[jsk_data] Use different value for duration and rate in tf_to_transform.py

[wkentaro]

Rate can be 50 - 100 for example, but duration should be ~1 [s] even so. In
previous implementation, the duration will be 1/100 - 1/50 [s] and it is
too small to resolve tf.

[wkentaro]

It needed to be Exception

Exception in thread Thread-5:
Traceback (most recent call last):
  File "/usr/lib/python2.7/threading.py", line 810, in __bootstrap_inner
    self.run()
  File "/opt/ros/indigo/lib/python2.7/dist-packages/rospy/timer.py", line 223, in run
    self._callback(TimerEvent(last_expected, last_real, current_expected, current_real, last_duration))
  File "/home/wkentaro/Projects/jsk_apc/src/jsk-ros-pkg/jsk_common/jsk_topic_tools/scripts/tf_to_transform.py", line 32, in _publish_transform
    timeout=rospy.Duration(self.duration)
Exception: Lookup would require extrapolation into the past.  Requested time 1489937523.406011105 but the earliest data is at time 1489937523.459053278, when looking up transform from frame [right_hand_camera_rgb_optical_frame] to frame [base]. canTransform returned after 1.00738 timeout was 1.

[knorth55]

what do you mean?

[wkentaro]

I explain why I needed this change.
Catching the Exception is not ideal solution.

[wkentaro]

@knorth55 Please review.

[knorth55]

+1

[wkentaro]

Please merge this.

[k-okada]

please list up the code that affected by this change.

[wkentaro]

Maybe the code used by @knorth55 .

[knorth55]

I made this program, and i think there is yet no one or repo using this program.
it is not that big change, and i can manage it.

[knorth55]

GOOD!

[k-okada]

@knorth55 : please provide link to the code using this program

[knorth55]

https://github.com/start-jsk/jsk_apc/blob/master/jsk_2016_01_baxter_apc/launch/bimanual_stow_data_collection.launch

[k-okada]

ok, merged, but if this code is only used one place, that is not a common program....

[knorth55]

i thought it can be used in various program for logging.
maybe no one knows its existence.

[k-okada]

btw why tf.LookupTransform is not ok? #1482 (comment) Too much data?

…

-- ◉ Kei Okada 2017-03-22 16:26 GMT+09:00 Shingo Kitagawa <notifications@github.com>:

i thought it can be used in various program for logging. maybe no one knows its existence. — You are receiving this because you modified the open/close state. Reply to this email directly, view it on GitHub <#1509 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AAeG3CRw8ZAtWYF9twNkS8JAymkWFOKYks5roM0NgaJpZM4MhxEG> .

[knorth55]

Im using lookupTransform in this program.

[wkentaro]

i thought it can be used in various program for logging.
maybe no one knows its existence.

Yeah. This program is originally created for data_collection_server.py.
It should be useful for people who have motivation for learning-based approach by collecting data with robot, but there are few people who have motivation about it.

It is used at projects like:

• Humanoids2016 TransformableMap
• APC2016 inBinSegment
• APC2016 inHandClassify
• APC2016 inStowBimanual
• APC2017 inBinSegment

but current active users are only me and @knorth55

[k-okada]

I mean why you do not log tf data? because it becomes too large?

ok, merged, but if this code is only used one place, that is not a common

program.... ok, at least this is used in 5 different places......

…

-- ◉ Kei Okada 2017-03-22 20:01 GMT+09:00 Shingo Kitagawa <notifications@github.com>: Im using lookupTransform in this program. — You are receiving this because you modified the open/close state. Reply to this email directly, view it on GitHub <#1509 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AAeG3FcMuOKcKD98KbwvzGxg9CJFjzfGks5roP-LgaJpZM4MhxEG> .

-- -- ◉ Kei Okada

[wkentaro]

I mean why you do not log tf data? because it becomes too large?

Because /tf cannot be subscribe for synchronizing with other message,
and this means it cannot be saved with data_collection_server.py.

ok, at least this is used in 5 different places......

Ah, actually I mean data_collection_server.py.
So this program tf_to_transform.py is only used once by his bthesis.

[k-okada]

I see, another aspect of understanding '/tf' is, Although tf is pub/sub through topic, but has different layer than other topics, because it has strong tf library and the message is cached in tf client or tf2 server. So we can use tf as a very basic information layer, like on euslisp, we use coordinates as a base class of most of class definition (https://github.com/euslisp/jskeus/blob/master/irteus/irtpointcloud.l#L31)
So another design of data_collection_server was whenever it collect data, it also LookUpTransform of all necessary transforms.

[wkentaro]

Yeah, that is the another way.
data_collection_server.py is in the early stage, so surely there will be some changes of design.