kim

Document Sample
kim Powered By Docstoc
					 DYNAMIC MOTION PREDICTION AND
ENERGY LEVEL DETERMINATION FOR
 A VIRTUAL SOLDIER’S UPPER-BODY
  1Kim,   J.H., 1Abdel-Malek, K., 1Yang, J., and 2Nebel, K.

    1Virtual  Soldier Research (VSR) Program
          Center for Computer-Aided Design
               The University of Iowa
            2U.S.Army TACOM/RDECOM
                            Virtual Soldier Research
  Virtual Soldier Research Program
       at The University of Iowa




• Funded primarily by the US Army, we conduct basic and
  applied research for creating new technologies dealing
  with digital human modeling and simulation. We are a
  group of 36 people (faculty, staff, scientists, engineers,
  clinical researchers, and graduate students) that have
  come together to create this new technology.
                                                   Our strengths and areas of expertise
               Real-time Simulation/gaming
               Real-time rendering/visualization                  Human Motion Prediction
                                                                  Kinematics (postures/trajectories)



     Task-based kinematics                                                 Balance and instability
     Task execution                                                        Dynamics and environment


                                                                                Real-time dynamics
 EMG                                                                            Performance
 Which muscles are active when!



                                                                      Muscle Fatigue
 Joint Stress                                                         Hills Models, energy
 Pain/injury prediction


                                                                   Clothing
Real-time optimization                                             Mathematical Modeling
Genetics and gradient based


                                                               Multi-objective Optimization
    Motion Capture                                             many human performance measures
    Model verification

                              Muscle contraction       Force Feedback Control
                              FEM models               Haptics
                             Motivations

    Motion Prediction for Various Tasks

• pulling a lever
• lifting an object
• turning a steering wheel
• pulling trigger
• moving tools
• pushing a button
• etc.
                             Basic Assumptions

Upper-Body Motion with External Loads

                 Our Assumptions
 Different tasks lead to different external loads and
paths (at the end-effector, or hand)
 Different external loads lead to different
motion/posture
 Human acts in such a way as to minimize certain
cost functions – human performance measures.
                                       Modeling
          Generalized Coordinates and
             Generalized Torques
                                        SANT
                                        OSTM


                                            Denavit-Hartenberg Representation
                                                          x ( q* )  0                     n x
                                                                     Tn ( q1 ,..., qn )  
                                                             1                           1
• Multiple muscles
contribute to a single
                                                     0
                                                         Tn (q )  0 T1 ( q1 )1 T1 ( q2 )...n 1 Tn ( qn )
degree-of-freedom joint
                                      L
                                      cos
                                      M    i    cos  i sin  i      sin  i sin  i    ai cos i    O
                                                                                                       P
motion.                                          cos  i cos i        sin  i cos i    ai sin  i
                                   T M                                                                P
                            i 1
                                      sin   i


• Complex muscle
                                   i
                                      M0
                                      M0            sin  i               cos  i             di       P
                                                                                                       P
configuration, which also
                                      N                0                     0                1        Q
varies as links move.
                                     Muscle components

                 Muscle Joint System

• Muscle (and tendon) = contractile components + elastic components
• Examples:
Hill’s muscle model (Hill, 1938),
Zajac’s Muscle model (Zajac, 1989)
• Joint torque vector due to
 muscle elasticity:


      τ  K  q - q N                   Typical Muscle Model
                               Equations of Motion

            Joint Torque Prediction

• Equation of motion                           q3
                                        b                    Fy, My
derived from Lagrangian        q2
mechanics                                           c
                               a
                                                                 Fx, Mx

                           s    q1
• General external loads                            Fz, Mz
are applied at the end-
effector (hand)
                                             Equations of Motion

 Dynamic Equations of Motion



                          Fy, My



                              Fx, Mx

             Fz, Mz


τ = M(q) q+ V(q,q) +  J i T mi g +  J k TFk  K q - q N 
   mass &inertia                       i                 k
                      Coriolis &                                          muscle elasticity
   matrix             Centrifugal      gravity forces   external forces
                   Muscle Energy

Muscle Energy Consumption


“Total muscle energy consumption
  = Work done by joint torques + Heat”


         Emuscle  Wjoint  Qin
                           Muscle Energy

  Muscle Energy Expenditure Rate


Energy rate as a function of each time instant:

               Emuscle  Wjoint  Qin
                           Muscle Energy

                Why Energy?
         Using the first law of thermodynamics

oxygen                               mechanical work
&                                    &
food                                 heat
                         Muscle Energy

     Mathematical Formulation of
        Energy Consumption
                                    
First law of thermodynamics: We  Qin  T  U
where,
 We : work done by external forces
   
 Qin : inlet heat energy
T : change of kinetic energy of the system
U : change of internal energy of the system
                                                 Muscle Energy

    Muscle Energy Rate in Joint Space


                              n            n            n
 Emuscle  Wjoint  Qin    i qi   hm  i   hsi  i qi  B
                                        i

                             i 1         i 1         i 1




where,
   i
 hm : coefficient of generalized maintenance heat of joint i
 hsi : coefficient of generalized shortening heat of joint i
 B : basal metabolic heat rate
                                       Optimization

Realistic Upper-Body Motion Prediction
                       Optimization algorithm

      Find Design Variables: pi , i  1, 2,...
                         (control points for B-Spline)
                                             Optimization

Realistic Upper-Body Motion Prediction
                         Optimization algorithm

      Find Design Variables: pi , i  1, 2,...
                         (control points for B-Spline)


      Minimize: Muscle Energy Consumption
                                   t2   n
        Or muscle work Wjoint     i (t )qi (t ) dt
                                   t1 i 1

     (assumption: nearly constant muscle mechanical efficiency)
                                              Optimization

Realistic Upper-Body Motion Prediction
                         Optimization algorithm

      Find Design Variables: pi , i  1, 2,...
                         (control points for B-Spline)


      Minimize: Muscle Energy Consumption
                                   t2   n
        Or muscle work Wjoint     i (t )qi (t ) dt
                                   t1 i 1

     (assumption: nearly constant muscle mechanical efficiency)



      Subject to:      joint limits, torque limits, etc.

                             qil  qi  qiu

         τ l  M(q)q+ V(q,q) +  J i T mi g +  J k T Fk  τ  τ u
                                   i             k
                                       Example

            Example of Motion Prediction




• SANTOSTM moving 5 Kg
weight from initial to final
position.



                               SANTOSTM moving an object
                                                                                                                      Example
                                  Inverse Dynamics of Example Task
                                         JOINT ANGLES OF RIGHT ARM

                                                                       joint13
                        1.5
Joint Angle (radians)




                                                                       joint14
                          1
                                                                       joint15
                        0.5
                                                                       joint16
                          0
                                                                       joint17
                        -0.5 0     0.5        1        1.5   2   2.5
                                                                       joint18
                         -1
                                                                       joint19
                        -1.5
                                              Time (sec.)
                                                                       joint20
                                                                                                                       JOINT TORQUE OF right ARM
                                                                                                                    Torque profiles for RIGHT upper limb
                                                                       joint21



                                                                                                       15000                                               joint13




                                                                                 Joint Torque (N.cm)
                                                                                                                                                           joint14
                                                                                                       10000
                                                                                                                                                           joint15
                                                                                                        5000                                               joint16
                                                                                                           0                                               joint17
                                                                                                                0      0.5    1       1.5   2    2.5       joint18
                                                                                                        -5000
                                                                                                                                                           joint19
                                                                                                       -10000                                              joint20
                                  JOINT ACCELERATIONS OF RIGHT ARM                                                            Time (sec.)                  joint21
                                                                       joint13
                           6
                                                                       joint14
Joint Acceleration




                           4
                                                                       joint15
  (rad/sec/sec)




                           2
                                                                       joint16
                           0
                                                                       joint17
                           -2 0     0.5       1        1.5   2   2.5
                                                                       joint18
                           -4
                                                                       joint19
                           -6
                                                                       joint20
                                              Time (sec.)
                                                                       joint21
                                       Example

       Local Biomechanical Analysis


Generalized joint torque from
Inverse Dynamics
                                          q1
(global motion analysis)
                                                 Fbiceps


                                                 Ftriceps
Muscle force distribution and configuration
(local biomechanical analysis)
                 Example

Muscle force Distribution
                                                           Example

Predicted Joint Mechanical Power
                                   MUSCLE POWER LEVEL

                  9000
                  8000
                  7000
 Power (N.cm/s)




                  6000
                  5000
                  4000
                  3000
                  2000
                  1000
                     0
                         0   0.5         1             1.5       2   2.5
                                             Time (sec.)


Joint mechanical power profiles for the example task
                             Example

  Prediction of Physiological Indexes
 Muscle Power Level
                       Monitoring
 Energy Consumption   Human Performance Measures
                        • Energy (Power)
                        • Discomfort
                        • Muscle fatigue
 Heart Rate            • Instability
                        • Effort
 Body Temperature      • Cardiovascular fatigue
                          (heart rate)
 Blood Pressure        • Biomechanical stress
 Water loss            • Etc.

 etc.
                      Motion/Posture prediction

              Conclusions
• Developed realistic human model called
  SantosTM with 89 degrees of freedom.
                       Motion/Posture prediction

               Conclusions
• Developed realistic human model called
  SantosTM with 89 degrees of freedom.
• Derived dynamic equation of motion and
  energy consumption formulation in joint space.
                       Motion/Posture prediction

               Conclusions
• Developed realistic human model called
  SantosTM with 89 degrees of freedom.
• Derived dynamic equation of motion and
  energy consumption formulation in joint space.
• Prediced realistic human motion based on
  energy minimization (less muscle fatigue).
                       Motion/Posture prediction

               Conclusions
• Developed realistic human model called
  SantosTM with 89 degrees of freedom.
• Derived dynamic equation of motion and
  energy consumption formulation in joint space.
• Prediced realistic human motion based on
  energy minimization (less muscle fatigue).
• Predicted joint torques and energy rate for
  biomechanical and physiological analysis.
    Motion/Posture prediction



Thank you
                    Motion/Posture prediction


        Presenter: Joo H. KIM

             Research Assistant
 Virtual Soldier Research (VSR) Program
Center for Computer-Aided Design (CCAD)
           The University of Iowa
            Iowa City, IA 52242
Tel: 319-384-0579      Fax: 319-384-0542,
       Email: joo-kim@uiowa.edu
      http://www.digital-humans.org/

				
DOCUMENT INFO