[01.02.16, 17:58] Greg Horn: my design variables used to be {parameters, x0, u0, x1, u1, .... xn-1, un-1, xn}
[01.02.16, 17:59] Greg Horn: i mean, i am doing dir coll but writing multiple shooting to save typing
[01.02.16, 17:59] Adrian Bürger: ok
[01.02.16, 18:00] Greg Horn: now my design variables are {parameters, x0, u0, p0, x1, u1, p1, .... xn-1, un-1, pn-1, xn} and I add constraints {parameters - p0, parameters - p1, ... parameters - pn-1} == 0
[01.02.16, 18:00] Greg Horn: casadi can color the hessian for this problem with better complexity
[01.02.16, 18:01] Greg Horn: i have seen it go from 30 seconds to 1 second recently, and i expect on larger horizons to go from 10 minutes to 10 seconds
[01.02.16, 18:01] Greg Horn: that is initialization
[01.02.16, 18:01] Greg Horn: runtime seems unaffected
[01.02.16, 18:01] Adrian Bürger: that's impressive!
[01.02.16, 18:02] Adrian Bürger: I'll write that down to try with casiopeia
[01.02.16, 18:02] Adrian Bürger: ☺
[01.02.16, 18:02] Greg Horn: (i am doing variable end time problems, and i group TF with p)
[01.02.16, 18:02] Greg Horn: so also do the same trick for TF
[01.02.16, 18:02] Greg Horn: you have to only constrain parameters though, not {p_k}, for LICQ
[01.02.16, 17:58] Greg Horn: my design variables used to be {parameters, x0, u0, x1, u1, .... xn-1, un-1, xn} [01.02.16, 17:59] Greg Horn: i mean, i am doing dir coll but writing multiple shooting to save typing [01.02.16, 17:59] Adrian Bürger: ok [01.02.16, 18:00] Greg Horn: now my design variables are {parameters, x0, u0, p0, x1, u1, p1, .... xn-1, un-1, pn-1, xn} and I add constraints {parameters - p0, parameters - p1, ... parameters - pn-1} == 0 [01.02.16, 18:00] Greg Horn: casadi can color the hessian for this problem with better complexity [01.02.16, 18:01] Greg Horn: i have seen it go from 30 seconds to 1 second recently, and i expect on larger horizons to go from 10 minutes to 10 seconds [01.02.16, 18:01] Greg Horn: that is initialization [01.02.16, 18:01] Greg Horn: runtime seems unaffected [01.02.16, 18:01] Adrian Bürger: that's impressive! [01.02.16, 18:02] Adrian Bürger: I'll write that down to try with casiopeia [01.02.16, 18:02] Adrian Bürger: ☺ [01.02.16, 18:02] Greg Horn: (i am doing variable end time problems, and i group TF with p) [01.02.16, 18:02] Greg Horn: so also do the same trick for TF [01.02.16, 18:02] Greg Horn: you have to only constrain parameters though, not {p_k}, for LICQ