Param space plot/nonconvexity

[avikde]

Param space plot / nonconvexity

Turn down wΔy=1e2

Opt Φ=90, Qdt=0.0, minal=180, τ2/1 lim=2.0 => -2, [22.748 2.473 1.183 4.477 4.921 91.06 0.096], fHz=130.7, al[mg]=191.4, u∞=108.5, FD∞=73.7, pow=13.6, J=79.5, AR=4.0, x=30.0 24.695961 seconds (162.23 M allocations: 21.301 GiB, 21.14% gc time) Js [0.18813421737405006, 79.23286479554557, 0.059007717007310626, 0.0, 0.0]

Lowering feasible space to make f >= 0.18

Opt Φ=90, Qdt=0.0, minal=180, τ2/1 lim=2.0 => 0, [16.365 2.462 0.85 4.084 4.924 65.457 0.069], fHz=180.0, al[mg]=181.2, u∞=100.3, FD∞=53.0, pow=15.1, J=71.3, AR=4.0, x=25.4 1.607173 seconds (8.63 M allocations: 1.162 GiB, 18.05% gc time) Js [0.24871056906253575, 70.99616216373754, 0.07337037931690314, 0.0, 0.0]

Plotted with objective at the first set (green = these params, red = other params)

with objective at the second set

[avikde]

In #140

[avikde]

A number of improvements in #140

• modify opt settings
• clearly found an example of the problem
• plot objective from both opts

Objective looks fairly simple from both opt results

[avikde]

It does actually look like it gets to a good part of the contours plotted. Red dot is what is returned in each case; top row = ret2, bottom row = ret3

[avikde]

Plotting cost between the two points along a line segment between them does not reveal any problems

Just the cost function J itself is different.

• need to figure out how to visualize that