Robert  RienerSensory-­‐Motor  Systems  LabIns1tute  of  Robo1cs  and  Intelligent  Systems,  ETH  ZurichUniversity  Hospital  Balgrist,  University  of  Zurich

Design  Principles  for  Intelligent  

Rehabilita5on  Robots  

ShanghAI  Lecture  Series

15  November  2012,  10:00  AM  CET

For  the  Therapist• Physically  exhaus.ng

• Ergonomically  inconvenient

For  the  Pa1ent• Limited  training  dura.on

• Gait  pa;ern  not  op.mal

SCI  Center,  Balgrist

Disadvantage  of  Manual  Training  

Human-­‐Robot  Gait  Rehabilita1on

Gait  Rehabilita1on  "Robots"GaitTrainer

Hap.c  Walker

G-­‐EO

Lopes

Autoambulator

HAL

ETH  Zurich/Balgrist,  Hocoma  AG

Exoskeletal  Robot  with  7  Degrees  of  Freedom

Nef,  Riener  et  al.  2006-­‐2011

ARMin  III

Arm  Rehabilita1on  RobotsMIT  Manus Bi-­‐Manu-­‐Track

Hap.c  Master,  GENTLE/s

MGA

PERCRO  Exoskeleton

Salford  PMA

KIST  Arm

Sensory-­‐Motor  Interac1on

Human Machine

Interac.on

Focus:    Exoskeleletons  for  Clinical  RehabilitaKon

Setups:    Lokomat,  ARMin

Sensory-­‐Motor  Characteris1cs

To  Be  Known  about  Human  Physiology

• Nervous  system  is  plas.c

To  Be  Known  about  Human  Physiology

Sensory-­‐Motor  Characteris1cs

• Nervous  system  is  plas.c

To  Be  Known  about  Human  Physiology

Sensory-­‐Motor  Characteris1cs

• Nervous  system  is  plas.c

• Par.cipa.on  increases  plas.city

Passive  vs  Ac1ve  Training

Limita1ons  of  Passive  Training  of  Healthy  Subjects• Physical  guidance  hinders  motor  learning  of  walking  balance                                        Domingo  &  

Ferris,  2009

• High  frequency  guiding  is  detrimental  for  learning  of  arm  movements  (guidance  hypothesis)                                                        Winstein  et  al.  1994;  Marchal-­‐Crespo  &  Reinkensmeyer  2008

Assist-­‐as-­‐Needed  Training  (ANN)• AAN  shows  higher  level  of  recovery  step  number,  periodicity,  and  consistency  (27  mice)                              Cai  et  al.  2006

Lokomat:  Pa1ent-­‐Coopera1ve  Control

Path  Control

• Robot  behaves  assis.ve,  correc.ve  or  transparent,  when  needed

• Free  .ming  for  pa.ent

• Support  pa.ent,  but  do  not  restrict  pa.ent

Duschau-­‐Wicke,  Vallery,  Riener  2009,  2010,  2011

Heart  Rate

Posi.oncontrol

Path  control

Muscle  AcKvity

0.06

0.08

0.1

0.12

0.14

0.16

0.18

0.2

0.22

0.24

Init. loading

Mid stance

Term. stance

Pre swing

Init. swing

Mid Swing

Term. swing

Nor

mal

ized

mus

cle

activ

ity (B

F)

Position controlPath control

Rela1ve  increase  of  h

eart  ra

te

11  incomplete  SCI  subjects

Path  Control  Increases  ParKcipaKon

To  Be  Known  about  Human  Physiology

Sensory-­‐Motor  Characteris1cs

• Nervous  system  is  plas.c

• Par.cipa.on  increases  plas.city

• Repe..on  without  repe..on

Path  Control  Enhances  Variability

Posi.on  Control

-20 -10 0 10 20 30 400

10

20

30

40

50

60

70

80

Hip angle [°]

Knee

ang

le [°

]

Path  Control

-20 -10 0 10 20 30 400

10

20

30

40

50

60

70

80

Hip angle [°]

Knee

ang

le [°

]

“RepeKKon  without  RepeKKon”

To  Be  Known  about  Human  Physiology

Sensory-­‐Motor  Characteris1cs

• Nervous  system  is  plas.c

• Par.cipa.on  increases  plas.city

• Repe..on  without  repe..on

• Provide  large  ROM,  avoid  joint  stress  

Main  Problem  with  Exoskeletons

Alignment  of  RoboKc  and  Human  Joints

Human

Robot

Main  Problem  with  Exoskeletons

Alignment  of  RoboKc  and  Human  Joints

Human

Robot

Human

Robot

Main  Problem  with  Exoskeletons

Alignment  of  RoboKc  and  Human  Joints

Human

Robot

Human

Robot

Humerus  MoKon

x

y

GH

SC

AC

φ

Δy  = 124  mm

Δx  =  28  mm

Body  height:  1.7  m

Humerus  MoKon

Ver1cal  displacementof  main  drive  

Ver1cal  shoulder  displacement

ARMin  II  Shoulder  Kinema1cs

ARMin  III:  Novel  Shoulder  Actua1on

Humerus  Head

ARMin  III:  Novel  Shoulder  Actua1on

Humerus  Head

COR

ARMin  III:  Novel  Shoulder  Actua1on

Fixa1on  screw

Axis  1

Axis  2

Axis  3

Axis  4α

Force/torque  sensor

2  Laser  pointers

Sensory-­‐Motor  Characteris1cs

To  Be  Known  about  Human  Physiology

• Nervous  system  is  plas.c

• Par.cipa.on  increases  plas.city

• Repe..on  without  repe..on

• Provide  large  ROM,  avoid  joint  stress  

• Each  pa.ent  is  different

ARMin  III:  Adjustments

Reference  trajectory

Actual  trajectory

Adapt  Field  with  a  IteraKve  Learning  Controller

Lokomat:  AdapKve  Force  Field

To  Be  Known  about  Human  Physiology

Sensory-­‐Motor  Characteris1cs

• Nervous  system  is  plas.c

• Par.cipa.on  increases  plas.city

• Repe..on  without  repe..on

• Provide  large  ROM,  avoid  joint  stress  

• Each  pa.ent  is  different

• Mo.va.on  increases  plas.city

Human-­‐Robot  Coopera1on

   Audiovisual  &  hap.c  displays  

     Task  specific      controller

S1muli

             Force  and  posi.on  recordings              

         Psychophysiological  recordings            

To  Be  Known  about  Human  Physiology

Sensory-­‐Motor  Characteris1cs

• Nervous  system  is  plas.c

• Par.cipa.on  increases  plas.city

• Repe..on  without  repe..on

• Provide  large  ROM,  avoid  joint  stress  

• Each  pa.ent  is  different

• Mo.va.on  increases  plas.city

• Train  ac.vi.es  of  daily  living

Task  Specific  Training

Chronic  Stroke  PaKent  (FMA=26)

Task  Specific  Training

Conclusion  and  Outlook

Taking  into  Account  Human  Physiology• Robots  can  allow  efficient  &  intensive  &  individual  training

• Robot  can  cooperate  to  keep  the  pa.ent  ac.ve

• Robots  can  mo.vate  the  pa.ent

Chances• Improve  health  status  and  quality  of  life  (pa.ents)

• Reduce  workload  of  clinical  staff  (therapists,  nurses)

To  do• Clinical  evalua1on  studies  on  pa1ents