g

Artificial and Natural Walking Machines, and Neural Networks

Seminar SS2002 Dr. Rolf Pfeifer (pfeifer@ifi.unizh.ch)
Chandana Paul (chandana@ifi.unizh.ch)

Now available: Student Papers and Presentations

In this class, we will explore the exciting world of Natural and Artificial Walking Machines, and Neural Networks. This class will introduce the latest ideas and insights into biological walking systems, such as humans and animals, and discuss the underlying neural structures which control them. It will also focus on the latest advances in the control of walking robots. For students who are interested in topics introduced in the class "New Artificial Intelligence", this seminar will be an opportunity to further develop insights into AI, with respect to biological and robotic motor behavior.

List of Topics

Introduction to Artificial and Natural Walking Machines
Speaker: Chandana Paul 5.4.02

A general introduction to the Seminar and overview of future topics.

Neural Oscillators
Speaker: Sonja Kramer , Philip Schoch 12.4.02

Many biologists, such as Grillner, hypothesize that neural structures called "oscillators" are responsible for producing rhythmic movements, during activities such as swimming, walking and running, in invertebrates to higher vertebrates. Taga has been able to use simple oscillator networks to control a 3D walking robot in simulation, have it adapt to unpredictable environments and anticipate obstacles. Might it be that oscillators are being used in humans as well?

References

Grillner, S. Neural networks for vertebrate locomotion. Scientific American, 48-53, Jan 1996.
Taga, G. (1995). A model of the neuro-musculo-skeletal system for human locomotion. Biol. Cybern. 73:113--121, 1995.
Taga, G. A model of the neuro-musculo-skeletal system for anticipatory adjustment of human locomotion during obstacle avoidance. Biological Cybernetics 78, 2 (1998), 9-17.
Computer Graphics. pp. 15--20. Taga, G., Y. Yamaguchi and H. Shimizu (1991). Selforganized control of bipedal locomotion by neural oscillators in unpredictable environment. Biological Cybernetics 65(3), 147--159.

Quadrupeds
Guest Speaker: Fumiya Iida 19.4.02
While having four legs makes balance easier, it means there are more legs to control. How does a complex creature like a four legged animal coordinate all its legs? And why do gaits change as the speed of motion increases? These interesting questions can be partially answered by studying the gaits of salamanders, horses and artificial robot dogs.

References

M. Hildebrand. Symmetrical gaits of horses. Science, 150:701--708, 1965.
Ijspeert, A.J. 2001. A connectionist central pattern generator for the aquatic and terrestrial gaits of a simulated salamander. Biological Cybernetics 84, 331-348. [PDF]
Kimura, M., H., Tsuchiya, K., Onat, A. Towards 3D Adaptive Dynamic Walking of a Quadruped Robot on Irregular Terrain by Using Neural System Model Proceedings of the IEEE Intl. Conf. on Intelligent, Robots and Systems, Oct, 2001.

"Cruse" Control
Speaker: Jonas Boesch , Patrice Egger 26.4.02
Holk Cruse, a biologist and roboticist, found that insects, which have many legs, do not use a central processor to keep track of all the movements of their multiple legs. Instead local interactions between sensory motor reflexes enable complex walking behaviors to emerge. In this session, you can find out how this surprising behavior occurs, and how these ideas have been used in other robots.

Cruse, H., Dean, J., and Ritter, H. Die Entdeckung der Intelligenz oder Konnen Ameisen denken? Intelligenz bei Tieren und Maschinen. C. H. Beck 1998
H. Cruse, U. Mueller-Wilm, and J. Dean. Artificial neural nets for controlling a 6-legged walking system. In [SAB92].
C. Ferrell, "Many Sensors, One Robot", IEEE/RSJ Int. Conf. on Intelligent Robots and Systems, Yokohama, Japan, pp. 399-406, July 1993.

Evolutionary Methods for Locomotion
Speaker: Martina Huber 3.5.02
Evolutionary methods have been successfully used to design controllers for all kinds of walking machines in simulation, from salamanders, to octapods to artificial insects and bipeds. This session will focus on understanding how these automated design algorithms can find solutions even better than human engineers!

References

Paul, C., and Bongard, J. The Road Less Travelled: Morphology in the Optimization of Biped Robot Locomotion, to appear in Proceedings of The IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS2001), Hawaii, USA.
Jakobi, N. Running Across the Reality Gap: Octopod Locomotion Evolved in Minimal Simulation In P. Husbands and J-A Meyer, editors, Proceedings of Evorobot. Springer Verlag, 1998,
Kodjabachian and J.-A. Meyer, Evolution and development of neural networks controlling locomotion, gradient-following, and obstacle avoidance in artificial insects 1997.

Balance Control
Speaker: Sebastian Kay Belle , Tobias Kleyer 10.5.02
Control of balance is one of the most sophisticated tasks performed by biological systems. Not only is it essential for walking and running in most animals and humans, but it also enable humans to perform fine motor activities, such as dancing, ice-skating, skiing and even snowboarding! In this session, you can find out how balance control for such activities could be accomplished by neural systems in robots.

G. Taga, 1997, A Model of Integration of Posture and Locomotion, Proceedings of International Symposium on Computer Simulation in Biomechanics.
A. Kun and W. T. Miller, III, Adaptive Dynamic Balance of a Biped Robot using Neural Networks . Proc. IEEE Int. Conf. on RA, pp. 240--245, 1996
Kun A. L., Miller W. T. (1997), Adaptive Static Balance of a Biped Robot Using Neural Networks, Proc. of the Int.

Passive Dynamic Walking
Speaker: Stefan Mueller , Lukas Fray 17.5.02
In 1990, Tad McGeer showed for the first time that a mechanical structure, without sensors, motors or control, could walk on its own down a slope! This was a shock to the biped robot community who had for years built elaborate robots with many sensors and motors and complex control. Since then, several simple passive dynamic walkers have been built. In this session, one can discover the secrets behind the design of such autonomous walking robots. Students with mechanical inclinations can also try to build simple passive dynamic walking machines and report on their own work.

References

T. McGeer. Passive dynamic walking. Int. J. Robotics Research, 9(2), 1990.
M. Coleman and A. Ruina. A Tinkertoy R model that walks.Physical Review Letters, 1997. accepted for publication.
Collins, S. H., Wisse, M., and Ruina, A. (2001). "A 3-D passive-dynamic walking robot with two legs and knees." International Journal of Robotics Research, In Press.

Rapid Locomotion
Sandro Bocuzzo, Daniel Steiner 24.5.02
Extraordinary athletic feats such as running, sprinting, and long jumps require the body to save and reuse as much of its energy as possible. The biological mechanisms which enable us to perform these activities, are also the same ones which enable cheetahs to run faster than 100 km/h. This session will focus on how biological systems exploit the material properties of their bodies to accomplish these activities.

References

Alexander, R.M. (1984a). Elastic energy stores in running vertebrates. Amer. Zool., 24, 85-94.
Alexander, R. (1984) Walking and running. American Scientist. 72 348-354.
Alexander, R. M. (1991). Energy-saving mechanisms in walking and running. J. Exp. Biol, 160:55--69.

Artificial Muscles
Speaker: Boris Neubert , Thomas Franken 31.5.02
It is difficult to match the effectiveness of muscles; their strength to weight ratio, their reaction speeds, and their ability to relax and let passive dynamics take over is mostly unparalled in mechanical actuators to date. But many attempts have been made to partially replicate some of their properties. In this session, find out the interesting ways in which research has made progress in this direction.

References

G. K. Klute, J. M. Czerniecki, and B. Hannaford, "McKibben Artificial Muscles: Pneumatic Actuators with Biomechanical Intelligence", Proc. IEEE/ASME International Conference on Advanced Intelligent Mechatronics.
Robinson, D. W., Pratt, J. E., Paluska, D. J. & Pratt, G. A. (1999), `Series Elastic Actuator Development for a Biomimetic Robot', IEEE/ASME International Conference on Advanced Intelligent Mechatronics.
David L. Brock, "Review of Artificial Muscle Based on Contractile Polymers," MIT Artificial Intelligence Laboratory Memo 1330, November, 1991.

Neural Modelling of Human Walking
Speaker: Christoph Dahl 7.6.02
At the AILab, University of Zurich, research has been focused on understanding the neural mechanisms in the spinal cord which give rise to locomotion. In the undertaking, an elaborate artificial neural network spinal cord model has been designed, with current knowledge from biology, and it has been used to control a simulated biped robot. Find out more about how this biorobotics project is helping to understand human locomotion and help paraplegic patients at the University Hospital.

References

Kueng, T. Neural Oscillator Networks and Coupling. Diploma Thesis, AILab, UNIZH
Durrer, A. Energy Efficieny in Neural Oscillator Networks Diploma Thesis, AILab, UNIZH
Baumgartner, D. Learning and Neural Plasticity Diploma Thesis. AILab, UNIZH

Humanoid Robots
Speaker: Daniel Gerteis , Michael Sutter 14.6.02
The newest generation of Humanoid robots, developped by Honda and Sony, almost look like an Android species landed on earth from outer space. Yet behind their shiny exteriors they are highly technically advanced. In this session, take a look "inside" these humanoid robots and see how they are constructed and controlled.

References

K. Hirai, M. Hirose, Y. Haikawa, and T. Takenaka. The development of honda humanoid robot. IEEE Conference on Robotics and Automation, 1998.
J . Yamaguchi and A. Takanishi. Development of a Humanoid Robot - Design of a Biped Walking Robot Having Antagonistic Driven Joints Using Nonlinear Spring Mechanism. Proc. 1st Int. Symposium on HUmanoid RObots, pages102-111, 1996. 3.
S. Schaal. Is Imitation Learning the Route to Humanoid Robots. Trends in Cognitive Sciences
G. Cheng, A. Nagakubo and Y. Kuniyoshi. Continuous Humanoid Interaction: An Integrated Perspective - Gaining Adaptivity, Redundancy, Flexibility - In One. Robotics and Autonomous Systems, Volume 37, No. 2-3, pp. 161-183, November 2001.

Learning to Walk
Speaker: Richard Hefti 21.6.02
Understanding how babies learn locomotion, by first crawling and then walking, gives us insight into the development of the complex locomotor apparatus which eventually enables us to walk. Join researchers who study infant motor development, in the quest to understand the human locomotor system.

References

Thelen, E. (1986). Treadmill-elicited stepping in seven-month-old infants. Child Development, 57, 1498-1506.
Burnside, L. H. (1927). Coordination in the Locomotion of Infants, Genetic Psychology Monographs, 2, 283-341.
Bertenthal, B. I. & Bai, D. L., (1989). Infants' sensitivity to optical flow for controlling posture. Developmental Psychology, 25, 936- 945.
Benson, J?B. (1990). The significance and development of crawling in human infancy. In J. E. Clark, J. H. Humphrey, (Eds.) Advances in motor development research, Vol. 3, 91?142. New York: AMS Press, Inc.

Exotic Locomotion
Speaker: Ivo Schindler, Daniel Oberhoff 28.6.02
In addition to the biological forms of bipeds, quadrupeds, hexapods, octapods, centipedes and millipeds, there is the strange and bizarre world of artificial walking machines which arise from the creative minds of researchers and have no biological likeness at all. Among these, are the inverted double pendulum walker, Stumpy, the Ball robot, and the Rectiblob. In this session learn these and other strange artificial creatures. Students are also encouraged to try to design strange and interesting locomoting robots of their own!

References

M. Yim, A reconfigurable modular robot with multiple modes of locomotion, in Proceedings of the 1993 JSME Conference on Advanced Mechatronics, Tokyo, 1993.
Homsy, G.E., Allen, A.R., Pratt, G.A. The Recti-Blob: A conformal shape-changing robot for robust locomotion over rough terrain Unpublished as of 10/2/98

Final Session
Speakers: Dr. Rolf Pfeifer, Chandana Paul 5.7.02

Review, Conclusions

Additional References

This section was going to contain a list of many other interesting publications related to this topic. But then I realized... the list is endless! So, if you're interested in finding more publications on a particular topic, the NEC Research Index is a good database.

Interesting Links
(under construction...)

Introduction to Artificial and Natural Walking Machines Speaker: Chandana Paul	5.4.02
A general introduction to the Seminar and overview of future topics.
Neural Oscillators Speaker: Sonja Kramer , Philip Schoch	12.4.02
Many biologists, such as Grillner, hypothesize that neural structures called "oscillators" are responsible for producing rhythmic movements, during activities such as swimming, walking and running, in invertebrates to higher vertebrates. Taga has been able to use simple oscillator networks to control a 3D walking robot in simulation, have it adapt to unpredictable environments and anticipate obstacles. Might it be that oscillators are being used in humans as well? References Grillner, S. Neural networks for vertebrate locomotion. Scientific American, 48-53, Jan 1996. Taga, G. (1995). A model of the neuro-musculo-skeletal system for human locomotion. Biol. Cybern. 73:113--121, 1995. Taga, G. A model of the neuro-musculo-skeletal system for anticipatory adjustment of human locomotion during obstacle avoidance. Biological Cybernetics 78, 2 (1998), 9-17. Computer Graphics. pp. 15--20. Taga, G., Y. Yamaguchi and H. Shimizu (1991). Selforganized control of bipedal locomotion by neural oscillators in unpredictable environment. Biological Cybernetics 65(3), 147--159.
Quadrupeds Guest Speaker: Fumiya Iida	19.4.02
While having four legs makes balance easier, it means there are more legs to control. How does a complex creature like a four legged animal coordinate all its legs? And why do gaits change as the speed of motion increases? These interesting questions can be partially answered by studying the gaits of salamanders, horses and artificial robot dogs. References M. Hildebrand. Symmetrical gaits of horses. Science, 150:701--708, 1965. Ijspeert, A.J. 2001. A connectionist central pattern generator for the aquatic and terrestrial gaits of a simulated salamander. Biological Cybernetics 84, 331-348. [PDF] Kimura, M., H., Tsuchiya, K., Onat, A. Towards 3D Adaptive Dynamic Walking of a Quadruped Robot on Irregular Terrain by Using Neural System Model Proceedings of the IEEE Intl. Conf. on Intelligent, Robots and Systems, Oct, 2001.
"Cruse" Control Speaker: Jonas Boesch , Patrice Egger	26.4.02
Holk Cruse, a biologist and roboticist, found that insects, which have many legs, do not use a central processor to keep track of all the movements of their multiple legs. Instead local interactions between sensory motor reflexes enable complex walking behaviors to emerge. In this session, you can find out how this surprising behavior occurs, and how these ideas have been used in other robots. Cruse, H., Dean, J., and Ritter, H. Die Entdeckung der Intelligenz oder Konnen Ameisen denken? Intelligenz bei Tieren und Maschinen. C. H. Beck 1998 H. Cruse, U. Mueller-Wilm, and J. Dean. Artificial neural nets for controlling a 6-legged walking system. In [SAB92]. C. Ferrell, "Many Sensors, One Robot", IEEE/RSJ Int. Conf. on Intelligent Robots and Systems, Yokohama, Japan, pp. 399-406, July 1993.
Evolutionary Methods for Locomotion Speaker: Martina Huber	3.5.02
Evolutionary methods have been successfully used to design controllers for all kinds of walking machines in simulation, from salamanders, to octapods to artificial insects and bipeds. This session will focus on understanding how these automated design algorithms can find solutions even better than human engineers! References Paul, C., and Bongard, J. The Road Less Travelled: Morphology in the Optimization of Biped Robot Locomotion, to appear in Proceedings of The IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS2001), Hawaii, USA. Jakobi, N. Running Across the Reality Gap: Octopod Locomotion Evolved in Minimal Simulation In P. Husbands and J-A Meyer, editors, Proceedings of Evorobot. Springer Verlag, 1998, Kodjabachian and J.-A. Meyer, Evolution and development of neural networks controlling locomotion, gradient-following, and obstacle avoidance in artificial insects 1997.

Balance Control Speaker: Sebastian Kay Belle , Tobias Kleyer	10.5.02
Control of balance is one of the most sophisticated tasks performed by biological systems. Not only is it essential for walking and running in most animals and humans, but it also enable humans to perform fine motor activities, such as dancing, ice-skating, skiing and even snowboarding! In this session, you can find out how balance control for such activities could be accomplished by neural systems in robots. G. Taga, 1997, A Model of Integration of Posture and Locomotion, Proceedings of International Symposium on Computer Simulation in Biomechanics. A. Kun and W. T. Miller, III, Adaptive Dynamic Balance of a Biped Robot using Neural Networks . Proc. IEEE Int. Conf. on RA, pp. 240--245, 1996 Kun A. L., Miller W. T. (1997), Adaptive Static Balance of a Biped Robot Using Neural Networks, Proc. of the Int.
Passive Dynamic Walking Speaker: Stefan Mueller , Lukas Fray	17.5.02
In 1990, Tad McGeer showed for the first time that a mechanical structure, without sensors, motors or control, could walk on its own down a slope! This was a shock to the biped robot community who had for years built elaborate robots with many sensors and motors and complex control. Since then, several simple passive dynamic walkers have been built. In this session, one can discover the secrets behind the design of such autonomous walking robots. Students with mechanical inclinations can also try to build simple passive dynamic walking machines and report on their own work. References T. McGeer. Passive dynamic walking. Int. J. Robotics Research, 9(2), 1990. M. Coleman and A. Ruina. A Tinkertoy R model that walks.Physical Review Letters, 1997. accepted for publication. Collins, S. H., Wisse, M., and Ruina, A. (2001). "A 3-D passive-dynamic walking robot with two legs and knees." International Journal of Robotics Research, In Press.
Rapid Locomotion Sandro Bocuzzo, Daniel Steiner	24.5.02
Extraordinary athletic feats such as running, sprinting, and long jumps require the body to save and reuse as much of its energy as possible. The biological mechanisms which enable us to perform these activities, are also the same ones which enable cheetahs to run faster than 100 km/h. This session will focus on how biological systems exploit the material properties of their bodies to accomplish these activities. References Alexander, R.M. (1984a). Elastic energy stores in running vertebrates. Amer. Zool., 24, 85-94. Alexander, R. (1984) Walking and running. American Scientist. 72 348-354. Alexander, R. M. (1991). Energy-saving mechanisms in walking and running. J. Exp. Biol, 160:55--69.
Artificial Muscles Speaker: Boris Neubert , Thomas Franken	31.5.02
It is difficult to match the effectiveness of muscles; their strength to weight ratio, their reaction speeds, and their ability to relax and let passive dynamics take over is mostly unparalled in mechanical actuators to date. But many attempts have been made to partially replicate some of their properties. In this session, find out the interesting ways in which research has made progress in this direction. References G. K. Klute, J. M. Czerniecki, and B. Hannaford, "McKibben Artificial Muscles: Pneumatic Actuators with Biomechanical Intelligence", Proc. IEEE/ASME International Conference on Advanced Intelligent Mechatronics. Robinson, D. W., Pratt, J. E., Paluska, D. J. & Pratt, G. A. (1999), `Series Elastic Actuator Development for a Biomimetic Robot', IEEE/ASME International Conference on Advanced Intelligent Mechatronics. David L. Brock, "Review of Artificial Muscle Based on Contractile Polymers," MIT Artificial Intelligence Laboratory Memo 1330, November, 1991.
Neural Modelling of Human Walking Speaker: Christoph Dahl	7.6.02
At the AILab, University of Zurich, research has been focused on understanding the neural mechanisms in the spinal cord which give rise to locomotion. In the undertaking, an elaborate artificial neural network spinal cord model has been designed, with current knowledge from biology, and it has been used to control a simulated biped robot. Find out more about how this biorobotics project is helping to understand human locomotion and help paraplegic patients at the University Hospital. References Kueng, T. Neural Oscillator Networks and Coupling. Diploma Thesis, AILab, UNIZH Durrer, A. Energy Efficieny in Neural Oscillator Networks Diploma Thesis, AILab, UNIZH Baumgartner, D. Learning and Neural Plasticity Diploma Thesis. AILab, UNIZH
Humanoid Robots Speaker: Daniel Gerteis , Michael Sutter	14.6.02
The newest generation of Humanoid robots, developped by Honda and Sony, almost look like an Android species landed on earth from outer space. Yet behind their shiny exteriors they are highly technically advanced. In this session, take a look "inside" these humanoid robots and see how they are constructed and controlled. References K. Hirai, M. Hirose, Y. Haikawa, and T. Takenaka. The development of honda humanoid robot. IEEE Conference on Robotics and Automation, 1998. J . Yamaguchi and A. Takanishi. Development of a Humanoid Robot - Design of a Biped Walking Robot Having Antagonistic Driven Joints Using Nonlinear Spring Mechanism. Proc. 1st Int. Symposium on HUmanoid RObots, pages102-111, 1996. 3. S. Schaal. Is Imitation Learning the Route to Humanoid Robots. Trends in Cognitive Sciences G. Cheng, A. Nagakubo and Y. Kuniyoshi. Continuous Humanoid Interaction: An Integrated Perspective - Gaining Adaptivity, Redundancy, Flexibility - In One. Robotics and Autonomous Systems, Volume 37, No. 2-3, pp. 161-183, November 2001.
Learning to Walk Speaker: Richard Hefti	21.6.02
Understanding how babies learn locomotion, by first crawling and then walking, gives us insight into the development of the complex locomotor apparatus which eventually enables us to walk. Join researchers who study infant motor development, in the quest to understand the human locomotor system. References Thelen, E. (1986). Treadmill-elicited stepping in seven-month-old infants. Child Development, 57, 1498-1506. Burnside, L. H. (1927). Coordination in the Locomotion of Infants, Genetic Psychology Monographs, 2, 283-341. Bertenthal, B. I. & Bai, D. L., (1989). Infants' sensitivity to optical flow for controlling posture. Developmental Psychology, 25, 936- 945. Benson, J?B. (1990). The significance and development of crawling in human infancy. In J. E. Clark, J. H. Humphrey, (Eds.) Advances in motor development research, Vol. 3, 91?142. New York: AMS Press, Inc.
Exotic Locomotion Speaker: Ivo Schindler, Daniel Oberhoff	28.6.02
In addition to the biological forms of bipeds, quadrupeds, hexapods, octapods, centipedes and millipeds, there is the strange and bizarre world of artificial walking machines which arise from the creative minds of researchers and have no biological likeness at all. Among these, are the inverted double pendulum walker, Stumpy, the Ball robot, and the Rectiblob. In this session learn these and other strange artificial creatures. Students are also encouraged to try to design strange and interesting locomoting robots of their own! References M. Yim, A reconfigurable modular robot with multiple modes of locomotion, in Proceedings of the 1993 JSME Conference on Advanced Mechatronics, Tokyo, 1993. Homsy, G.E., Allen, A.R., Pratt, G.A. The Recti-Blob: A conformal shape-changing robot for robust locomotion over rough terrain Unpublished as of 10/2/98
Final Session Speakers: Dr. Rolf Pfeifer, Chandana Paul	5.7.02
Review, Conclusions