Upload
queenie-jane-perez
View
126
Download
0
Embed Size (px)
Citation preview
Human Motion Simulation Using
Back Propagation as
Inverse Kinematics Solver
Goals 1. To make a 3D representation of a
human body. 2. To implement an inverse
kinematics solver for motion simulation.
Human Motion Simulation Using Back Propagation as Inverse Kinematics Solver
Human Motion Simulation Using Back Propagation as Inverse Kinematics Solver
Modeling the Human Motions being simulated Inverse Kinematics Back Propagation
Human Motion Simulation Using Back Propagation as Inverse Kinematics Solver
Human Model : The Making
The human body The skeleton The human motion
Human Model : The Skeletal System
Human Motion Simulation Using Back Propagation as Inverse Kinematics Solver
Bones Joints Joint as Vertex Bone as Link (x1,y1,z1)
(x2,y2,z2)
Human Motion Simulation Using Back Propagation as Inverse Kinematics Solver
Human Model : DH Convention
Given these parameters d – link offset a – link lengthα – link twist (rotation on x-axis)θ – link angle (rotation on z-axis)
Describes the position and orientation of a link with respect to a frame of reference
Models the human motion
Human Model : Forward Kinematics
Human Motion Simulation Using Back Propagation as Inverse Kinematics Solver
Inverse Kinematics
Human Motion Simulation Using Back Propagation as Inverse Kinematics Solver
- compute the angles of the joints to reach a desired location.
Example of Inverse Kinematics
Back Propagation Is a artificial neural network that has 3 layers,
the input layer, hidden layer and output layer.
Human Motion Simulation Using Back Propagation as Inverse Kinematics Solver
Illustration of a back propagation neural network
Back Propagation Configuration
Number of input neurons – 24 Number of hidden layers – 2 Number of hidden neurons of each layer – 54 Number of output neurons – 54 Initial learning rate – 0.1 Final learning rate – 0.001 Initial momentum – 0.95 Final momentum – 0.2
Human Motion Simulation Using Back Propagation as Inverse Kinematics Solver
Back Propagation as Inverse Kinematics Solver
Four back propagations were used.1. Left leg 2. Right Leg3. Left arm 4. Right arm
Each were trained using randomly generated end effector’s positions and joint angles.
Human Motion Simulation Using Back Propagation as Inverse Kinematics Solver
Output Tolerance
82.24 84.3 81.4 81.5
16.24 15.7 18.6 18.5
0
10
20
30
40
50
60
70
80
90
Kinematics Chain
Per
cen
tag
e
Passed
Failed
Result
Human Motion Simulation Using Back Propagation as Inverse Kinematics Solver
Left Leg Right Leg Left Arm Right Arm
Significance
Human Motion Simulation Using Back Propagation as Inverse Kinematics Solver
•Can be used in different fields like computer animation and robotics.
•Our human model can be used as substitutes for the real humanoid robot participants into virtual environments.
Significance (Cont.)
Our human model can be used on experimenting right configurations of a human robot before building an actual robot.
Another neural network can be used as inverse kinematics solver to make a comparative study.
Human Motion Simulation Using Back Propagation as Inverse Kinematics Solver
References
[1] Robotics Appin Knowledge Solutions; Infinity Science Press LLC Hingham, Massachusetts New Delhi; p. 14
[2] The World Book Encyclopedia, 1995 [3] The World Book Encyclopedia, 1998 [4] Scientists try to make robots more human http//www.
usatoday.com/tech/news/robotics/2006-11-22
References [5] Interactive Simulation of Stylized Human Locomotion [6] Real time Virtual Humans http://www.cis.upenn.edu/~badler/bcs/Paper.htm [7] Jack, The Making of a Virtual Human Being http://www.dp.upenn.edu/street/032097/: [8] Robot a Programmer’s Bonanza; John Blankenship and Samuel
Mishal; McGraw-Hill Companies; 2008; pp.3-6 [9] Simulation of Human Body Kinematics http:// www.cescg.org/CESCG-2000/RFilkorn/ [10] Realistic Human Simulation From Simulating real-world objects, to
creating comprehensive simulations of reality
http://www.engineer.ucla.edu/newsroom/featurednews/archive/2009/realistic-human-simulation/
[11] Human Motion Modelling and Simulation http://academic.research.microsoft.com/Paper/6012143.aspx [12] Humanoid Robots http://www.robotmatrix.org/humanoidrobot.htm
References [13]Ontology forVirtual Humans http://vrlab.cpfl.ch/~alcgarcia/VHOntology/long.html [14]CS 490 Human Motion Simulation, Philip Yen, Spring 1998
http://www.nbb.cornell.edu/neurobio/land/OldStudentProjects/cs490-97to98/yen/ [15] SOK, K. W., KIM, M., AND LEE, J. 2007. Simulating biped behaviors from human motion data. ACM Transactions on Graphics 26, 3 (July), 107:1–107:9. [16] ABDALLAH, M.; GOSWAMI, A. 2005. A biomechanically motivated two-phase
strategy for biped upright balance control. In International Conference on Robotics and Automation (ICRA), 1996–2001.
[17] History of Robots http://pages.cpsc.ucalgary.ca/~jaeger/visualMedia/robotHistory.html http://www.space.gc.ca/pdf/educator-story_robot.pdf [18] http://www.intelligent-systems.com.ar/intsyst/artis.htm [19] Humanoid robots: A New Kind of Tool http://people.csail.mit.edu/brooks/papers/IEEE-cog.pdf [20] Digital Representations of Human Movement http://design.osu.edu/carlson/history/PDFs/badler-smoliar.pdf [21] Keyframe Animation http://www.cadtutor.net/dd/bryce/anim/anim.html
References [22] Humanoid Animation http://www.ia.hiof.no/vv/pages/moduler/animation/
humanoid_a.html [23 ] Specification for a Standard VRML Humanoid http://h-anim.org/Specifications/H-Anim1.0/ [24] Motion Capture http://www.ia.hiof.no/vv/pages/moduler/animation/
motion_capture.html [25] Surface Reconstruction from Point Cloud of Human
Body by Clustering; Systems and Computers in Japan, Vol. 37, No. 11, 2006 http://www.mm.media.kyoto-u.ac.jp/research/doc/727/
Publication.pdf
References [26]Kallmann, M. (2008), Analytical inverse kinematics with
body posture control. Computer Animation and Virtual Worlds, 19: 79–91. doi: 10.1002/cav.176
[27] Modelling Inverse Kinematics in a robotic armhttp://www.mathworks.com/products/fuzzylogic/demos.html?file=/products/demos/shipping/fuzzy/invkine_codepad.html
[28] Inverse Kinematics Solution of 3DOF Planar Robot using ANFIS; Srinivasan Alavandar, M.J. Nigam; Int. J. of Computers, Communications & Control, ISSN 1841-9836, E-ISSN 1841-9844 ;Vol. III (2008), Suppl. issue: Proceedings of ICCCC 2008, pp. 150-155
[29] Forward Kinematics for Virtual Agents; Antonio Benitez R., Guillermo de los Santos T. and Daniel Vallejo R.
[30] Simulation of Betty: A Humanoid Robot with OpenGL; Meng Cheng, Lau.
[31] Robot Modeling and Kinematics. Manseur, Rachid