How to interpret moco simulation parameter information
-
Jingke Song
- Posts: 34
- Joined: Tue Oct 19, 2021 4:52 am
How to interpret moco simulation parameter information
Hello everyone, what is the meaning of each parameter in each iteration when using Moco for optimization? As shown in the figure, and how to evaluate the results of each optimization iteration based on them?
- Attachments
-
- 1.JPG (33.27 KiB) Viewed 397 times
-
Nicholas Bianco
- Posts: 1044
- Joined: Thu Oct 04, 2012 8:09 pm
Re: How to interpret moco simulation parameter information
Hi Jingke,
These parameters are produced by IPOPT, which is the non-linear solver we use by default in Moco. You can find more information about these parameters here: https://coin-or.github.io/Ipopt/OUTPUT.html.
Best,
Nick
These parameters are produced by IPOPT, which is the non-linear solver we use by default in Moco. You can find more information about these parameters here: https://coin-or.github.io/Ipopt/OUTPUT.html.
Best,
Nick
-
Jingke Song
- Posts: 34
- Joined: Tue Oct 19, 2021 4:52 am
Re: How to interpret moco simulation parameter information
Hi Nick,
Thank you very much for your answer, but at the same time, I would like to ask you another question. In MOCO optimization, I often achieve ideal results within 100 iterations. However, MOCO will continue to optimize and iterate, and may not find the optimal solution until the 500th iteration. But the optimization results of both are almost identical. In the subsequent iterations, the objective value remained almost unchanged, which undoubtedly wasted a lot of simulation time. The control goal weight has been set within 0-10, and I have attempted to adjust the optimization convergence error, but it has almost no effect on the number of iterations during convergence. Can I quickly find an optimized solution through some measures.
Best,
Jingke
Thank you very much for your answer, but at the same time, I would like to ask you another question. In MOCO optimization, I often achieve ideal results within 100 iterations. However, MOCO will continue to optimize and iterate, and may not find the optimal solution until the 500th iteration. But the optimization results of both are almost identical. In the subsequent iterations, the objective value remained almost unchanged, which undoubtedly wasted a lot of simulation time. The control goal weight has been set within 0-10, and I have attempted to adjust the optimization convergence error, but it has almost no effect on the number of iterations during convergence. Can I quickly find an optimized solution through some measures.
Best,
Jingke
-
Nicholas Bianco
- Posts: 1044
- Joined: Thu Oct 04, 2012 8:09 pm
Re: How to interpret moco simulation parameter information
Hi Jingke,
The number of iterations and total time to convergence can vary greatly depending on the problem that you are trying to solve. It could take a complicated problem well over 500 iterations to solve if the mesh is too course or the objective function not scaled correctly.
What does your solution look like? Are there any actuators or joints hitting bounds? Anything that might indicate that the problem is having difficulty meeting the constraints?
-Nick
The number of iterations and total time to convergence can vary greatly depending on the problem that you are trying to solve. It could take a complicated problem well over 500 iterations to solve if the mesh is too course or the objective function not scaled correctly.
What does your solution look like? Are there any actuators or joints hitting bounds? Anything that might indicate that the problem is having difficulty meeting the constraints?
-Nick