Stop motion when collisions are detected

[k-okada]

start-jsk/rtmros_common#247
@kenji0919

[130s]

I take this over

[130s]

Title changed, becauase servo-off at collision situation might just make things even worse by releasing held objects etc.

• old) servo off (or stop motion) when collisions are detected
• new) Stop motion when collisions are detected

After #89 is merged, which is a visualizer of detected collision, we can add a feature to stop the motion (maybe in hrpsys_ros_robotics). Also, related to the comment #34 (comment), we need a "stop" feature in the upstream (hrpsys-base is the good one to ask at).

[130s]

Ticketed

[k-okada]

this is always discussion on which is better because servo off has less
chance to break robot (because another approach is that robot have chance
to stop motion even if there are collision, due to model error, that means
the servo system keep running and increase motor/circuit temperature and
may lead to burn), but offcourse we do not want to drop off the
tools/object if servo off. Usually academic research says ok for servo off
and industrial people ok with stop motion, because in the industrial
people, they think there are always one person who hold emergency button,
when ever robot stops motion, then main power can cut off by human, but is
is not realistic for small lab.

• old) servo off (or stop motion) when collisions are
• new) Stop motion when collisions are detected

On Saturday, May 3, 2014, Isaac Isao Saito notifications@github.com wrote:

Title changed, becauase servo-off at collision situation might just make
things even worse by releasing held objects etc.

• old) servo off (or stop motion) when collisions are detected
• new) Stop motion when collisions are detected

After #89 #89 is merged,
which is a visualizer of detected collision, we can add a feature to stop
the motion (maybe in hrpsys_ros_robotics).

—
Reply to this email directly or view it on GitHubhttps://github.com//issues/36#issuecomment-42090879
.

[130s]

Thank you for the valuable summary (@fkanehiro cites the similar concern at fkanehiro/hrpsys-base#215 (comment) as well). Humbly speaking I have no idea what the ultimate decision must become like, but at this moment it looks like the discussion diverges after good length of period, so I'd suggest to add first a small, simple feature to Stop Motion to give users more options (and also we can test and evaluate the feature). Thoughts anyone?

[k-okada]

currently you can achieve stop motion by

• get current angle vector
• send that data to the robot with time = 0.1 ( I'm not sure if time = 0 works)

so you can implement this in hironx_client.py level and may be this is good starting point.

[k-okada]

Of course, if you already working on this, you do have to stop that.

[k-okada]

@130s, @emijah what does stop mean? if this is just servo off, then I can fix soon, if this means stop motion and need to resume, we have to wait for fkanehiro/hrpsys-base#215 discussion

[130s]

@130s, @emijah what does stop mean? if this is just servo off, then I can fix soon, if this means stop motion and need to resume, we have to wait for fkanehiro/hrpsys-base#215 discussion

I meant stop motion, resume as per the title, and wonder what I can do to proceed fkanehiro/hrpsys-base#215 discussion.

[k-okada]

i think fkanehiro/hrpsys-base#215 may takes long
for both discussion and implementation, so my suggestion is to start with
just stop function.

On Fri, Jul 11, 2014 at 4:32 PM, Isaac I.Y. Saito notifications@github.com
wrote:

@130s https://github.com/130s, @emijah https://github.com/emijah what
does stop mean? if this is just servo off, then I can fix soon, if this
means stop motion and need to resume, we have to wait for
fkanehiro/hrpsys-base#215
fkanehiro/hrpsys-base#215 discussion

I meant stop motion, resume as per the title, and wonder what I can do to
proceed fkanehiro/hrpsys-base#215
fkanehiro/hrpsys-base#215 discussion.

—
Reply to this email directly or view it on GitHub
#36 (comment)
.

[DaikiMaekawa]

I took over being in charge of this issue.

[emijah]

I think servo-stop is best, because the purpose of this collision detection is to prevent collision, so the problem is detected in advance. If there was an unforeseen collision, it's probably best to leave it to the safety features of the servo control, and let them switch the servo motors off.
Ideally, only the servo detecting excessive force should switch off, and allow switching on again, but the current system cuts the power supply to all motors so this is not possible.

[k-okada]

nice,
please follow discussion here ->
start-jsk/rtmros_common#618,
fkanehiro/hrpsys-base#431,
fkanehiro/hrpsys-base#425. They do not reach to
the final solution, but it might be interesting to you.

I always like to start from most easiest and simplest approach and move to
more difficult version, so

• running collision detector component on ubuntu machine, when it detect
  collision, just to send servo off command (
  古いor特殊な体内環境等でのCollisionDetectorの運用 fkanehiro/hrpsys-base#425 (comment),
  https://github.com/start-jsk/rtmros_common/blob/master/hrpsys_ros_bridge/src/hrpsys_ros_bridge/hrpsys_dashboard.py#L178
  )
• running collision detector component on QNX machine, this "should" work,
  but not sure, because vclip is not able to work on 32bit machine until
  recently (32bitマシンでのvclipの挙動の修正 fkanehiro/hrpsys-base#352), so we're
  afraid that there might be some problem on QNX.
• running collision detector plugin inside of hrpsys(rtcd), this fill
  suppress motor commend when it find collision, so you do have to "servo
  off" in this setup

Also using fcl (https://github.com/flexible-collision-library/fcl) instead
of vclip is also related topic (
fkanehiro/hrpsys-base#97)

◉ Kei Okada

On Mon, Dec 22, 2014 at 1:39 AM, Daiki Maekawa notifications@github.com
wrote:

I took over being in charge of this issue.

—
Reply to this email directly or view it on GitHub
#36 (comment)
.

[130s]

See #250, which may or may not be the complete replacement of this. Feel free to reopen if further discussing is needed.