Upload
others
View
1
Download
0
Embed Size (px)
Citation preview
Tam
im A
sfou
rU
nive
rsitä
tKa
rlsru
he
F 4.13 “F 4.13 “SteuerungSteuerung und und RegelungRegelung von von RoboternRobotern””24. 24. JanuarJanuar 2007, Karlsruhe 2007, Karlsruhe
Mechatronics und Bewegungssteuerung des humanoiden Roboters ARMAR-III
T. Asfour and R. Dillmann
University of Karlsruhe (TH)Institute of Computer Science and Engineering (CSE)
http://wwwiaim.ira.uka.de
http://sfb588.uni-karlsruhe.de
Tam
im A
sfou
rU
nive
rsitä
tKa
rlsru
he
F 4.13 “F 4.13 “SteuerungSteuerung und und RegelungRegelung von von RoboternRobotern””24. 24. JanuarJanuar 2007, Karlsruhe 2007, Karlsruhe
SFB 588: Project Research Topics
Cooperation
Learning
MultimodaleInteraction
Programming by Demonstration
InteractiveLearning
Human-RobotInteraction
Analysis of Human Motion
Speech
Gesture/mimic
TactileInteraction
Mechatronics and Motor Control
Tam
im A
sfou
rU
nive
rsitä
tKa
rlsru
he
F 4.13 “F 4.13 “SteuerungSteuerung und und RegelungRegelung von von RoboternRobotern””24. 24. JanuarJanuar 2007, Karlsruhe 2007, Karlsruhe
Outline of the talk
The humanoid robot ARMAR-IIIMechatronicsHardware architecture Software framework
Implemented Skills • Object recognition and localization • Inverse Kinematics• Path Planning/Collision detection
Tam
im A
sfou
rU
nive
rsitä
tKa
rlsru
he
F 4.13 “F 4.13 “SteuerungSteuerung und und RegelungRegelung von von RoboternRobotern””24. 24. JanuarJanuar 2007, Karlsruhe 2007, Karlsruhe
Kinematics of upper body
40Upper body
446
3
43
16
222
1
1
2
223
3
4
8
WristElbowShoulder
Torso
NeckEyes
Hand
totalamountD.O.FJoint
θ1
θ7θ6
θ4
θ2
θ11
θ9
θ10
θ8
θ5
θ3
θ12
θ13
θ14θ1
θ7θ6
θ4
θ2
θ11
θ9
θ10
θ8
θ5
θ3
θ12
θ13
θ14
Tam
im A
sfou
rU
nive
rsitä
tKa
rlsru
he
F 4.13 “F 4.13 “SteuerungSteuerung und und RegelungRegelung von von RoboternRobotern””24. 24. JanuarJanuar 2007, Karlsruhe 2007, Karlsruhe
Range of motion of upper body
θ1
θ7θ6
θ4
θ2
θ11
θ9
θ10
θ8
θ5
θ3
θ12
θ13
θ14θ1
θ7θ6
θ4
θ2
θ11
θ9
θ10
θ8
θ5
θ3
θ12
θ13
θ14
-180° to 180°-10° to 60°-20° to 20°
θ12θ13θ14
Torso
-180° to 180°-45° to 45°-45° to 45°-60° to 60°
θ8θ9θ10θ11
Neck
-180° to 180°-45° to 180°-10° to 180°
θ5θ6θ7
Shoulder
-90° to 90°-10° to 150°
θ3θ4
Elbow
-30° to 30°-60° to 60°
θ1θ2
Wrist
Tam
im A
sfou
rU
nive
rsitä
tKa
rlsru
he
F 4.13 “F 4.13 “SteuerungSteuerung und und RegelungRegelung von von RoboternRobotern””24. 24. JanuarJanuar 2007, Karlsruhe 2007, Karlsruhe
ARMAR-III: Head
Neck• 4 DOF (Pitch-Roll-Yaw-Pitch)• Incremental and absolute
position sensors for all DOFs
Head: • 4 Dragonfly Color cameras,
640x480 (30 Hz), Firewire,2,9 mm und 12 mm micro lenses
• Framerate: 30fps max.• Resolution: 640x480 max.• common tilt and independent pan• Variable stereo calibration• 6 microphones
Tam
im A
sfou
rU
nive
rsitä
tKa
rlsru
he
F 4.13 “F 4.13 “SteuerungSteuerung und und RegelungRegelung von von RoboternRobotern””24. 24. JanuarJanuar 2007, Karlsruhe 2007, Karlsruhe
ARMAR III: Upper body
7 DOF Arms• Weight: 6 kg, max. Payload: 3 kg• Sensors: each joint
- motor encoders- Optical encoder in the axis- Joint torque sensor
(www.novatechuk.demon.co.uk)
• 6 DOF Force/Torque sensor (ATI) • Artificial skin
- Shoulder- Upper arm and forearm
3 DOF Torso• 2 Embedded PCs
Tam
im A
sfou
rU
nive
rsitä
tKa
rlsru
he
F 4.13 “F 4.13 “SteuerungSteuerung und und RegelungRegelung von von RoboternRobotern””24. 24. JanuarJanuar 2007, Karlsruhe 2007, Karlsruhe
ARMAR III: 5-Finger Hands
Anthropomorphic5-finger lightweight hand
• 8 DOF• weight ca. 200 g• Holding force 2 kg• On-board controller• Fluidic/pneumatic actuators• Anthropomorphic shape
and motion characteristics
Tam
im A
sfou
rU
nive
rsitä
tKa
rlsru
he
F 4.13 “F 4.13 “SteuerungSteuerung und und RegelungRegelung von von RoboternRobotern””24. 24. JanuarJanuar 2007, Karlsruhe 2007, Karlsruhe
ARMAR III: Platform
Application area: Kitchen
holonomic mobile platform• wheels with passive rolls at the
circumference (Mecanum wheels or Omniwheels)
• 3 laser scanner (Hokuyo Automatic)
Platform specification:• Height: 700 mm• Weight: 90 kg• Max. speed: 1.2 m/s• Power supply (whole system):
8 h with 25% drive• Spring-damper combination to reduce
vibrations
Tam
im A
sfou
rU
nive
rsitä
tKa
rlsru
he
F 4.13 “F 4.13 “SteuerungSteuerung und und RegelungRegelung von von RoboternRobotern””24. 24. JanuarJanuar 2007, Karlsruhe 2007, Karlsruhe
Computer architecture
2 GHz PC/104• vision
• tracker
• object recognition
2 GHz PC/104• vision
• tracker
• object recognition
933 MHz PC/104• position control
• torque control
• movement coordination
933 MHz PC/104• position control
• torque control
• movement coordination
933 MHz PC/104• platform control
• localization
• collision avoidance
933 MHz PC/104• platform control
• localization
• collision avoidance
1,6 GHz IPC• speech recognition
• speech synthesis
• acoustic localization
1,6 GHz IPC• speech recognition
• speech synthesis
• acoustic localization
UCoM #1• motor control
• encoder acquisition
• current sensing
UCoM #1• motor control
• encoder acquisition
• current sensing
1,6 GHz IPC• Global models
• Fileserver
1,6 GHz IPC• Global models
• Fileserver
UCoM #12• motor control
• encoder acquisition
• current sensing
UCoM #12• motor control
• encoder acquisition
• current sensing
Task planning level
Task coordination level
Task execution level
Gigabit Ethernet
4 CAN-Busses
Tam
im A
sfou
rU
nive
rsitä
tKa
rlsru
he
F 4.13 “F 4.13 “SteuerungSteuerung und und RegelungRegelung von von RoboternRobotern””24. 24. JanuarJanuar 2007, Karlsruhe 2007, Karlsruhe
Universal Controller Module (UCoM)
Motor board• Size: 80mm * 70mm * 20mm
• Up to 3 Motors (24V up to 5A)
• 6 Encoder-Ports (6-pin, 2 power supply, 4 IOs)
• Serial Peripheral Interface (SPI)
• 3 General Purpose IOs
DSP/FPGA:• 80MHz DSP: Motorola Hybrid Controller DSP56F803
- Size: 57mm * 70mm * 15mm- Interfaces: CAN, SCI, SPI, JTAG- Analog Ports: 2 * 4 channels ADC - Motorcontrol: 6 PWM
• External RAM • FPGA: Altera EPF10k30• FPGA-Configuration: Altera EPC2-ROM• All Components programmable via JTAG• DSP programmable via CAN-Bus
Tam
im A
sfou
rU
nive
rsitä
tKa
rlsru
he
F 4.13 “F 4.13 “SteuerungSteuerung und und RegelungRegelung von von RoboternRobotern””24. 24. JanuarJanuar 2007, Karlsruhe 2007, Karlsruhe
Software Framework
Modular Controller Architecture http://www.mca2.org
• modular, network transparent and real-time capable C/C++ framework for controlling robots
• Supports distributed computing • The main platform is Linux/RTLinux;
support for Win32 and MCA OS/X also exists.
• Easy to parameterize; to administrate• Modules with defined interfaces• Combination of modules into groups, which
act like any other module but with more complex behavior.
• Reusability of modules
Linux
RT
Linux
RT
Tcp-socket
Linux
RT
Linux
RT
RTAI-Linux
Linux/Windows
RT-FIFO
Planning Level
Coordination Level
Execution Level
Firewire
CAN
Tam
im A
sfou
rU
nive
rsitä
tKa
rlsru
he
F 4.13 “F 4.13 “SteuerungSteuerung und und RegelungRegelung von von RoboternRobotern””24. 24. JanuarJanuar 2007, Karlsruhe 2007, Karlsruhe
Implementation in ARMAR-III
Tam
im A
sfou
rU
nive
rsitä
tKa
rlsru
he
F 4.13 “F 4.13 “SteuerungSteuerung und und RegelungRegelung von von RoboternRobotern””24. 24. JanuarJanuar 2007, Karlsruhe 2007, Karlsruhe
Object Recognition
Recognition of colored objects Segmentation by color [IROS 2006]
• Appearance-based recognition using a global approach
• Model-based generation of view sets• Combination of stereo vision and stored
orientation information for 6D-localization (with respect to a given 3D model)
Recognition of textured objects Recognition using local features [Humanoids 2006]
• Calculation of consistent featureswith respect to the pose of the object by using the Hough transform
• 2D-localization using image pointcorrespondences
• 3D-localization using stereo vision
The robot‘s view(left camera)
Recognition/Localizationresult
Correspondences between learned view and view in scene
Tam
im A
sfou
rU
nive
rsitä
tKa
rlsru
he
F 4.13 “F 4.13 “SteuerungSteuerung und und RegelungRegelung von von RoboternRobotern””24. 24. JanuarJanuar 2007, Karlsruhe 2007, Karlsruhe
Integrating Vision Toolkit (IVT)
Developed by Pedram Azad (IAIM, Karlsruhe)
Available under: http://ivt.sourceforge.net
Features:• Integrating various cameras and camera interfaces (Dragonfly,
Digiclops/Bumblee, Videre, CMU1394, Unicap, Quicktime, uncompressed AVIs, bitmap sequences, ...)
• Using QT for visualization and creation of GUIs• Encapsulating OpenCV for functionality which has not been
implemented yet• Platform independent: being constantly used and tested under
Linux, Mac OS and Windows
IVT applications are portable automatically
Tam
im A
sfou
rU
nive
rsitä
tKa
rlsru
he
F 4.13 “F 4.13 “SteuerungSteuerung und und RegelungRegelung von von RoboternRobotern””24. 24. JanuarJanuar 2007, Karlsruhe 2007, Karlsruhe
Natural Interaction
Multimodal human-robot dialogs• For robot control, requesting
services, knowledge acquisition• Task oriented dialogs• Error handling• Learning dialog strategies with
reinforcement learning
Speech recognition for continuous speech (German/ English)
• Semantic context free grammars for recognition and understanding
• Tight coupling of speech recognition and dialog manager
• Close and distant speech
Audio and visual tracking, localization and sound classification
ARMAR-III, Kitchen
Tam
im A
sfou
rU
nive
rsitä
tKa
rlsru
he
F 4.13 “F 4.13 “SteuerungSteuerung und und RegelungRegelung von von RoboternRobotern””24. 24. JanuarJanuar 2007, Karlsruhe 2007, Karlsruhe
Inverse Kinematics for Reaching and PointingInverse Kinematics for Reaching and Pointing
SoechtingSoechting and Flanders:and Flanders: SSensorimotor transformation model that maps the Cartesian wrist position to a natural arm posture using four representation parameters
ShShoulderoulder
ElbowElbow
WristWrist
Tam
im A
sfou
rU
nive
rsitä
tKa
rlsru
he
F 4.13 “F 4.13 “SteuerungSteuerung und und RegelungRegelung von von RoboternRobotern””24. 24. JanuarJanuar 2007, Karlsruhe 2007, Karlsruhe
Inverse Kinematics for Reaching and PointingInverse Kinematics for Reaching and Pointing
Mapping of the posture representation parameters into the joint angles of the arm, which are responsible for the position of the wrist (θθ11, , θθ22, , θθ33, , θθ44))
Final adjustment step: Use the remaining DOFs (θθ55, , θθ66, , θθ77) ) to match the exact position und orientation of the TCP
Tam
im A
sfou
rU
nive
rsitä
tKa
rlsru
he
F 4.13 “F 4.13 “SteuerungSteuerung und und RegelungRegelung von von RoboternRobotern””24. 24. JanuarJanuar 2007, Karlsruhe 2007, Karlsruhe
Collision Detection/Motion Planning
Armar-III Model• Detailed visualization (~20.000 triangles)• Simplified model for collision detection (~450
triangles)
Collisions with environment / Self collisions• Hierarchical model of scene and objects
(triangle meshes)• test on triangle level only for nearby parts of
the meshes (No unnecessary collision tests )• Real-time performance (query <1 ms)
Distance calculations ~ 3 msMotion planning with Rapidly Exploring Random Trees (RRT)
• Sampling-based approach
Tam
im A
sfou
rU
nive
rsitä
tKa
rlsru
he
F 4.13 “F 4.13 “SteuerungSteuerung und und RegelungRegelung von von RoboternRobotern””24. 24. JanuarJanuar 2007, Karlsruhe 2007, Karlsruhe
Manipulation Planning Among Movable Obstacles
ARMAR: 20 DOF
Objects:• 4 fixed• 18 movable
Goal: Ketchup
Joint work with Mike Stilman, Jan-Ullrich Schamburek, and James Kuffner [ICRA 2007]
Tam
im A
sfou
rU
nive
rsitä
tKa
rlsru
he
F 4.13 “F 4.13 “SteuerungSteuerung und und RegelungRegelung von von RoboternRobotern””24. 24. JanuarJanuar 2007, Karlsruhe 2007, Karlsruhe
ARMAR III: Retrieve Ketchup
Tam
im A
sfou
rU
nive
rsitä
tKa
rlsru
he
F 4.13 “F 4.13 “SteuerungSteuerung und und RegelungRegelung von von RoboternRobotern””24. 24. JanuarJanuar 2007, Karlsruhe 2007, Karlsruhe
Conclusions
ARMAR-III• Humanoid robot for household applications • distributed hardware architecture • Software framework• Integration of motor control, vision and speech• All components on-board (Power supply, computers, ... )
Contributions • Mechanics developed at IPEK
(Prof. Albers, www.ipek.uni-karlsruhe.de)• 5-Finger hand at developed FZK
(Prof. Bretthauer, www.fzk.de) • Artificial skin developed at IPR
(Prof. Wörn, wwwipr.ira.uka.de)
Tam
im A
sfou
rU
nive
rsitä
tKa
rlsru
he
F 4.13 “F 4.13 “SteuerungSteuerung und und RegelungRegelung von von RoboternRobotern””24. 24. JanuarJanuar 2007, Karlsruhe 2007, Karlsruhe
Thank you ...
... for your attention.