Let the param opt change the timestep

[avikde]

This is needed for W2D if leaving the stiffnesses fixed and scaling since inertial terms get higher

Identified the problem in #101

• currently directly have Hk*pt
• this will have many side-effects: need to modify the fixeddt idea and use the one from params
• what if it tries to set it too low or too high? shouldn't since i+s cancelling is a good way to minimize torque
• on the plus side, with the various non-dimensional templates, usually there is some time normalization, and this may be needed in general

Conclusions of this PR

• This is working and is quite useful: see the itot, gtot in the plots in Let the param opt change the timestep #102 (comment) or Let the param opt change the timestep #102 (comment) -- they are quite close up to some minal
• after some minal this doesn't work any more -> due to the large wing, the inertial force is too high. switches to act force being in opp phase to the stroke.

[avikde]

Affine test previous commit: f729b6c

 0.0815238  0.320188  0.45894   0.453797  0.346806  0.175307  …  0.751371  0.417824  0.152845  -0.141091  -0.258705  -0.127218
 1.19676    0.85242   0.657917  0.517529  0.359011  0.172148     0.35619   0.243886  0.382706   0.963683   1.45447    1.58034

affine test after the dt split

 0.0815238  0.320188  0.45894   0.453797  0.346806  0.175307  …  0.751371  0.417824  0.152845  -0.141091  -0.258705  -0.127218
 1.19676    0.85242   0.657917  0.517529  0.359011  0.172148     0.35619   0.243886  0.382706   0.963683   1.45447    1.58034

i.e. the dt split is good. (These are both bad because this is from the truncated sim without fix)

[avikde]

Opt ran with the dt in the optimization: Opt minal=1.0, τ2/1 lim=2.0 => [3.499 17.501 0.565 6.184 3.929 15.432 0.132]

Slightly different params than the last of #101

[avikde]

Just need to make sure this helps alleviate the issue of needing to change stiffness. Scale params for lift and let it change dt, and check that i+s is small in each case.

Debug components results at different minal; letting it opt freq

0.6: Opt minal=0.6, τ2/1 lim=2.0 => [2.096 16.95 0.335 2.307 2.418 33.134 0.076], fHz=290.7

0.8: Opt minal=0.8, τ2/1 lim=2.0 => [2.795 17.62 0.447 4.294 3.659 10.752 0.091], fHz=243.3

1.0: Opt minal=1.0, τ2/1 lim=2.0 => [3.493 17.942 0.559 7.208 5.11 0.001 0.103], fHz=216.4

[avikde]

Scale params for lift

This is for minal = 0.6:0.2:1.6

• Unfortunately T2 goes to 0 again.

Debugging components

• Components @ 1.6

Initial traj δt=0.135, opt.fixedδt=0.135
Avg lift initial [mN]=0.916, in [mg]=93.4
Opt minal=1.6, τ2/1 lim=2.0 => [5.59 17.942 0.894 24.701 12.744 0.0 0.11], fHz=202.8, al[mg]=165.1, u∞=96.5

• i+s is quite large. looks like stiffness terms should be a lot higher
• Increase stiffnesses to ko=0.55, ka=200

Initial traj δt=0.135, opt.fixedδt=0.135
Avg lift initial [mN]=0.635, in [mg]=64.7
Opt minal=1.6, τ2/1 lim=2.0 => [8.063 14.529 1.29 78.51 32.985 0.0 0.093], fHz=238.3, al[mg]=164.7, u∞=100.2

• increasing cbar?? and increasing freq?? but i+s remaining large

Concrete fixes to make

• the new dt is not used in plotTrajs x axis Let the param opt change the timestep #102 (comment)
• is dt relevant in computing the avg lift?
• need to debug why the initial traj stroke is changing: can be seen with makeplot=true Let the param opt change the timestep #102 (comment)

[avikde]

Unfortunately now I'm not getting the same results as in #102 (comment)

Opt minal=0.6, τ2/1 lim=2.0 => [2.158 18.88 0.345 2.493 2.719 0.001 0.076], fHz=293.5, al[mg]=61.1, u∞=14.9

With lower gains to createInitialTraj
Opt minal=0.6, τ2/1 lim=2.0 => [2.078 17.039 0.333 2.224 2.402 33.128 0.076], fHz=294.1, al[mg]=61.0, u∞=16.4

Need to think about:

• why does this change with gain?
• makes more sense to use an input sine to generate the output traj?

Use open-loop force input to generate initial traj

Opt minal=0.6, τ2/1 lim=2.0 => [2.029 17.657 0.325 2.064 2.372 34.147 0.075], fHz=297.2, al[mg]=60.1, u∞=15.1

Opt minal=1.0, τ2/1 lim=2.0 => [3.382 18.681 0.541 6.695 4.889 0.001 0.101], fHz=220.0, al[mg]=101.7, u∞=39.7

Opt minal=1.6, τ2/1 lim=2.0 => [5.411 18.681 0.866 22.633 12.037 0.0 0.109], fHz=203.8, al[mg]=165.5, u∞=93.1

[avikde]

Scale params for lift with new traj - almost identical results as before.