Control - ISIR

Multi‐modal control of an anthropomorphic hand in the p
p
context of grasping and in‐hand manipulation
context of grasping and in‐hand manipulation
Kien Cuong NGUYEN
kien‐[email protected]
ki
@
f
PhD St d t f
PhD Student of Prof Véronique Perdereau and Prof. Mohamed Abderrahim
Prof. Véronique
and Prof Mohamed Abderrahim
ISIR Intelligent Systems and Robotics Institute
ISIR ‐
I t lli t S t
d R b ti I tit t
Pierre and Marie Curie University of Paris France
Pierre and Marie Curie University of Paris, France
Main Objective
Simulation Tools
Control of
C
t l f an anthropomorphic hand th
hi h d Matlab, Marilou Simulation b d
based on a multi‐sensorial system for l
l
f
Environment and ODE – Open the tasks of grasping and in‐hand Dynamics Engine
y
g
manipulation. The work will be driven by quality criteria such as yq
y
dexterity, stability, compliance and de
te ty, stab ty, co p a ce a d
manipulability.
Shadow hand with diverse types of sensors: position, force, visual and p
,
,
tactile sensingg
• Position control
• Force – Torque control
q
• Tactile Sensing control
g
• Visual Servoingg
• Multi‐Finger Coordination
Control Parameters:
Control Parameters:
Learning g
Engine
Action Decomposition
Action Decomposition
The whole action that the control system is involved in can be y
decomposed into 5 basic tasks:
p
T d Di
Tendon‐Driven
Basic Bricks:
They are essentially the gains or They
are essentially the gains or
parameters that are specified based on
parameters that are specified based on robot–object characteristics control
robot–object characteristics, control strategy and some specific needs
strategy and some specific needs.
Experimental Platform
Experimental Platform
Force Sensor for f
each tendon
each tendon
Control Architecture
Control Architecture
Control Parameters
High‐Level Planning
Control Monitor
Position Control
Force Control
R hi
Reaching
In‐hand Manipulation
p
Tactile sensing
i
Camera System
Camera System
Position Sensor Position Sensor
for each joint
j
G
Grasping
i
Transporting
Releasing
Each basic task need a specific E hb i t k
d
ifi
control strategy.
l
www.handle‐project.eu
FP7‐2008 – 231640
Sensing System
Obstacles
Tactile Control
Visual Servoing
Visual Servoing
Object
b
Interaction with Learning & High‐level Planning:
High
level Planning:
g
p
Online Learningg & High‐level
planningg will
y
monitor the execution of the tasks by determining the type of control and the g
yp
corresponding parameters.
HANDLE: Developmental p
pathway towards autonomy
y
y
and dexterity in robot in‐hand
and dexterity in robot in
hand manipulation, is a Large Scale manipulation, is a Large Scale
Integrated Project coordinated by the University Pierre and
Integrated Project coordinated by the University Pierre and Marie Curie of Paris and includes a consortium formed by nine
Marie Curie of Paris and includes a consortium formed by nine partners from six EU countries: France UK Spain Portugal
partners from six EU countries: France, UK, Spain, Portugal, Sweden and Germany
Sweden and Germany.