stulp / dmpbbo

Python/C++ library for Dynamical Movement Primitives and Black-Box Optimization
GNU Lesser General Public License v2.1
224 stars 90 forks source link

pass the via point #81

Closed wangziyang123456 closed 1 year ago

wangziyang123456 commented 1 year ago

Excuse me, in the task of passing a point, I used a custom track and changed the point to pass. After learning, the generation curve passed through the via point, but it did not reach the end point. What is the reason? I have studied it myself and did not understand it.

stulp commented 1 year ago

Does #79 help?

stulp commented 1 year ago

Someone else was mentioning this, so I should have a closer look. Did you fork the code? Can you show me the exact commit where the problem arises? Thank you.

stulp commented 1 year ago

Thank you. The data alone is not sufficient to reproduce your observations. Please provide a link to the fork you have made, and the commit in which the problem arises. Only then can I reproduce what you see.

stulp commented 1 year ago

Sorry, I don't time to debug code in zip files.

I ran the scripts myself and could not reproduce the bug, but made several improvements to the code in 0781bc4cfed32c which should help you and others:

Note that demo_bbo_of_dmps_step_by_step.py writes intermediate results to a temporary directory. So make sure to delete this temporary directory (or make a backup somewhere in case it is useful) when running the script again (otherwise the script may read old data). You can also pass it a directory with demo_bbo_of_dmps_step_by_step.py --dir my_output_directory.

In general, if you do not need the intermediate results stored (i.e. you're just testing things to start out), then I recommend using demo_bbo_of_dmps.py. It has the same functionality, but is much faster because it does not write files to disk.

Hope it helps.

wangziyang123456 commented 1 year ago

Thank you for your reply. I've tried everything you said before. I'll try to find out what the problem is later. Thanks again.