automatic selection of ergonomic indicators for the design...
TRANSCRIPT
Automatic selection of ergonomic indicators for the design ofcollaborative robots: a virtual-human in the loop approach
P. Maurice, P. Schlehuber, Y. Measson, V. Padois, P. Bidaud
F R O M R E S E A R C H T O I N D U S T R Y
2014 IEEE-RAS International Conference on Humanoid RobotsNovember 19, 2014
Introduction
Work-related musculoskeletal disorders: A major health problem
Pauline MAURICE Selection of ergonomic indicators for collaborative robotics using a DHM Humanoids 2014 1 / 11
Statistics [Schneider and Irastorza, 2010]
I Affect over 35 % of workers in EuropeI Represent the 1st occupational diseaseI Increase by 15 % per yearI Cost about $50B a year in the US
Carpal tunnel syndrome
Rotator cuff tendinitis
Bursitis
Epicondylitis
Achilles tendinitis
Low back pain
Tension necksyndrome
Main biomechanical risk factorsI Extreme posturesI Considerable effortsI Static workI High frequency of the gestures
Introduction
Collaborative robotics: A physical assistance for complex tasks
Pauline MAURICE Selection of ergonomic indicators for collaborative robotics using a DHM Humanoids 2014 2 / 11
Robot [Colgate et al., 2003]
I Weight compensationI Strength amplificationI Guidance via virtual paths
HumanI Technical expertiseI AdaptabilityI Decision
Co-manipulation of objects or tools
Introduction
Limitations of DHM based ergonomic assessments
Pauline MAURICE Selection of ergonomic indicators for collaborative robotics using a DHM Humanoids 2014 3 / 11
Macroscopic modelsI Products: Delmia, Jack 1,
Sammie 1, 3DSSPP 2 . . .I Ergonomic assessments:
RULA 5, OWAS 5, Snook tables 6,NIOSH 7, Low-back analysis . . .
Rough or Task-specificOne global criterion
Biomechanical modelsI Products: OpenSim 3, Anybody 4,
LifeMOD, Santos . . .I Ergonomic assessments:
Joint force, Muscle force, Tendonlength . . .
Numerous criteriaAccurate and Generic
1[Delleman et al., 2004], 2[Chaffin et al., 2006], 3[Delp et al., 2007], 4[Damsgaard et al., 2006],5[Li and Buckle, 1999], 6[Snook and Ciriello, 1991], 7[Waters et al., 1993]
Introduction
Limitations of DHM based ergonomic assessments
Pauline MAURICE Selection of ergonomic indicators for collaborative robotics using a DHM Humanoids 2014 3 / 11
Macroscopic modelsI Products: Delmia, Jack 1,
Sammie 1, 3DSSPP 2 . . .I Ergonomic assessments:
RULA 5, OWAS 5, Snook tables 6,NIOSH 7, Low-back analysis . . .
Rough or Task-specificOne global criterion
Biomechanical modelsI Products: OpenSim 3, Anybody 4,
LifeMOD, Santos . . .I Ergonomic assessments:
Joint force, Muscle force, Tendonlength . . .
Numerous criteriaAccurate and Generic
1[Delleman et al., 2004], 2[Chaffin et al., 2006], 3[Delp et al., 2007], 4[Damsgaard et al., 2006],5[Li and Buckle, 1999], 6[Snook and Ciriello, 1991], 7[Waters et al., 1993]
Selection of relevant ergonomic indicatorsI Dedicated to the comparison of
collaborative robotsI Dependent on task featuresI Independent from robot designI Automatic DHM-based process
One global criterion Accurate and Generic
Method
Pauline MAURICE Selection of ergonomic indicators for collaborative robotics using a DHM Humanoids 2014 3 / 11
1 Introduction
2 Method
3 Results
4 Conclusion
Method
Indicators relevance: Differentiating various ways of performing a task
Pauline MAURICE Selection of ergonomic indicators for collaborative robotics using a DHM Humanoids 2014 4 / 11
Dynamic simulation
Task description
Robot controllerForce amplification
Manikin controller
LQP
Parameters set #NParameters set #1
Human and robotparameters
Selection
...
Analysis
Relevant ergonomic indicators
Indicators set #1 Indicators set #N...
Method
Parameters selection: Creating a variety of situations
Pauline MAURICE Selection of ergonomic indicators for collaborative robotics using a DHM Humanoids 2014 5 / 11
MK
B
Frob = α Fvh
Frob
Collaborative robot Abstraction of the robot
τrob = α JT Fvh + g(q)
ParametersI Amplification coefficientI Robot massI Upper body joint limitsI Pelvis positionI Upper body reference postureI Upper body tasks weightsI Step lengthI Human sizeI Human body mass index
ExplorationFourier amplitude sensitivity
testing (FAST) [Saltelli et al., 1999]
Method
A dynamic DHM for indicators calculation
Pauline MAURICE Selection of ergonomic indicators for collaborative robotics using a DHM Humanoids 2014 6 / 11
TasksI Balance: ZMP preview controlI Hands trajectories and forcesI Whole body postureI Torques minimization
Optimization [Salini et al., 2011]
Linear Quadratic Programmingwith weighting strategy
ConstraintsI Dynamical model equationI Joint limitsI Joint torques saturationI Non sliding contacts
Joint torquesContact forcesManikin stateaa
bb
{ }Ergonomicindicators
Dynamic simulationXDE framework (CEA-LIST)
”Robot” controllerForce amplification
”Torques”
Robot stateInteraction force
Method
Ergonomic indicators selection: A variance-based analysis
Pauline MAURICE Selection of ergonomic indicators for collaborative robotics using a DHM Humanoids 2014 7 / 11
Indicator 1 Indicator N Local indicatorsI PositionsI VelocitiesI AccelerationsI TorquesI Power
I BackI Right armI Left armI Legs
Global indicatorsI Kinetic energyI Force transmission ratioI Velocity transmission ratioI Balance robustnessI Dynamic balance
∫task
I1(t) dt
Scaling
Variance
...
...
...
...
∫task
IN(t) dt
Scaling
Variance
Scree test: elbow criterion
Discriminating indicators
Method
Ergonomic indicators selection: A variance-based analysis
Pauline MAURICE Selection of ergonomic indicators for collaborative robotics using a DHM Humanoids 2014 7 / 11
Indicator 1 Indicator N
∫task
I1(t) dt
Scaling
Variance
...
...
...
...
∫task
IN(t) dt
Scaling
Variance
Scree test: elbow criterion
Discriminating indicators
Scaling Scaling
Scaling valueI mean(Ii/I ref
i ) = 1I variance(Ii/I ref
i ) 6= 1
I refi =
∑m∈T
∑n∈P
Im,ni
NT NP
T : tasks, P: parameters sets
Method
Ergonomic indicators selection: A variance-based analysis
Pauline MAURICE Selection of ergonomic indicators for collaborative robotics using a DHM Humanoids 2014 7 / 11
Indicator 1 Indicator N
∫task
I1(t) dt
Scaling
Variance
...
...
...
...
∫task
IN(t) dt
Scaling
Variance
Scree test: elbow criterion
Discriminating indicators
Scaling valueI mean(Ii/I ref
i ) = 1I variance(Ii/I ref
i ) 6= 1
I refi =
∑m∈T
∑n∈P
Im,ni
NT NP
T : tasks, P: parameters sets
Scree test: elbow criterion
Scree plot
I1 I2 I3 I4 I5 I7I6
Varia
nces
Indicators
elbow
selected not selected
Results
Pauline MAURICE Selection of ergonomic indicators for collaborative robotics using a DHM Humanoids 2014 7 / 11
1 Introduction
2 Method
3 Results
4 Conclusion
Results
Application to various tasks: Selected indicators
Pauline MAURICE Selection of ergonomic indicators for collaborative robotics using a DHM Humanoids 2014 8 / 11
Walk Reach Fast traj. Push Bend andsideways 35 cm tracking 100 N Carry 3 kg
81 % 80 % 86 % 81 % 82 %
Walk Reach Fast traj. Push Bend andsideways 35 cm tracking 100 N Carry 3 kg
Kinetic energyVelocity Transmission Ratio
Force Transmission RatioDynamic balance
Balance robustnessLegs torqueLegs power
Legs accelerationLegs velocityLegs position
Left arm torqueLeft arm power
Left arm accelerationLeft arm velocityLeft arm positionRight arm torqueRight arm power
Right arm accelerationRight arm velocityRight arm position
Back torqueBack power
Back accelerationBack velocityBack position
I 3 to 8 indicators selectedout of 29
I Variance information loss< 20 %
Results
Application to various tasks: Illustration of the parameters effects
Pauline MAURICE Selection of ergonomic indicators for collaborative robotics using a DHM Humanoids 2014 9 / 11
Conclusion
Pauline MAURICE Selection of ergonomic indicators for collaborative robotics using a DHM Humanoids 2014 9 / 11
1 Introduction
2 Method
3 Results
4 Conclusion
Conclusion
Conclusion: Automatic selection of ergonomic indicators
Pauline MAURICE Selection of ergonomic indicators for collaborative robotics using a DHM Humanoids 2014 10 / 11
ConstraintsI Dedicated to collaborative
roboticsI Independent from robot designI Dependent on task featuresI AutomaticI Differentiate various ways of
performing a task
MethodI Parameters: Varying human and
robot (abstraction) featuresI Dynamic simulation: DHM with
LQP based controllerI Indicators: Variance-based
analysis of mutliple biomechanicalquantities
ResultsI Physically consistent selectionI 6 relevant indicators on averageI > 80 % information remains
Conclusion
Future work: Application to industrial jobs
Pauline MAURICE Selection of ergonomic indicators for collaborative robotics using a DHM Humanoids 2014 11 / 11
Application to complex tasksI Indicator
∫task I(t) dt → Suitable for elementary tasks only
I Industrial tasks = Succession of elementary tasksI Where is the limit between 2 tasks?
subtask 1 subtask 2 subtask 3
influence influence
limit? limit?
Optimal design of collaborative robotsI Identify the most influential parameters to orient the design
work: Sensitivity analysis
I Combine assessment method with evolutionary algorithmI Optimize mechanical and/or control parameters of the robot
Conclusion
Pauline MAURICE Selection of ergonomic indicators for collaborative robotics using a DHM Humanoids 2014 11 / 11
Thank you
Appendices
LQP Controller [Salini et al., 2011]
Pauline MAURICE Selection of ergonomic indicators for collaborative robotics using a DHM Humanoids 2014
Optimization
argminX
∑i
ωiTi(X) where X = (τ , wc, q)T
s.t.{
M(q)q + C(q, q) + g(q) = S τ − JTc (q)wc
GX � h
Tasks
• Joint torque ‖τ − τ ∗‖2
• Operational space wrench ‖wi − w∗i ‖2
• Joint acceleration ‖q− q∗‖2
• Operational space acceleration ‖Ji q + Ji q− X∗i ‖2
with X∗ = Xgoal + Kv (Xgoal − X) + Kp(Xgoal − X)