I propose this PR. Unfortunately It turned out being a little too "bulky".
1) I am working on a third solver based on qpOASES and to avoid errors and duplicated code, I propose to use a base class and to have ChainIkSolverPos_TL and NLOPT_IK inheriting from it.
2) Some minor cleanups. Here and there. In particular, I suggest a more consistent namespace and less copy and paste. I also prefer less cryptic names. I understand that this might occasionally fall into the "personal taste" field.
3) A less misleading information from the benchmark test. Instead of informing the user of the average time including both successful and failed IK, inform only of the average speed of successful samples. This is IMO a less biased source of information, since it actually inform about the ability of the algorithm to converge.
Hi,
I propose this PR. Unfortunately It turned out being a little too "bulky".
1) I am working on a third solver based on qpOASES and to avoid errors and duplicated code, I propose to use a base class and to have ChainIkSolverPos_TL and NLOPT_IK inheriting from it.
2) Some minor cleanups. Here and there. In particular, I suggest a more consistent namespace and less copy and paste. I also prefer less cryptic names. I understand that this might occasionally fall into the "personal taste" field.
3) A less misleading information from the benchmark test. Instead of informing the user of the average time including both successful and failed IK, inform only of the average speed of successful samples. This is IMO a less biased source of information, since it actually inform about the ability of the algorithm to converge.