Present Study

1
Patient Specific Motion Modeling and Assistive
Devices
S Russell, P Sheth, B Bennett, P Allaire, M Abel
University of Virginia, Motion Analysis and Motor
Performance Laboratory, Charlottesville, VA.
INTRODUCTION
RESULTS (cont)
METHODS (cont)
Angular Momentum Control

• Foot Model
• Previous model consisted of
• only heel and toe ellipsoids
• Resulted in piece wise motion
• of modeled CoP
• Non-smooth GRF
• New model includes contact
• ellipsoids along metatarsals
• connecting heel/toe ellipsoids
• Smooth/continuous GRF
• Facilitates smooth CoP motion in
• both sagittal and frontal planes

• Angular Momentum used to develop stable walking
  in bipedal gait models (Goswami, Popovic)
• Previously used to
• Determine energy lost at heel strike/foot contact
• Control strategy for full body model in single
  support

Present Study

• Develop patient specific full body gait model
• Develop applied joint torques from desired
  Angular Momentum
• Implementation of newly developed foot model
• Application and validation of Optimized contact
  model between foot and floor

Angular Momentum Control

• Angular momentum of each segment about the body
  CoM calculated and averaged for each test subject
• Data used to populate 19xN matrix A where
• N data points over gait cycle
• PD control used to determine joint torques for
  each point of gait cycle, 1,2,N
• Simulations run using angular momentum A as
  negative feedback for PD control resulting in
  minimized error between desired and simulated
  angular momentum

Comparison of actual patient vertical ground
reaction force, measured experimentally, and the
vertical component of the ground reaction force
predicted by the model using Angular momentum
control and the new foot and ground contact model.
METHODS
DISCUSSION
Tested Subjects

• Stable walking patterns are predicted using the
  angular momentum about the full body CoM as the
  high level control
• Analysis found angular momentum computed from
  experimentally measured kinematics was invariant
  with speed ( 20)
• Consistent with published data (Popovic)
• Allows simulation of various speed from same
  angular momentum control matrix A
• Subject specific models make no assumption
  regarding symmetry and model entire gait cycle
• Our research has found that children with
  Cerebral Palsy walk with angular momentum
  patterns similar to typical walkers
• Extension of this work is under way
• Full 3-D simulations
• Prediction pathologic gait

• 3-D kinematics collected via Vicon implementing a
  38 marker Helen-Hayes full body set
• 3 subjects with no history of lower extremity
  pathology were used to develop/validate patient
  specific models

RESULTS

• Simulations completed successfully implementing,
  patient specific model, foot/floor contact model,
  and PD control for both single and double support
  phase
• Model successfully predicted gait kinematics and
  ground reaction force for individual patients

Model Development

• Full body model developed for
• each patient via MSC.Adams
• using LifeMod
• Patient anthropometrics used
• with the GeBOD data base to
• create patient specific models
• Model
• 19 body segments
• 16 joints
• Full 3-D with motion
• constrained to sagittal plane
• Ground Contact
• Ground placement optimized resulting in GRF
  equaling Body weight
• Coulomb friction applied between floor and foot
  geometries
• Contact parameters (stiffness/damping) optimized
  in both planes to match experimentally measured
  GRF
• Similar solution parameters suggest optimization
  process is not mandatory

References
Goswami, A., Kallen, V. (2004). Proc IEEE ICRA
04, 3785-3790. Popovic, M., et al. (2004). Proc
IEEE ICRA 04, 2405-2411. Popovic, M., et al.
(2004). Proc IEEE/RSJ IROS 04, 1685-1691.
Acknowledgements
The authors would like to thank the staff at the
Gait and Motion Analysis Lab, Kluge Childrens
Rehabilitation Center at the University of
Virginia, where experiments were conducted. This
work was supported in part by NSF grant 0503256.
Plots show lower extremity joint kinematics of a
single patient. Results shown include
experimentally measured joint angles for the
ankle, knee, and hip joint. Also shown are the
lower extremity joint angles predicted using the
patient specific model, revised foot model,
improved floor contact model, and angular
momentum based PD control for joint torques
2
Patient Specific Motion Modeling and Assistive
Devices
S Russell, P Sheth, B Bennett, P Allaire, M Abel
University of Virginia, Motion Analysis and Motor
Performance Laboratory, Charlottesville, VA.
INTRODUCTION
Present Studies
Cerebral Palsy

• Develop plantar flexion assist ankle foot
  orthosis (AFO) to promote heel strike while
  facilitating 3 rockers of stance
• Develop passive brace to store energy lost at
  heel strike and return it during pre swing
• Develop motorized walker to predict gait events
  and assist subjects in walking and turning

• 764,000 people in the United States have symptoms
  of Cerebral Palsy (CP)
• Metabolic costs of walking 2-3 times higher in
  individuals with CP
• 50 of people with CP are prescribed Ankle Foot
  Orthotics (AFO)
• Previous research equivocal on effectiveness of
  AFOs

Smart Walker
Energy Return AFO
Plantar Flexion Assist AFO
Objectives
Objectives
Objectives

• Develop an AFO which limits plantar flexion
  during swing due to conditions such as equinus or
  drop foot
• Allow patient kinematics to exploit all three
  rockers during stance (Figure 1)

• Energy added from ankle plantar flexion at push
  off for CP gait typically 15-20 less than normal
  gait
• Develop an AFO to store the energy lost during
  heel strike and return the energy later in gait
  cycle (i.e. push off)
• Return .4 J/kg of energy to gait at push off (15
  total normal energy added at push off)
• Allow patient kinematics to exploit all three
  rockers during stance (Figure 1)

• Preform real-time prediction of gait events (i.e.
  heel strike, toe off) via forces applied to
  walker handles
• Create shared control of a motorized posterior
  walker to facilitate better walking in children
  with CP

Previous Solutions
Figure 1. 1st Rocker represents rotation about
the heel at initial heel contact allowing the
foot to lay flat, 2nd Rocker allows the body to
progress forward rotating about the ankle with
the foot flat in stance, and the 3rd rocker
allows the subject to rotate onto the ball of
there foot to facilitate push off during pre
swing.

• Post-processing prediction of gait events from
  handle forces validated using VICON motion
  analysis
• Implementation of shared control on steering
  angle to control anterior walker trajectories

Current AFOs

• Solid ankle AFOs are unable to return energy to
  gait cycle while PLS, ground reaction, and carbon
  toe off AFOs store and return energy at push off
  but do so by inhibiting the 2nd rocker

Current AFOs

• Solid ankle AFOs and posterior leaf spring (PLS)
  restrict or inhibit one or more kinematic rockers
• Hinged AFOs facilitate rockers but cannot
  inhibit foot drop or equinus

Current Solution Design
Current Solution Design

• Based on a double upright AFO with dual action
  joints
• Patient specific conical compression springs
  located between foot bed and sole of shoe
• Ankle plantar flexion compresses the springs via
  the cable, moment arm, and pulleys
• Spring tension is adjusted to apply patient
  specific plantar flexion assist during swing
  phase
• In stance springs are compressed as weight is
  transferred forward releasing tension on moment
  arm facilitating full range of motion and all 3
  rockers

• Development of real time prediction of Gait
  events from forces applied to walker handles
  during walking
• Development of shared control algorithm to
  control electric motors for desired walker motion
• Hold walker position fixed in cases of impending
  fall (instability)
• Interject energy during strategic gait events
  (push off)
• Facilitate directional control of walker
• Negate additional work of dragging a walker
  during gait

• An additional benefit of the design is the energy
  return applied during pre swing to aid push off
  as the springs decompress as body weight is
  removed

Current Solution Design

• Energy stored at heel strike
• Body weight compresses springs
• Springs held in compression by multi-tooth
  ratchet system
• Tension in cables released to facilitate full
  range of motion
• Energy returned during push off
• Body rotates forward over foot (2nd rocker)
• Foot is dorsi-flexed taking slack out of cable
• Weight in on ball of foot activating mechanism to
  release springs
• Springs create plantar flexion moment about ankle
  joint via cables