[Bug]: manipulability reverses start and end when selecting a robot subchain

未关闭 适合新手
#943 1 条评论 0 个 reaction 已指派 0 人 在 GitHub 查看

维护者通常 1 天内回复

还没有人认领这个 Issue。

评估

难度
2/5
预计耗时
1-3 小时
新手友好度
76/100
Issue 类型
缺陷
描述清晰度
描述清楚
活跃度
活跃
技术栈
numpy, python
领域
backend, robotics

调研方向

找到 robot wrapper 的 manipulability 方法,并检查它如何将 start 和 end 转发给 self.ets;issue 指出了参数顺序颠倒的问题。针对显式默认 end、仅 end、仅 start 以及两端都指定的子链,添加有针对性的检查,并使用报告中的 Panda 示例和预期值。选定链的结果与 jacob0 一致,且显式默认调用与隐式调用结果相同,即为完成。

由索引模型根据 Issue 内容生成。

描述

bug
RTB version

1.5.0, source main cee0023535988e4ae9490c81d639a46f2311b8fd; spatialmath-python 1.1.19; NumPy 2.5.5; native _fknm_c available

Python version

3.12.14

Operating system

Linux

Expected behavior

Selecting start and end should use the same kinematic chain as jacob0 with the same arguments. Supplying the default end explicitly should also agree with leaving it implicit. For the configuration below, the inverse-condition measure for panda_hand should be approximately 0.09476812042044401. For the subchain from panda_link2 to panda_link6, the minimum singular value should be approximately 0.175997031437505.

Actual behavior

Explicit end="panda_hand" raises ValueError: cond is not defined on empty arrays, whereas the default-end call and the measure computed from the selected Jacobian agree. Selecting panda_link2 to panda_link6 returns approximately 0.27532097160945784 instead of 0.175997031437505.

The robot wrapper builds its sequence with self.ets(end, start), but the sequence API takes start first and end second. This chooses a different chain (or a joint-free chain for an explicitly selected default end). The default call with both arguments omitted is unaffected. This was reproduced with the compiled backend on the source revision above. A focused fix would correct argument forwarding and add controls for explicit defaults, end-only, start-only, and two-ended subchains, without changing the manipulability formulas or public API.

Minimal reproducible example
import numpy as np
import roboticstoolbox as rtb

panda = rtb.models.Panda()
q = np.array([0.3, -0.5, 0.2, -1.8, -0.4, 1.4, 0.6])

selection = dict(start="panda_link2", end="panda_link6")
J = panda.jacob0(q, **selection)
print(panda.manipulability(J=J, method="minsingular"))
print(panda.manipulability(q, method="minsingular", **selection))
# Expected both: 0.175997031437505
# Actual second: 0.27532097160945784

print(panda.manipulability(q, method="invcondition"))
# 0.09476812042044401
print(panda.manipulability(q, end="panda_hand", method="invcondition"))
# ValueError: cond is not defined on empty arrays
主要语言
C++
星标
3.6k
派生
652
平均合并
1 天 23 小时
30 天内合并 PR
92

环境准备

从这里开始

  1. 先读完整个 Issue,再读项目的贡献指南。
  2. 在 Issue 下留言说明你要接手 —— 这能避免两个人做同样的事。
  3. Fork 仓库,在一个分支上完成修改。
  4. 提交 Pull Request,并在描述里引用这个 Issue 编号。

petercorke/robotics-toolbox-python 的其他 Issue

查看 petercorke/robotics-toolbox-python 的全部 Issue

相似的 Issue

更多 C++ Issue

把新 issue 发到你的邮箱

精选适合新手参与的 GitHub issue 摘要。