28 dev kepler add examples

[antoinepichon03]

No description provided.

[ocots]

Todo :

• Augmenter poussée max (par exemple 60N) et diminuer le temps final (tf = 50)
• Augmenter nombre de pas de temps : 200, 500, 1000...
• Faire de la conso min régularisée avec la barrière log
• Faire du Trapèze au lieu de Euler

[ocots]

@antoinepichon03 Can you please push your updates and make a point on what you have done since last meeting and what you still have to do? You can put some details, regarding what we wrote on the blackboard (which was white).

[antoinepichon03]

@antoinepichon03 Can you please push your updates and make a point on what you have done since last meeting and what you still have to do? You can put some details, regarding what we wrote on the blackboard (which was white).

Hello,
Sorry, I hadn’t seen the message. I am currently trying to initialize the two-body problem with minimal consumption using the minimum time solution so that the model converges with Optimal Control. For now, it still does not converge… I’m drawing inspiration from what has already been done to fix this.
Regarding the solution with Jump, it seems to work even when I vary the parameters (as long as the solution remains admissible).

[ocots]

@antoinepichon03 Can you please push your updates and make a point on what you have done since last meeting and what you still have to do? You can put some details, regarding what we wrote on the blackboard (which was white).

Hello, Sorry, I hadn’t seen the message. I am currently trying to initialize the two-body problem with minimal consumption using the minimum time solution so that the model converges with Optimal Control. For now, it still does not converge… I’m drawing inspiration from what has already been done to fix this. Regarding the solution with Jump, it seems to work even when I vary the parameters (as long as the solution remains admissible).

It is interesting to archive the fact that JuMP is more robust wrt to initialization. Could you write directly in the issue #28, some results showing this? Maybe you can write the problem, the different resolutions and the trace.

[jbcaillau]

@antoinepichon03 @ocots will be worth to give a try with :exa (and Ipopt / MadNLP) solving in v1.1. should do as well or better than JuMP 🤞🏽

It is interesting to archive the fact that JuMP is more robust wrt to initialization. Could you write directly in the issue #28, some results showing this? Maybe you can write the problem, the different resolutions and the trace.