Solving the mechanical singularity of the stretched Knee?

[NovaOne]

A friend of mine, subscribes to http://www.biomch-l.org/ and showed me this thread yesterday, ( I was going to just post the takeanishi URL, but is it better to quote Massimiliano fully?)

Dear All,

There is a robot capable of solving the mechanical
singularity of the stretched knee:
http://www.takanishi.mech.waseda.ac.jp/research/wabian/index.htm

Suggested readings:
Human-like Walking with Knee Stretched, Heel-contact and
Toe-off Motion by a Humanoid Robot
Yu Ogura; Kazushi Shimomura; Hideki Kondo; Akitoshi
Morishima; Tatsu Okubo; Shimpei Momoki; Hun-ok Lim; Atsuo
Takanishi
Intelligent Robots and Systems, 2006 IEEE/RSJ
International Conference on
Volume , Issue , Oct. 2006 Page(s):3976 - 3981

Development of a new humanoid robot WABIAN-2
Yu Ogura; Aikawa, H.; Shimomura, K.; Kondo, H.; Morishima,
A.; Hun-ok Lim; Takanishi, A.
Robotics and Automation, 2006. ICRA 2006. Proceedings 2006
IEEE International Conference on
Volume , Issue , May 15-19, 2006 Page(s): 76 - 81

and related references.

Kind regards
Massimiliano


Takanishi Lab - http://www.takanishi.mech.waseda.ac.jp/
RoboCasa - http://www.robocasa.net

---------------------------------------------------------------
Information about BIOMCH-L: http://www.Biomch-L.org
Archives: http://listserv.surfnet.nl/archives/Biomch-L.html
---------------------------------------------------------------

I have been looking for papers relating to bipeds to help me understand the basics of things like: biped balance, ZMP, COM and inverse kinematics, and came across this excellent masters thesis. (I have found no reference to Felix or Jupp on Robosavvy but he may be known to some of you?) .
http://www.informatik.uni-freiburg.de/~faber/Learning_Fast_Bipedal_Locomotion.pdf

Heres an quotes from Felix Faber's work:

Furthermore, ZMP-controlled gaits result in ’bent-Knee’ motions in order to
avoid inverse kinematic singularities. These motions do not look natural and
place significant limits on the types of walking styles that can be produced.
In contrast to ZMP-based approaches my approach does not require an accu-
rate model of the robot and the environment. The approach presented here
does also not rely on highly precise controllers for following precomputed
joint trajectories, but works with low-cost servo-motors. As no dynamic
equations of motion have to be solved, this approach has a low computational
complexity. Using the gait presented in this thesis, a more human-like mo-
tion is possible using straight legs for walking, compared to the ’bent-knee’
motions of ZMP-controlled gaits.

Interesting stuff

Its a bit beyond me at the moment,(understatement!) but I hope this post is useful for some of you involved in tackling some of the more challenging aspects of biped development?

[plingboot]

really interesting info - i'm not going to pretend much make sense, but on a simplistic level it's interesting to study the mpg files frame-by-frame to see how the joints move - with a view to translating similar movement over to the hobby 'bot.

[JonHylands]

In order to walk the way that robot walks, your robot will need a pelvis. That allows the hips to pivot in a couple directions, which is essential.

- Jon