SlideShare a Scribd company logo
1 of 1
Download to read offline
www.data61.csiro.au
Why Simulation?
• No wear or damage to real-world hardware;
• Many instantiations of a simulation can run in parallel;
• (Often) Faster than real-time operation;
• Instant access to robots without having to purchase;
• Human intervention is not required; and
• Safety.
The ‘Reality Gap’ is the disparity seen when transferring a controller
generated, evolved or learnt in simulation to the real-world. Many things
cause the reality gap including:
Results
We compare real-world data to simulation data using 3D Euclidean distance.
Scene 1 - Designed to compare single joint kinematics.
Scene 2 - Two joint kinematics over an extended time period.
Scene 3 - Multiple joint kinematics and interaction dynamics.
Figure 1: Real-world and simulated environments with the Kinova arm and cube visualised. (A) Real-world setup
with tracking markers attached to dynamic elements; (B) the MuJoCo environment; (C) the PyBullet
environment; (D) the V-Rep environment.
REFERENCES
J. Collins, D. Howard, and J. Leitner, “Quantifying the Reality Gap in Robotic
Manipulation Tasks,” in 2019 IEEE International Conference on Robotics and
Automation (ICRA).
ACKNOWLEDGEMENTS
This research was supported by a Data61 PhD
Scholarship. This research was further supported by
the Australian Research Council Centre of Excellence
for Robotic Vision (project number CE140100016).
QUANTIFYING THE REALITY GAP IN
ROBOTIC MANIPULATION TASKS
We quantify the accuracy of various simulators compared to real-world robotic reaching and interaction tasks. Making
use of a Kinova robotic manipulator and a motion capture system, we record ground truth data enabling comparisons
with various simulators. We show the relative strengths and weaknesses of numerous contemporary simulators,
highlighting areas of significant discrepancy.
Jack Collins1,2, David Howard2 and Jürgen Leitner1,3
1 Queensland University of Technology (QUT), Brisbane, Australia
2 Data61/Commonwealth Scientific and Industrial Research Organisation, Brisbane, Australia
3 Australian Centre for Robotic Vision (ACRV)
Figure 3: (a) Plotted lines represent the wrist's Euclidean distance from the final goal position; and (b) the x-axis is the
motion capture with plotted lines being the Euclidean distance from the ground truth. (i-iv) are areas of note within the
plot. MuJoCo and Bullet 2.83 oscillate like the motion capture although neither are able to imitate the exact real world
motion. Evident in (iii) and from the large spikes seen in (b), MuJoCo, Bullet 2.83 and Bullet 2.78 have large error-full
periods when moving against gravity. Newton and Vortex follow the motion capture path the closest.
Method
The setup consists of a Kinova Mico2 6DOF arm with an attached KG-3
gripper. The arm is controlled using joint velocities and a proportional
controller running at 5Hz.
Real World Ground Truth
Motion capture data was recorded at 100Hz to an accuracy of <1mm. The
arm was controlled through ROS.
Simulation
Three leading robotic simulators were used to collect simulated data:
• V-Rep
• PyBullet
• MuJoCo
The majority of settings are left to their default value as we want to see how
well a generic scene can perform without the knowledge of an expert.
FOR FURTHER INFORMATION
Jack Collins
e jack.collins@data61.csiro.au
• Mathematical Precision
• Temporal Dynamics
• Under-modelled actuators/sensors
• Synthetic Image Data
• Physical Interactions
Figure 4: (a) Plotted lines represent the wrist's Euclidean distance from the final goal position; (b) the x-axis represents
the start position of the cube with plotted lines being the cubes Euclidean distance away; and (c) the pitch of the cube
(where pitch is the rotation around the y-axis). (i-iv) are areas of note within the plot. Graph (a) visually shows that there
is no simulator that follows the trajectory of the motion capture remarkably well. Between plots (b) and (c) we see that
MuJoCo and Bullet 2.83 are closest at replicating the displacement of the cube but that these two simulators also flip the
cube 90° onto another face unlike the motion capture. Other simulators interact with the cube minimally or not at all.
Figure 2: Plotted is the mean Euclidean distance from the goal position (calculated over 20 runs), (i) and (ii) are areas of note
within the plot. V-Rep ODE is not plotted as it was found to be unstable from self contacts. Oscillation is evident in the plots
of the motion capture, Bullet 2.83 and Bullet 2.78 and is likely due to overshoot by the proportional controller. The
simulators that follow the motion capture closest are Vortex, Newton and PyBullet.

More Related Content

What's hot

What's hot (6)

Human arm tracking
Human arm trackingHuman arm tracking
Human arm tracking
 
Luigy Bertaglia Bortolo - Poster Final
Luigy Bertaglia Bortolo - Poster FinalLuigy Bertaglia Bortolo - Poster Final
Luigy Bertaglia Bortolo - Poster Final
 
The flow of baseline estimation using a single omnidirectional camera
The flow of baseline estimation using a single omnidirectional cameraThe flow of baseline estimation using a single omnidirectional camera
The flow of baseline estimation using a single omnidirectional camera
 
Development of a robust filtering algorithm for inertial sensor based navigation
Development of a robust filtering algorithm for inertial sensor based navigationDevelopment of a robust filtering algorithm for inertial sensor based navigation
Development of a robust filtering algorithm for inertial sensor based navigation
 
Virtual Dance Game Using Kinect and ICP Algorithm
Virtual Dance Game Using Kinect and ICP AlgorithmVirtual Dance Game Using Kinect and ICP Algorithm
Virtual Dance Game Using Kinect and ICP Algorithm
 
Controlling Anisotropy In Mass-Spring Systems
Controlling Anisotropy In Mass-Spring SystemsControlling Anisotropy In Mass-Spring Systems
Controlling Anisotropy In Mass-Spring Systems
 

Similar to ICRA2019 Poster

Control Systems For Projectile DefenseRyan MendivilMar.docx
Control Systems For Projectile DefenseRyan MendivilMar.docxControl Systems For Projectile DefenseRyan MendivilMar.docx
Control Systems For Projectile DefenseRyan MendivilMar.docx
maxinesmith73660
 
Object Distance Detection using a Joint Transform Correlator
Object Distance Detection using a Joint Transform CorrelatorObject Distance Detection using a Joint Transform Correlator
Object Distance Detection using a Joint Transform Correlator
Alexander Layton
 
Place Cell Latex report
Place Cell Latex reportPlace Cell Latex report
Place Cell Latex report
Jacob Senior
 

Similar to ICRA2019 Poster (20)

report
reportreport
report
 
Control Systems For Projectile DefenseRyan MendivilMar.docx
Control Systems For Projectile DefenseRyan MendivilMar.docxControl Systems For Projectile DefenseRyan MendivilMar.docx
Control Systems For Projectile DefenseRyan MendivilMar.docx
 
Fuzzy-proportional-integral-derivative-based controller for object tracking i...
Fuzzy-proportional-integral-derivative-based controller for object tracking i...Fuzzy-proportional-integral-derivative-based controller for object tracking i...
Fuzzy-proportional-integral-derivative-based controller for object tracking i...
 
Estimation of Terrain Gradient Conditions & Obstacle Detection Using a Monocu...
Estimation of Terrain Gradient Conditions & Obstacle Detection Using a Monocu...Estimation of Terrain Gradient Conditions & Obstacle Detection Using a Monocu...
Estimation of Terrain Gradient Conditions & Obstacle Detection Using a Monocu...
 
Final Project
Final ProjectFinal Project
Final Project
 
AUTO LANDING PROCESS FOR AUTONOMOUS FLYING ROBOT BY USING IMAGE PROCESSING BA...
AUTO LANDING PROCESS FOR AUTONOMOUS FLYING ROBOT BY USING IMAGE PROCESSING BA...AUTO LANDING PROCESS FOR AUTONOMOUS FLYING ROBOT BY USING IMAGE PROCESSING BA...
AUTO LANDING PROCESS FOR AUTONOMOUS FLYING ROBOT BY USING IMAGE PROCESSING BA...
 
I0343065072
I0343065072I0343065072
I0343065072
 
Object Distance Detection using a Joint Transform Correlator
Object Distance Detection using a Joint Transform CorrelatorObject Distance Detection using a Joint Transform Correlator
Object Distance Detection using a Joint Transform Correlator
 
Visual Environment by Semantic Segmentation Using Deep Learning: A Prototype ...
Visual Environment by Semantic Segmentation Using Deep Learning: A Prototype ...Visual Environment by Semantic Segmentation Using Deep Learning: A Prototype ...
Visual Environment by Semantic Segmentation Using Deep Learning: A Prototype ...
 
A NOVEL NAVIGATION STRATEGY FOR AN UNICYCLE MOBILE ROBOT INSPIRED FROM THE DU...
A NOVEL NAVIGATION STRATEGY FOR AN UNICYCLE MOBILE ROBOT INSPIRED FROM THE DU...A NOVEL NAVIGATION STRATEGY FOR AN UNICYCLE MOBILE ROBOT INSPIRED FROM THE DU...
A NOVEL NAVIGATION STRATEGY FOR AN UNICYCLE MOBILE ROBOT INSPIRED FROM THE DU...
 
Ca36464468
Ca36464468Ca36464468
Ca36464468
 
“A Computer Vision System for Autonomous Satellite Maneuvering,” a Presentati...
“A Computer Vision System for Autonomous Satellite Maneuvering,” a Presentati...“A Computer Vision System for Autonomous Satellite Maneuvering,” a Presentati...
“A Computer Vision System for Autonomous Satellite Maneuvering,” a Presentati...
 
VARIATIONAL MONTE-CARLO APPROACH FOR ARTICULATED OBJECT TRACKING
VARIATIONAL MONTE-CARLO APPROACH FOR ARTICULATED OBJECT TRACKINGVARIATIONAL MONTE-CARLO APPROACH FOR ARTICULATED OBJECT TRACKING
VARIATIONAL MONTE-CARLO APPROACH FOR ARTICULATED OBJECT TRACKING
 
SkyStitch: a Cooperative Multi-UAV-based Real-time Video Surveillance System ...
SkyStitch: a Cooperative Multi-UAV-based Real-time Video Surveillance System ...SkyStitch: a Cooperative Multi-UAV-based Real-time Video Surveillance System ...
SkyStitch: a Cooperative Multi-UAV-based Real-time Video Surveillance System ...
 
Comparison of Some Motion Detection Methods in cases of Single and Multiple M...
Comparison of Some Motion Detection Methods in cases of Single and Multiple M...Comparison of Some Motion Detection Methods in cases of Single and Multiple M...
Comparison of Some Motion Detection Methods in cases of Single and Multiple M...
 
VARIATIONAL MONTE-CARLO APPROACH FOR ARTICULATED OBJECT TRACKING
VARIATIONAL MONTE-CARLO APPROACH FOR ARTICULATED OBJECT TRACKINGVARIATIONAL MONTE-CARLO APPROACH FOR ARTICULATED OBJECT TRACKING
VARIATIONAL MONTE-CARLO APPROACH FOR ARTICULATED OBJECT TRACKING
 
E0333021025
E0333021025E0333021025
E0333021025
 
Place Cell Latex report
Place Cell Latex reportPlace Cell Latex report
Place Cell Latex report
 
Conference Publication
Conference PublicationConference Publication
Conference Publication
 
פוסטר דר פרידמן
פוסטר דר פרידמןפוסטר דר פרידמן
פוסטר דר פרידמן
 

Recently uploaded

XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX
XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX
XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX
ssuser89054b
 
scipt v1.pptxcxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx...
scipt v1.pptxcxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx...scipt v1.pptxcxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx...
scipt v1.pptxcxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx...
HenryBriggs2
 
"Lesotho Leaps Forward: A Chronicle of Transformative Developments"
"Lesotho Leaps Forward: A Chronicle of Transformative Developments""Lesotho Leaps Forward: A Chronicle of Transformative Developments"
"Lesotho Leaps Forward: A Chronicle of Transformative Developments"
mphochane1998
 
1_Introduction + EAM Vocabulary + how to navigate in EAM.pdf
1_Introduction + EAM Vocabulary + how to navigate in EAM.pdf1_Introduction + EAM Vocabulary + how to navigate in EAM.pdf
1_Introduction + EAM Vocabulary + how to navigate in EAM.pdf
AldoGarca30
 
DeepFakes presentation : brief idea of DeepFakes
DeepFakes presentation : brief idea of DeepFakesDeepFakes presentation : brief idea of DeepFakes
DeepFakes presentation : brief idea of DeepFakes
MayuraD1
 
Cara Menggugurkan Sperma Yang Masuk Rahim Biyar Tidak Hamil
Cara Menggugurkan Sperma Yang Masuk Rahim Biyar Tidak HamilCara Menggugurkan Sperma Yang Masuk Rahim Biyar Tidak Hamil
Cara Menggugurkan Sperma Yang Masuk Rahim Biyar Tidak Hamil
Cara Menggugurkan Kandungan 087776558899
 

Recently uploaded (20)

XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX
XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX
XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX
 
NO1 Top No1 Amil Baba In Azad Kashmir, Kashmir Black Magic Specialist Expert ...
NO1 Top No1 Amil Baba In Azad Kashmir, Kashmir Black Magic Specialist Expert ...NO1 Top No1 Amil Baba In Azad Kashmir, Kashmir Black Magic Specialist Expert ...
NO1 Top No1 Amil Baba In Azad Kashmir, Kashmir Black Magic Specialist Expert ...
 
Orlando’s Arnold Palmer Hospital Layout Strategy-1.pptx
Orlando’s Arnold Palmer Hospital Layout Strategy-1.pptxOrlando’s Arnold Palmer Hospital Layout Strategy-1.pptx
Orlando’s Arnold Palmer Hospital Layout Strategy-1.pptx
 
scipt v1.pptxcxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx...
scipt v1.pptxcxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx...scipt v1.pptxcxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx...
scipt v1.pptxcxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx...
 
Moment Distribution Method For Btech Civil
Moment Distribution Method For Btech CivilMoment Distribution Method For Btech Civil
Moment Distribution Method For Btech Civil
 
Introduction to Serverless with AWS Lambda
Introduction to Serverless with AWS LambdaIntroduction to Serverless with AWS Lambda
Introduction to Serverless with AWS Lambda
 
AIRCANVAS[1].pdf mini project for btech students
AIRCANVAS[1].pdf mini project for btech studentsAIRCANVAS[1].pdf mini project for btech students
AIRCANVAS[1].pdf mini project for btech students
 
Thermal Engineering Unit - I & II . ppt
Thermal Engineering  Unit - I & II . pptThermal Engineering  Unit - I & II . ppt
Thermal Engineering Unit - I & II . ppt
 
"Lesotho Leaps Forward: A Chronicle of Transformative Developments"
"Lesotho Leaps Forward: A Chronicle of Transformative Developments""Lesotho Leaps Forward: A Chronicle of Transformative Developments"
"Lesotho Leaps Forward: A Chronicle of Transformative Developments"
 
Navigating Complexity: The Role of Trusted Partners and VIAS3D in Dassault Sy...
Navigating Complexity: The Role of Trusted Partners and VIAS3D in Dassault Sy...Navigating Complexity: The Role of Trusted Partners and VIAS3D in Dassault Sy...
Navigating Complexity: The Role of Trusted Partners and VIAS3D in Dassault Sy...
 
FEA Based Level 3 Assessment of Deformed Tanks with Fluid Induced Loads
FEA Based Level 3 Assessment of Deformed Tanks with Fluid Induced LoadsFEA Based Level 3 Assessment of Deformed Tanks with Fluid Induced Loads
FEA Based Level 3 Assessment of Deformed Tanks with Fluid Induced Loads
 
Generative AI or GenAI technology based PPT
Generative AI or GenAI technology based PPTGenerative AI or GenAI technology based PPT
Generative AI or GenAI technology based PPT
 
Hostel management system project report..pdf
Hostel management system project report..pdfHostel management system project report..pdf
Hostel management system project report..pdf
 
Computer Networks Basics of Network Devices
Computer Networks  Basics of Network DevicesComputer Networks  Basics of Network Devices
Computer Networks Basics of Network Devices
 
Theory of Time 2024 (Universal Theory for Everything)
Theory of Time 2024 (Universal Theory for Everything)Theory of Time 2024 (Universal Theory for Everything)
Theory of Time 2024 (Universal Theory for Everything)
 
1_Introduction + EAM Vocabulary + how to navigate in EAM.pdf
1_Introduction + EAM Vocabulary + how to navigate in EAM.pdf1_Introduction + EAM Vocabulary + how to navigate in EAM.pdf
1_Introduction + EAM Vocabulary + how to navigate in EAM.pdf
 
Employee leave management system project.
Employee leave management system project.Employee leave management system project.
Employee leave management system project.
 
DeepFakes presentation : brief idea of DeepFakes
DeepFakes presentation : brief idea of DeepFakesDeepFakes presentation : brief idea of DeepFakes
DeepFakes presentation : brief idea of DeepFakes
 
A Study of Urban Area Plan for Pabna Municipality
A Study of Urban Area Plan for Pabna MunicipalityA Study of Urban Area Plan for Pabna Municipality
A Study of Urban Area Plan for Pabna Municipality
 
Cara Menggugurkan Sperma Yang Masuk Rahim Biyar Tidak Hamil
Cara Menggugurkan Sperma Yang Masuk Rahim Biyar Tidak HamilCara Menggugurkan Sperma Yang Masuk Rahim Biyar Tidak Hamil
Cara Menggugurkan Sperma Yang Masuk Rahim Biyar Tidak Hamil
 

ICRA2019 Poster

  • 1. www.data61.csiro.au Why Simulation? • No wear or damage to real-world hardware; • Many instantiations of a simulation can run in parallel; • (Often) Faster than real-time operation; • Instant access to robots without having to purchase; • Human intervention is not required; and • Safety. The ‘Reality Gap’ is the disparity seen when transferring a controller generated, evolved or learnt in simulation to the real-world. Many things cause the reality gap including: Results We compare real-world data to simulation data using 3D Euclidean distance. Scene 1 - Designed to compare single joint kinematics. Scene 2 - Two joint kinematics over an extended time period. Scene 3 - Multiple joint kinematics and interaction dynamics. Figure 1: Real-world and simulated environments with the Kinova arm and cube visualised. (A) Real-world setup with tracking markers attached to dynamic elements; (B) the MuJoCo environment; (C) the PyBullet environment; (D) the V-Rep environment. REFERENCES J. Collins, D. Howard, and J. Leitner, “Quantifying the Reality Gap in Robotic Manipulation Tasks,” in 2019 IEEE International Conference on Robotics and Automation (ICRA). ACKNOWLEDGEMENTS This research was supported by a Data61 PhD Scholarship. This research was further supported by the Australian Research Council Centre of Excellence for Robotic Vision (project number CE140100016). QUANTIFYING THE REALITY GAP IN ROBOTIC MANIPULATION TASKS We quantify the accuracy of various simulators compared to real-world robotic reaching and interaction tasks. Making use of a Kinova robotic manipulator and a motion capture system, we record ground truth data enabling comparisons with various simulators. We show the relative strengths and weaknesses of numerous contemporary simulators, highlighting areas of significant discrepancy. Jack Collins1,2, David Howard2 and Jürgen Leitner1,3 1 Queensland University of Technology (QUT), Brisbane, Australia 2 Data61/Commonwealth Scientific and Industrial Research Organisation, Brisbane, Australia 3 Australian Centre for Robotic Vision (ACRV) Figure 3: (a) Plotted lines represent the wrist's Euclidean distance from the final goal position; and (b) the x-axis is the motion capture with plotted lines being the Euclidean distance from the ground truth. (i-iv) are areas of note within the plot. MuJoCo and Bullet 2.83 oscillate like the motion capture although neither are able to imitate the exact real world motion. Evident in (iii) and from the large spikes seen in (b), MuJoCo, Bullet 2.83 and Bullet 2.78 have large error-full periods when moving against gravity. Newton and Vortex follow the motion capture path the closest. Method The setup consists of a Kinova Mico2 6DOF arm with an attached KG-3 gripper. The arm is controlled using joint velocities and a proportional controller running at 5Hz. Real World Ground Truth Motion capture data was recorded at 100Hz to an accuracy of <1mm. The arm was controlled through ROS. Simulation Three leading robotic simulators were used to collect simulated data: • V-Rep • PyBullet • MuJoCo The majority of settings are left to their default value as we want to see how well a generic scene can perform without the knowledge of an expert. FOR FURTHER INFORMATION Jack Collins e jack.collins@data61.csiro.au • Mathematical Precision • Temporal Dynamics • Under-modelled actuators/sensors • Synthetic Image Data • Physical Interactions Figure 4: (a) Plotted lines represent the wrist's Euclidean distance from the final goal position; (b) the x-axis represents the start position of the cube with plotted lines being the cubes Euclidean distance away; and (c) the pitch of the cube (where pitch is the rotation around the y-axis). (i-iv) are areas of note within the plot. Graph (a) visually shows that there is no simulator that follows the trajectory of the motion capture remarkably well. Between plots (b) and (c) we see that MuJoCo and Bullet 2.83 are closest at replicating the displacement of the cube but that these two simulators also flip the cube 90° onto another face unlike the motion capture. Other simulators interact with the cube minimally or not at all. Figure 2: Plotted is the mean Euclidean distance from the goal position (calculated over 20 runs), (i) and (ii) are areas of note within the plot. V-Rep ODE is not plotted as it was found to be unstable from self contacts. Oscillation is evident in the plots of the motion capture, Bullet 2.83 and Bullet 2.78 and is likely due to overshoot by the proportional controller. The simulators that follow the motion capture closest are Vortex, Newton and PyBullet.