weixin_39843151
weixin_39843151
2020-11-26 10:20

[amcl_laser] Revert the measurement computation to what is in Probabilistic Robotics textbook.

Originally raised in this QandA.

I'm also wondering how this ad-hoc weight was developed. I understand that this ad-hoc value has been functioning for the purpose, but ideally should better be to be documented with a mathematical rationale considering it's based on a textbook.

I don't expect this to be merged as it is, or even understand if rejected; if the rationale for the current ad-hoc modification is provided, I think that would be fine. And we might want to keep using a long-standing logic rather than modifying it, unless new logic is tested thoroughly or gone through a heavy review process.

该提问来源于开源项目:ros-planning/navigation

  • 点赞
  • 写回答
  • 关注问题
  • 收藏
  • 复制链接分享
  • 邀请回答

5条回答

  • weixin_39612023 weixin_39612023 4月前

    Note that this has been this way since at least 2009 based on blame

    点赞 评论 复制链接分享
  • weixin_40004960 weixin_40004960 4月前

    Have you tested this change and seen it working effectively? I tried making the change in some stuff that I've been working on and it resulted in scoring not working correctly.

    点赞 评论 复制链接分享
  • weixin_39843151 weixin_39843151 4月前

    Not tested exhaustively. As mentioned I created this PR more for raising an attention.

    But if you have a .bag file that you can share publicly, that'll be great (not just for me. We might be able to integrate that into the system test).

    点赞 评论 复制链接分享
  • weixin_40004960 weixin_40004960 4月前

    I'm not able to release what I've done, sorry. :disappointed: But, I understand the point you've raised and I appreciate it.

    点赞 评论 复制链接分享
  • weixin_39612023 weixin_39612023 4月前

    I'm pretty sure if we were to change this, everyone would get pretty angry. If you really want to get something along these lines in, I would suggest adding a new parameterized approach that can be enabled by choice and opening a new PR against K or L.

    点赞 评论 复制链接分享