Index of /Walkin8r/Favorites

      Name                    Last modified      Size  Description
Parent Directory - 225.tar.bz2 2014-02-25 10:34 115K Aopt0002.png 2013-05-18 15:58 53K Aopt0022.png 2013-05-17 07:21 49K a040.svg 2014-02-25 07:59 86K b033.gif 2014-02-27 14:43 3.4M b033.svg 2014-02-25 07:59 86K b049.gif 2014-02-27 14:43 3.4M b049.svg 2014-02-25 07:59 86K b064.svg 2014-02-25 07:59 86K b087.svg 2014-02-25 07:59 86K b103.gif 2014-02-27 14:43 3.5M b103.svg 2014-02-25 07:59 86K b177.svg 2014-02-25 07:59 86K b198.svg 2014-02-25 07:59 86K c039.svg 2014-02-25 07:59 86K c046.gif 2014-02-27 14:43 3.5M c046.svg 2014-02-25 07:59 86K c060.gif 2014-02-27 14:43 3.5M c060.svg 2014-02-25 07:59 86K c090.svg 2014-02-25 07:59 86K c116.svg 2014-02-25 07:59 86K c293.svg 2014-02-25 07:59 86K d033.svg 2014-02-25 07:59 86K kopt0038.png 2013-05-18 07:48 33K o14_0114.gif 2013-08-15 10:51 4.1M o14_0114.png 2013-08-16 06:07 55K o14_114.svg 2013-09-05 15:14 350K o17_237.gif 2013-09-09 06:08 4.1M o17_237.png 2013-09-09 06:07 55K
Selected Optimized Linkage Configurations
$URL: svn+ssh:// $
$Id: README.html 410 2014-02-26 15:05:51Z mrwhat $

Selected Optimized Linkage Configurations

Optimizer was re-written and simplified circa 2/2014. The main metrics on this optimizer are the bumpiness of the step path (std. dev. of y in active phase), the change of speed during the active part of the step (std. dev. of foot speed), length of step, median height of return phase of step, symetry of return phase of step and reach (gradient) of inside part of step. There were also some boundary conditions applied, keeping the knee from dropping too low on return phase, how far the CH link stays from the B axle, and how far the linkage state is from a known impossible state. I am still trying to decide between these for the next version of the walker:
On this version of the optimizer, I start at a known, good configuration with small optimization steps. I watch the slow evolution of the configuration as it is optimized, and stop at places where it looks good to my human judgement.
They tend to get over-optimized and converge to degenerate, but high-scoring solutions.
I alter the weights of various components of the metric to try and influence the trajectory of this optimization to pass through more desirable sub-optimal solutions shown here.
An optimizer, with some simulated annealing, was run on the Jansen linkage. The metric is still being played with, but the statistics reported at the top of each linkage plot are fairly consistant.
Duty cycle when foot is on (or near) ground, in percent.
St.Dev. of each foot delta-x when walking, multiplied by number of steps (to keep it invariant w.r.t. no steps analysed) (less is better)
St.Dev. of foot height durring walking cycle. (less is better)
Minimum cm/m foot position from central axis.
Usually the peak of the foot height above ground. Later experiments are trying to use the median foot height above ground on the return cycle.

Later versions used a similarity of the footpath to a tank tread as a quality indicator (trackSD) instead of the flatness and footspeed (flatSD, speedSD). This proved to be a more reliable metric for generating reasonable optimized linkages.

The numbers across the bottom are the linkage lengths. See top page Jansen linkage drawing for definition of hinges/linkage names. The linkages shown are normalized for a 17.5 cm crank length (AC). The remaining measurements, shown at the bottom of the plot are


Klann Linkage

The numbers across the bottom of optimized Klann linkage plots are
     Bx By Dx Dy BH CF CH DE EF EG FG FH
respectively, for a linkage normalized to a crank length (AC) of 17.5.
Last modified: Tue May 14 10:36:35 MDT 2013