control of biped locomotion inspired from walking in monkeys€¦ · 18 june 2010 / final...
TRANSCRIPT
Control ofbiped
locomotioninspired from
walking inmonkeys
Charles-HenryHoudemer
IntroductionInsights on BMIs
Previous work
Goal of the project
Challenges
Long term goals
ResultsWebots model
Numerical tests
CPG model
Results
Futureimprovements
References
Control of biped locomotion inspired fromwalking in monkeys
Charles-Henry Houdemer
BiorobEcole polytechnique fédérale de Lausanne
LSROÉcole polytechnique fédérale de Lausanne
18 June 2010 / Final Presentation
Supervisors : Sarah Dégallier, Jesse Van den Kieboom, Solaiman Shokur, Auke Jan Ijspeert
Control ofbiped
locomotioninspired from
walking inmonkeys
Charles-HenryHoudemer
IntroductionInsights on BMIs
Previous work
Goal of the project
Challenges
Long term goals
ResultsWebots model
Numerical tests
CPG model
Results
Futureimprovements
References
Outline
1 IntroductionInsights on BMIsPrevious workGoal of the projectChallengesLong term goals
2 ResultsWebots modelNumerical testsCPG modelResults
3 Future improvements
Control ofbiped
locomotioninspired from
walking inmonkeys
Charles-HenryHoudemer
IntroductionInsights on BMIs
Previous work
Goal of the project
Challenges
Long term goals
ResultsWebots model
Numerical tests
CPG model
Results
Futureimprovements
References
Insights on BMIs
Three different forms :InvasivePartially-invasiveNon-invasive
Two types :Motor BMISensory BMI
Transform data either from or to the brain by predictingand interpreting correct behaviour
Control ofbiped
locomotioninspired from
walking inmonkeys
Charles-HenryHoudemer
IntroductionInsights on BMIs
Previous work
Goal of the project
Challenges
Long term goals
ResultsWebots model
Numerical tests
CPG model
Results
Futureimprovements
References
Insights on BMIs
Three different forms :InvasivePartially-invasiveNon-invasive
Two types :Motor BMISensory BMI
Transform data either from or to the brain by predictingand interpreting correct behaviour
Control ofbiped
locomotioninspired from
walking inmonkeys
Charles-HenryHoudemer
IntroductionInsights on BMIs
Previous work
Goal of the project
Challenges
Long term goals
ResultsWebots model
Numerical tests
CPG model
Results
Futureimprovements
References
Insights on BMIs
Three different forms :InvasivePartially-invasiveNon-invasive
Two types :Motor BMISensory BMI
Transform data either from or to the brain by predictingand interpreting correct behaviour
Control ofbiped
locomotioninspired from
walking inmonkeys
Charles-HenryHoudemer
IntroductionInsights on BMIs
Previous work
Goal of the project
Challenges
Long term goals
ResultsWebots model
Numerical tests
CPG model
Results
Futureimprovements
References
Insights on BMIs
Three different forms :InvasivePartially-invasiveNon-invasive
Two types :Motor BMISensory BMI
Transform data either from or to the brain by predictingand interpreting correct behaviour
Control ofbiped
locomotioninspired from
walking inmonkeys
Charles-HenryHoudemer
IntroductionInsights on BMIs
Previous work
Goal of the project
Challenges
Long term goals
ResultsWebots model
Numerical tests
CPG model
Results
Futureimprovements
References
Insights on BMIs
Three different forms :InvasivePartially-invasiveNon-invasive
Two types :Motor BMISensory BMI
Transform data either from or to the brain by predictingand interpreting correct behaviour
Control ofbiped
locomotioninspired from
walking inmonkeys
Charles-HenryHoudemer
IntroductionInsights on BMIs
Previous work
Goal of the project
Challenges
Long term goals
ResultsWebots model
Numerical tests
CPG model
Results
Futureimprovements
References
Previous work
Fitzsimmons’ work on monkeys1 :Model try to predict the position of each jointQuality of predictions is good for one direction butdegrades for two directionsMust switch between two models for forward andbackward
Control ofbiped
locomotioninspired from
walking inmonkeys
Charles-HenryHoudemer
IntroductionInsights on BMIs
Previous work
Goal of the project
Challenges
Long term goals
ResultsWebots model
Numerical tests
CPG model
Results
Futureimprovements
References
Experimental setup
Control ofbiped
locomotioninspired from
walking inmonkeys
Charles-HenryHoudemer
IntroductionInsights on BMIs
Previous work
Goal of the project
Challenges
Long term goals
ResultsWebots model
Numerical tests
CPG model
Results
Futureimprovements
References
Achievements
Control ofbiped
locomotioninspired from
walking inmonkeys
Charles-HenryHoudemer
IntroductionInsights on BMIs
Previous work
Goal of the project
Challenges
Long term goals
ResultsWebots model
Numerical tests
CPG model
Results
Futureimprovements
References
Goal of the project
Build a CPG with the data from Nicolelis Lab in DukeUniversityData consists in Cartesian positions of hip, knee andankle positions, speed of walking and indication if thefoot is on the groundReproduce the walking gait of a rhesus monkey withthis CPG recieving basic inputs from a BMIGeneralize it if possible to :
Different sessions;Different speeds;Different monkeys;Both directions.
Control ofbiped
locomotioninspired from
walking inmonkeys
Charles-HenryHoudemer
IntroductionInsights on BMIs
Previous work
Goal of the project
Challenges
Long term goals
ResultsWebots model
Numerical tests
CPG model
Results
Futureimprovements
References
Goal of the project
Build a CPG with the data from Nicolelis Lab in DukeUniversityData consists in Cartesian positions of hip, knee andankle positions, speed of walking and indication if thefoot is on the groundReproduce the walking gait of a rhesus monkey withthis CPG recieving basic inputs from a BMIGeneralize it if possible to :
Different sessions;Different speeds;Different monkeys;Both directions.
Control ofbiped
locomotioninspired from
walking inmonkeys
Charles-HenryHoudemer
IntroductionInsights on BMIs
Previous work
Goal of the project
Challenges
Long term goals
ResultsWebots model
Numerical tests
CPG model
Results
Futureimprovements
References
Goal of the project
Build a CPG with the data from Nicolelis Lab in DukeUniversityData consists in Cartesian positions of hip, knee andankle positions, speed of walking and indication if thefoot is on the groundReproduce the walking gait of a rhesus monkey withthis CPG recieving basic inputs from a BMIGeneralize it if possible to :
Different sessions;Different speeds;Different monkeys;Both directions.
Control ofbiped
locomotioninspired from
walking inmonkeys
Charles-HenryHoudemer
IntroductionInsights on BMIs
Previous work
Goal of the project
Challenges
Long term goals
ResultsWebots model
Numerical tests
CPG model
Results
Futureimprovements
References
Goal of the project
Build a CPG with the data from Nicolelis Lab in DukeUniversityData consists in Cartesian positions of hip, knee andankle positions, speed of walking and indication if thefoot is on the groundReproduce the walking gait of a rhesus monkey withthis CPG recieving basic inputs from a BMIGeneralize it if possible to :
Different sessions;Different speeds;Different monkeys;Both directions.
Control ofbiped
locomotioninspired from
walking inmonkeys
Charles-HenryHoudemer
IntroductionInsights on BMIs
Previous work
Goal of the project
Challenges
Long term goals
ResultsWebots model
Numerical tests
CPG model
Results
Futureimprovements
References
Challenges
Monkeys are not naturally bipedalMonkeys’ behaviour is not always predictable
Play with holding barMake feet slide on the groundVideoCauses variability in measurements
Control ofbiped
locomotioninspired from
walking inmonkeys
Charles-HenryHoudemer
IntroductionInsights on BMIs
Previous work
Goal of the project
Challenges
Long term goals
ResultsWebots model
Numerical tests
CPG model
Results
Futureimprovements
References
Typical variations in the hip angles
Control ofbiped
locomotioninspired from
walking inmonkeys
Charles-HenryHoudemer
IntroductionInsights on BMIs
Previous work
Goal of the project
Challenges
Long term goals
ResultsWebots model
Numerical tests
CPG model
Results
Futureimprovements
References
Long term goals
Allow paraplegics to regain the control of their legsAllow for more robust predictions
Less informations are neededLess neurons are neededMay lose more neurons before losing efficency
Control ofbiped
locomotioninspired from
walking inmonkeys
Charles-HenryHoudemer
IntroductionInsights on BMIs
Previous work
Goal of the project
Challenges
Long term goals
ResultsWebots model
Numerical tests
CPG model
Results
Futureimprovements
References
Long term goals
Allow paraplegics to regain the control of their legsAllow for more robust predictions
Less informations are neededLess neurons are neededMay lose more neurons before losing efficency
Control ofbiped
locomotioninspired from
walking inmonkeys
Charles-HenryHoudemer
IntroductionInsights on BMIs
Previous work
Goal of the project
Challenges
Long term goals
ResultsWebots model
Numerical tests
CPG model
Results
Futureimprovements
References
Long term goals
Allow paraplegics to regain the control of their legsAllow for more robust predictions
Less informations are neededLess neurons are neededMay lose more neurons before losing efficency
Control ofbiped
locomotioninspired from
walking inmonkeys
Charles-HenryHoudemer
IntroductionInsights on BMIs
Previous work
Goal of the project
Challenges
Long term goals
ResultsWebots model
Numerical tests
CPG model
Results
Futureimprovements
References
Creation
Built from the webots human model of Jesse van denKieboom
Only hand size and total weight were knownSize and masse extrapolated with data from Hamada2
Main differences :Tail added;Head lighter than with normal division to reflect that amonkey head is lighter than a human one.
Uses a grannywalker to prevent it from falling on its side
Control ofbiped
locomotioninspired from
walking inmonkeys
Charles-HenryHoudemer
IntroductionInsights on BMIs
Previous work
Goal of the project
Challenges
Long term goals
ResultsWebots model
Numerical tests
CPG model
Results
Futureimprovements
References
Monkey and human models
Control ofbiped
locomotioninspired from
walking inmonkeys
Charles-HenryHoudemer
IntroductionInsights on BMIs
Previous work
Goal of the project
Challenges
Long term goals
ResultsWebots model
Numerical tests
CPG model
Results
Futureimprovements
References
Curves were normalized for testingCPGs can be easily scaled but not change shape
For each cycle :Amplitude was normalizedy = (y − y)/(ymax − ymin)Frequency was normalized by scaling each time intervalto a reference one
Control ofbiped
locomotioninspired from
walking inmonkeys
Charles-HenryHoudemer
IntroductionInsights on BMIs
Previous work
Goal of the project
Challenges
Long term goals
ResultsWebots model
Numerical tests
CPG model
Results
Futureimprovements
References
Duty factor
Duty factorDecreases with speedRight leg different than left oneVariance lower with high speeds
Equation of the form (for one cycle) :stancetimetotaltime
Control ofbiped
locomotioninspired from
walking inmonkeys
Charles-HenryHoudemer
IntroductionInsights on BMIs
Previous work
Goal of the project
Challenges
Long term goals
ResultsWebots model
Numerical tests
CPG model
Results
Futureimprovements
References
Duty factor
Duty factorDecreases with speedRight leg different than left oneVariance lower with high speeds
Equation of the form (for one cycle) :stancetimetotaltime
Control ofbiped
locomotioninspired from
walking inmonkeys
Charles-HenryHoudemer
IntroductionInsights on BMIs
Previous work
Goal of the project
Challenges
Long term goals
ResultsWebots model
Numerical tests
CPG model
Results
Futureimprovements
References
Duty factor
Duty factorDecreases with speedRight leg different than left oneVariance lower with high speeds
Equation of the form (for one cycle) :stancetimetotaltime
Control ofbiped
locomotioninspired from
walking inmonkeys
Charles-HenryHoudemer
IntroductionInsights on BMIs
Previous work
Goal of the project
Challenges
Long term goals
ResultsWebots model
Numerical tests
CPG model
Results
Futureimprovements
References
Duty factor
Duty factorDecreases with speedRight leg different than left oneVariance lower with high speeds
Equation of the form (for one cycle) :stancetimetotaltime
Control ofbiped
locomotioninspired from
walking inmonkeys
Charles-HenryHoudemer
IntroductionInsights on BMIs
Previous work
Goal of the project
Challenges
Long term goals
ResultsWebots model
Numerical tests
CPG model
Results
Futureimprovements
References
Control ofbiped
locomotioninspired from
walking inmonkeys
Charles-HenryHoudemer
IntroductionInsights on BMIs
Previous work
Goal of the project
Challenges
Long term goals
ResultsWebots model
Numerical tests
CPG model
Results
Futureimprovements
References
Square errors
Square errorsSteady stepMean error lowVariance relatively low
Equation of the form (for one cycle) :∑ni=1(yiknown − yipredicted )2
Control ofbiped
locomotioninspired from
walking inmonkeys
Charles-HenryHoudemer
IntroductionInsights on BMIs
Previous work
Goal of the project
Challenges
Long term goals
ResultsWebots model
Numerical tests
CPG model
Results
Futureimprovements
References
Square errors
Square errorsSteady stepMean error lowVariance relatively low
Equation of the form (for one cycle) :∑ni=1(yiknown − yipredicted )2
Control ofbiped
locomotioninspired from
walking inmonkeys
Charles-HenryHoudemer
IntroductionInsights on BMIs
Previous work
Goal of the project
Challenges
Long term goals
ResultsWebots model
Numerical tests
CPG model
Results
Futureimprovements
References
Square errors
Square errorsSteady stepMean error lowVariance relatively low
Equation of the form (for one cycle) :∑ni=1(yiknown − yipredicted )2
Control ofbiped
locomotioninspired from
walking inmonkeys
Charles-HenryHoudemer
IntroductionInsights on BMIs
Previous work
Goal of the project
Challenges
Long term goals
ResultsWebots model
Numerical tests
CPG model
Results
Futureimprovements
References
Square errors
Square errorsSteady stepMean error lowVariance relatively low
Equation of the form (for one cycle) :∑ni=1(yiknown − yipredicted )2
Control ofbiped
locomotioninspired from
walking inmonkeys
Charles-HenryHoudemer
IntroductionInsights on BMIs
Previous work
Goal of the project
Challenges
Long term goals
ResultsWebots model
Numerical tests
CPG model
Results
Futureimprovements
References
Square errors of Nectarine - Hip
Control ofbiped
locomotioninspired from
walking inmonkeys
Charles-HenryHoudemer
IntroductionInsights on BMIs
Previous work
Goal of the project
Challenges
Long term goals
ResultsWebots model
Numerical tests
CPG model
Results
Futureimprovements
References
Square errors of Nectarine - Knee
Control ofbiped
locomotioninspired from
walking inmonkeys
Charles-HenryHoudemer
IntroductionInsights on BMIs
Previous work
Goal of the project
Challenges
Long term goals
ResultsWebots model
Numerical tests
CPG model
Results
Futureimprovements
References
Design of the CPGs
Data were in Cartesian form and contained thepositions of the knee, hip and ankleCPG is of the form :
xi = γi (fi (θi ) − xi ) + dfidθi
· θi + Ki
Phase coupling is of the form :p = sin(from.p ∗ 2 ∗ π − to.p ∗ 2 ∗ π − bias)
A cycle was arbitrarily chosen for the building of theCPG
AverageStarts and ends approximatively at the same height
Control ofbiped
locomotioninspired from
walking inmonkeys
Charles-HenryHoudemer
IntroductionInsights on BMIs
Previous work
Goal of the project
Challenges
Long term goals
ResultsWebots model
Numerical tests
CPG model
Results
Futureimprovements
References
Design of the CPG
Joint angle cycle was cut in three parts for interpolationParts were interpolated with third degree polynomialsPolynomials were used with cpgstudioTesting on webots was done with the cpg file obtainedthrough cpgstudio and the libcpg, both from Jesse vanden KieboomThe necessary data for building a CPG for the ankleand adapting the foot were obtained by trackingpositions on a video with the help of KostasKarakasiliotis. The implementation of the CPG wasabandoned in order not to add more complexity to analready well performing model.
Control ofbiped
locomotioninspired from
walking inmonkeys
Charles-HenryHoudemer
IntroductionInsights on BMIs
Previous work
Goal of the project
Challenges
Long term goals
ResultsWebots model
Numerical tests
CPG model
Results
Futureimprovements
References
CPG output
Control ofbiped
locomotioninspired from
walking inmonkeys
Charles-HenryHoudemer
IntroductionInsights on BMIs
Previous work
Goal of the project
Challenges
Long term goals
ResultsWebots model
Numerical tests
CPG model
Results
Futureimprovements
References
Real and interpolated curves - Hip
Control ofbiped
locomotioninspired from
walking inmonkeys
Charles-HenryHoudemer
IntroductionInsights on BMIs
Previous work
Goal of the project
Challenges
Long term goals
ResultsWebots model
Numerical tests
CPG model
Results
Futureimprovements
References
Real and interpolated curves - Knee
Control ofbiped
locomotioninspired from
walking inmonkeys
Charles-HenryHoudemer
IntroductionInsights on BMIs
Previous work
Goal of the project
Challenges
Long term goals
ResultsWebots model
Numerical tests
CPG model
Results
Futureimprovements
References
Visual tests
Can walk forward and backwardHandles different speeds
Limited by ground adherence
Video
Control ofbiped
locomotioninspired from
walking inmonkeys
Charles-HenryHoudemer
IntroductionInsights on BMIs
Previous work
Goal of the project
Challenges
Long term goals
ResultsWebots model
Numerical tests
CPG model
Results
Futureimprovements
References
Future improvements
Implement flexible feet for more efficient foot contactModify the polynomials to reduce the error ininterpolation further in order to obtain a interpolatedcurve even more close to a real one
Control ofbiped
locomotioninspired from
walking inmonkeys
Charles-HenryHoudemer
IntroductionInsights on BMIs
Previous work
Goal of the project
Challenges
Long term goals
ResultsWebots model
Numerical tests
CPG model
Results
Futureimprovements
References
References
1 Nathan A. Fitzsimmons, Mikhail A. Lebedev, Ian D. Peikon,andMiguel A. L. Nicolelis. Extracting kinematic parameters for monkeybipedal walking from cortical neuronal ensemble activity.
2 Yuzuru Hamada, Nontakorn Urasopon, Islamul Hadi, and SuchindaMalaivijitnond. Body Size and Proportions and Pelage Color ofFree-Ranging Macaca mulatta from a Zone of Hybridization inNortheastern Thailand.