sjoerdbruijn / footplacement

Foot placement prediction, based on the work of Wang & Srinivasan. Written by @moiravl and @Sjoerdbruijn

FootPlacement

Foot placement prediction, based on the work of Wang & Srinivasavan (2014), and as used in Mahaki et al (2019), Van Leeuwen et al (2020). Feel free to use this code, but please cite one of the above work, or the DOI itself;

'foot_placement_model_function_step' is a function to perform a linear regression model. It correlates the center of mass kinematic state during swing with subsequent foot placement.

We uploaded example data 'testdata.mat' which can be used to run the example code below. Here we run the foot placement model and plot the resulting relative explained variance (% explained variance).

Using the foot placement model function requires the Matlab Statistics toolbox.

clear all;
load 'testdata'

pred_samples = 1:51;
order        = 2;
removeorigin = 1;
centerdata   = 1;

[OUT,intermediates]=foot_placement_model_function_step(CoM_ML,Rfoot,Lfoot,events,fs_opto,pred_samples,order,removeorigin,centerdata)

figure;
plot((1:51)*2-2,OUT.Combined_pct.data*100)
ylabel(OUT.Combined_pct.ylabel)
title(OUT.Combined_pct.titel)
xlabel('step percentage (%)')

References

Wang, Yang, and Manoj Srinivasan. "Stepping in the direction of the fall: the next foot placement can be predicted from current upper body state in steady-state walking." Biology letters 10.9 (2014): 20140405.

Mahaki, Mohammadreza, Sjoerd M. Bruijn, and Jaap H. Van Dieën. "The effect of external lateral stabilization on the use of foot placement to control mediolateral stability in walking and running." PeerJ 7 (2019): e7939.

van Leeuwen, Anina Moira, et al. "Active foot placement control ensures stable gait: Effect of constraints on foot placement and ankle moments." bioRxiv (2020).

footplacement's Issues

nanR2

There is a call to nanr2 in the footplacementfunction that should be removed

Run possibility?

Basically, the run function used for Mohammed is almost the same, except for some lines at the start. Include an extra flag "isrunning" ?

documentation

Documentation is as of yet mostly lacking. Should be written

Improve to better handle shifts in data

get foot position not from sample 100, but (if input<50?) from the next step sample 25 (which should be mid-stance)

ORDER

Order now does relatively little as input. Should be changed. Also add option 'variable'

accumulate and plot functions

Accumulate data and plot functions should be included in the repo

Autocorrelation of CoM pos and Vel?

Jeroen smeets suggested that worse prediction in PD may be because COM more eratic. He suggested someting like autocorrelation to look into this. Maybe idea?

Time normalization hard coded

I was playing around with the code and tried to change time normalization (pred_samples) from 1:51 to 1:100, however 51 is hard coded in Normalizetimebase_step.m and therefore the function returned an error.

AP possibility

For AP, it may be safe to make sure that belt walking is transformed to overground walking. Also, some variable names should then be changed. This can be done by

1. adding input flag "direction"
2. if direction is AP, calculate speed during stance, and add to entire time series
3. switch between output names based on input "direction"