Companies…

slide-5
SLIDE 5

Nicolas Dalmasso | Innovation Director Predictive Simulation for Industrial Applications 26/03/2018 5

COMPUTER SCIENCE ACOUSTICS OPTICS

slide-6
SLIDE 6

Nicolas Dalmasso | Innovation Director Predictive Simulation for Industrial Applications 26/03/2018 6

A 0-PHYSICAL PROTOTYPE EXPERIENCE

 COMMUNICATE EASILY  REDUCE ECOLOGICAL FOOTPRINT  ENHANCE CREATIVITY  SAVE TIME AND MONEY  EXPERIENCE TO IMPROVE QUALITY

REPLACING

PHYSICAL PROTOTYPES TO…

slide-7
SLIDE 7

Nicolas Dalmasso | Innovation Director Predictive Simulation for Industrial Applications 26/03/2018 7

A GROWING PORTFOLIO OF BRANDS

LIGHT &

VISION

SIMULATION MATERIAL & COLOR SCANNER PERCEIVEDQUALITY EV

ALUA TION

D

YNAMIC DRIVING

EXPERIENCE HUMAN-INTEGRA

TED

MANUFACTURING REAL-TIME 3D

VISUALIZA TION

SOUNDSIMULA

TION

PERCEPTION

slide-8
SLIDE 8

Name | Job Title Presentation title 26/03/2018 8

How light works in real life vs How light works in SPEOS

Quick lesson of optics

▪ Light carries energy defined by spectrum in straight line, ▪ Path modified according to optical properties of materials it encounters, ▪ Until it reaches the human eye to be interpreted as color and intensity ▪ Light carries energy defined by spectrum in straight line, ▪ Path modified according to optical properties of materials it encounters, ▪ Until it reaches the luminance detector to be interpreted as color and intensity

𝑆(θ, λ)

slide-9
SLIDE 9

Nicolas Dalmasso | Innovation Director Predictive Simulation for Industrial Applications 26/03/2018 9

▪ Predictive means Accurate Engine (ie energy based, spectral based) ▪ Predictive means Accurate Inputs (ie measurement and Spec sheet based) ▪ Predictive means Accurate Restitution (ie calibrated LCD, calibrated ToneMapping, HDR+ displays)

Predictive means being Exhaustive

slide-10
SLIDE 10

Nicolas Dalmasso | Innovation Director Predictive Simulation for Industrial Applications 26/03/2018 10

Brief History of Optis in Predictive Rendering

What actually changed ?

1994 : Photometry Nowadays

▪ From Optical design to Energy Propagation

Legacy technology of OPTIS was Optical Design through Sequential Raytracing Then Optis developped Non Sequential Propagation for energy calculation = Photometry

▪ Photometry

Science of the measurement of light as perceived by the human eye. Mainly used to create and validate lighting systems. Derived from radiometry which is measurement of light. Completed by Colorimetry which separates the Photometric perception of the human eye according to its cones/rod tristimulus

Moore’s Law !

slide-11
SLIDE 11

Nicolas Dalmasso | Innovation Director Predictive Simulation for Industrial Applications 26/03/2018 11

▪ Spectral data and computations all along the process

▪ Surfaces and Volumes : Measured spectral BRDF ▪ Lightsources : Flux/Luminance (energy), Spectrum, Intensity diagrams, Spectral Sky, Polarization ▪ Renderers : Rasterizer (OpenGL), Deterministic Raytracing (CUDA), CPU/GPU Raytracing (OptiX)

▪ Human Vision and Camera Sensors simulation for accurate result restitution

The Importance of Accurate Engine

slide-12
SLIDE 12

Nicolas Dalmasso | Innovation Director Predictive Simulation for Industrial Applications 26/03/2018 12

slide-13
SLIDE 13

Nicolas Dalmasso | Innovation Director Predictive Simulation for Industrial Applications 26/03/2018 13

▪ BxDF (BRDF, BSDF, BTDF, …), and beyond !

▪ Predictive requires measurement ▪ No model fit : RAW measurement used for accuracy at all angles ▪ Optis developped its own brdf acquisition devices ▪ OMS4 :10^12 dynamics ▪ OMS2 : 1 minute HD BRDF acquisition

▪ HD BRDF : 360°capture, high dynamics, wavelength dependent

▪ Iridescence ▪ Anisotropy ▪ BTDF / BSDF ▪ Unpolished surfaces ▪ Polarizer, coating, grating, retroreflecting

The Importance of Accurate Inputs – Surface Properties

slide-14
SLIDE 14

Nicolas Dalmasso | Innovation Director Predictive Simulation for Industrial Applications 26/03/2018 14

The Importance of Accurate Inputs – Surface Properties

slide-15
SLIDE 15

Nicolas Dalmasso | Innovation Director Predictive Simulation for Industrial Applications 26/03/2018 15

▪ Volume Materials

▪ Spectral IoR ▪ Spectral linear absorption ▪ Volume scattering (MIE, Heinwey Greenstein, with wavelength dependent phase function) ▪ Transparent, translucid, smoky, foggy, milky ▪ Birefringent, fluorescent

▪ As always we ensure Energy Conservation all along the computations

The Importance of Accurate Inputs –Volume Properties

slide-16
SLIDE 16

Nicolas Dalmasso | Innovation Director Predictive Simulation for Industrial Applications 26/03/2018 16

The Importance of Accurate Inputs –Volume Properties

slide-17
SLIDE 17

Nicolas Dalmasso | Innovation Director Predictive Simulation for Industrial Applications 26/03/2018 17

▪ Result map is

▪ Radiometric : energy stored per pixel ▪ Photometric : amount of light as seen by human eye ▪ Colorimetric : accurate color info, even if display cannot show ▪ Spectral : stored per pixel for deeper analysis ▪ Layered per source : accurate dimming postprocessing ▪ Possible accurate Human Vision based tone mapping and Spectral Camera simulation

▪ Accurate display

▪ Does not exist, Wide Gamut / Wide dynamic, Away from any ambient light

The Importance of Accurate Restitution

slide-18
SLIDE 18

Nicolas Dalmasso | Innovation Director Predictive Simulation for Industrial Applications 26/03/2018 18

▪ Accurate Geometry ▪ Accurate Inputs

▪ Materials ▪ Light sources ▪ Natural light

▪ Accurate Simulation ▪ Accurate Results ▪ Accurate Reality ▪ The meaning of Predictive

Practical Example – Our office in T

  • ulon
slide-19
SLIDE 19

Nicolas Dalmasso | Innovation Director Predictive Simulation for Industrial Applications 26/03/2018 19

▪ Our French branch Office

▪ 2nd stage of a building ▪ South of France (stable sunny weather)

▪ Modeled in CAD software for high accuracy dimensions ▪ Using high accuracy measurement tools (well, +/- 1mm)

Practical Example –Accurate Geometry

slide-20
SLIDE 20

Nicolas Dalmasso | Innovation Director Predictive Simulation for Industrial Applications 26/03/2018 20

▪ Materials

▪ BRDF acquisition through our acquisition devices ▪ OMS2 BRDF is 2 minutes to get ▪ Glass modeled from Spectral Index of Refraction and Spectral Absorption

▪ Light Sources

▪ Neon tubes from datasheet ▪ Whole lighting system modelized ▪ LCD from measurement

▪ Natural Light

▪ GPS coordinates of Optis HQ location ▪ Date (year/month/day/hour/minute), 2012/03/20 at 10:10 am ▪ HDR capture of the surrounding

Practical Example –Accurate Inputs

slide-21
SLIDE 21

Nicolas Dalmasso | Innovation Director Predictive Simulation for Industrial Applications 26/03/2018 21

▪ Luminance value per pixel ▪ Colorimetry per pixel ▪ Human vision management ▪ Source layering ▪ Post-process filtering

Practical Example –Accurate Results

slide-22
SLIDE 22

Nicolas Dalmasso | Innovation Director Predictive Simulation for Industrial Applications 26/03/2018 22

  • Daytime, Sun

Practical Example – 1 Simulation, Multiple Results

slide-23
SLIDE 23

Nicolas Dalmasso | Innovation Director Predictive Simulation for Industrial Applications 26/03/2018 23

  • Daytime, no Sun, no Light

Practical Example – 1 Simulation, Multiple Results

slide-24
SLIDE 24

Nicolas Dalmasso | Innovation Director Predictive Simulation for Industrial Applications 26/03/2018 24

  • Daytime, all sources on

Practical Example – 1 Simulation, Multiple Results

slide-25
SLIDE 25

Nicolas Dalmasso | Innovation Director Predictive Simulation for Industrial Applications 26/03/2018 25

  • Nighttime, all sources on

Practical Example – 1 Simulation, Multiple Results

slide-26
SLIDE 26

Nicolas Dalmasso | Innovation Director Predictive Simulation for Industrial Applications 26/03/2018 26

  • Nighttime, only displays on

Practical Example – 1 Simulation, Multiple Results

slide-27
SLIDE 27

Nicolas Dalmasso | Innovation Director Predictive Simulation for Industrial Applications 26/03/2018 27

Practical Example – Our office in T

  • ulon

▪ Inception Result, predicted shadow pattern

slide-28
SLIDE 28

Nicolas Dalmasso | Innovation Director Predictive Simulation for Industrial Applications 26/03/2018 28

▪ Inception Result, predicted shadow pattern

Practical Example – Our office in T

  • ulon
slide-29
SLIDE 29

Nicolas Dalmasso | Innovation Director Predictive Simulation for Industrial Applications 26/03/2018 29

▪ Predictive Result

Practical Example – Our office in T

  • ulon
slide-30
SLIDE 30

Nicolas Dalmasso | Innovation Director Predictive Simulation for Industrial Applications 26/03/2018 30

▪ Predictive Result

Practical Example – Our office in T

  • ulon
slide-31
SLIDE 31

Nicolas Dalmasso | Innovation Director Predictive Simulation for Industrial Applications 26/03/2018 31

▪ Visual Accuracy

▪ Simulation settings: Daytime, Sunny day, March 29th 2012 at 10:00 ▪ Photo shot at 10:10

Same lighting pattern Same soft shadows Different sunlight position because of the 10 minutes difference between Photo and Simulation parameters

Comparison to Reality

slide-32
SLIDE 32

Nicolas Dalmasso | Innovation Director Predictive Simulation for Industrial Applications 26/03/2018 32

Comparison to Reality

Same lighting pattern Previous simulation of the very same view on display Same reflection

▪ Visual Accuracy

▪ Simulation settings: Daytime, Sunny day, March 29th 2012 at 10:00 ▪ Photo shot at 10:10

slide-33
SLIDE 33

Nicolas Dalmasso | Innovation Director Predictive Simulation for Industrial Applications 26/03/2018 33

Comparison to Reality

Same reflection position Same shadow

▪ Visual Accuracy

▪ Simulation settings: Daytime, Sunny day, March 29th 2012 at 10:00 ▪ Photo shot at 10:10

slide-34
SLIDE 34

Nicolas Dalmasso | Innovation Director Predictive Simulation for Industrial Applications 26/03/2018 34

▪ Visual Accuracy

▪ Wide dynamic, 50% shade on photometric values invisible to the eye

▪ Photometry Accuracy

▪ Linear Scale, 10% shift on photometric value is critical

▪ Certification Accuracy

▪ Standards in visibility and ledgivility for safety !

How Accurate is Predictive ?

slide-35
SLIDE 35

Nicolas Dalmasso | Innovation Director Predictive Simulation for Industrial Applications 26/03/2018 35

PHYSICAL BASED CAMERA MODEL

Raw camera sensor model.

– Fine modeling of camera sensor from lens system to imager. – Improve test of image processing base algorithm on a variety of scenery. – Raw image usable for fine Hardware In the Loop image injection inside ECU.

Exposure time Electronic noise Lens distortion

slide-36
SLIDE 36

Nicolas Dalmasso | Innovation Director Predictive Simulation for Industrial Applications 26/03/2018 36

▪ Lens System

▪ Distortion ▪ Natural Vignetting ▪ Lens Transmission ▪ Noise

PHYSICAL BASED CAMERA MODEL - Parameters

▪ Imager

▪ Color filter array ▪ Quantum Efficiency ▪ Dynamic Range ▪ ExposureTime ▪ Various noise sources ▪ Amplification ▪ Discretization

▪ Processing:

▪ Demosaicing ▪ Tonemapping ▪ Color space conversion ▪ RGB

slide-37
SLIDE 37

Nicolas Dalmasso | Innovation Director Predictive Simulation for Industrial Applications 26/03/2018 37

LIDAR IMPLEMENT

A TION – FIELD OFVIEWASSESSMENT

▪ Using physical based Solid State LiDAR model, preview of Field of view from : ▪ Emitter ▪ Receiver ▪ Intersection

Field of Vision of emitter (pink) and receiver (blue)

slide-38
SLIDE 38

Nicolas Dalmasso | Innovation Director Predictive Simulation for Industrial Applications 26/03/2018 38

LIDAR PERCEPTION – RAW SIMULA

TION

▪ Get from simulation of physical based Solid State LiDAR model: ▪ Raw optical signal on LiDAR receiver ▪ 3D mapping of perceived environment ▪ Define diffusive ambient medium

Perceived environment (distance map) by LiDAR Raw optical signal on a pixel

slide-39
SLIDE 39

Nicolas Dalmasso | Innovation Director Predictive Simulation for Industrial Applications 26/03/2018 39

USE CASE - LIDAR SIMULA

TION

slide-40
SLIDE 40

www.optis-world.com

Thank You