by Jason Moore
I looked into how the identified controller parameters varied among two rider for several runs at the same speed.
Here I examine 26 runs at the average speed of 5.0 +/- 0.2 m/s with three different riders. The assumptions are as follows:
- I use the Whipple model linear about the upright configuration and populate the model parameters by measuring the physical properties of the bicycle and rider.
- I assume our controller with 6 free parameters (5 gains and the neuromuscular frequency).
- I identified the parameters based on five outputs: roll angle, steer angle, roll rate, steer rate, yaw rate.
This first plot gives an idea of how varialbe the identified parameters were for each rider. There is rather large variability in the identified parameters for Jason, less so for Charlie and even less so for Luke. Jason’s parameters don’t tend to have a normal distribution.
- These are boxplots showing the variability in each parameter of the control model as a function of each rider.
You can also look at this in the frequency domain. The following graphs show the lateral force transfer functions with respect to each output. Each rider is identified by a color (Charlie = blue, Jason = red, Luke = green). I’ve plotted a dark solid line as the mean magnitude and frequency with dotted lines bounding the standard deviation of each. This may not be the best visualization yet, but it shows that the system response is similar regardless of the variability in the identified parameters. This implies that the parameters may not be very unique identifiers of the controller. I’m not sure what to make of that just yet, except that relying only on the identified parameters to identify control has some issues.
- The phase plot is incorrect here, because of incorrect alignment of the various phase curves (i.e. the phases need to be matched because they could be +/- 360 degrees off…some extra programming to get it right).
