search

maidenone / atrias

Automatically exported from code.google.com/p/atrias
0 stars 0 forks source link

data is missing in log #167

Closed GoogleCodeExporter closed 8 years ago

GoogleCodeExporter commented 8 years ago 
In log files (observed for atc_eq_gait) hug amounts of samples get missing once 
in a while. The images below show the controller time over samples, that 
exhibits jumps and leg angles at one specific time of these jumps.

We have to figure out, if this problem is caused by rtt-ros integration or 
rosbag and how it can be addressed, e.g.

- increasing buffer size
- reducing the amount of data logged
- maybe there is a way to prioritize data

Original issue reported on code.google.com by tra...@gmail.com on 5 Feb 2013 at 5:32

Attachments:

GoogleCodeExporter commented 8 years ago 
I have yet to ask about this on ROS answers. I've searched for other people who 
might have had this problem, but it seems that no one else is logging as much 
data as fast as we are. I'll try to get on that tonight, once I have a bulk of 
my homework out of the way.

> - increasing buffer size

The buffer is not the issue in either Orocos or ROS. Currently, Orocos has a 
buffer of 100000 messages, and rosbag keeps an internal 250MB buffer. The data 
skips happen before even those 100000 messages or 250MB's worth of data is even 
generated.

> - reducing the amount of data logged

We could see if data is dropped even when, for example, something as simple as 
the motor torque controller is run.

> - maybe there is a way to prioritize data

Maybe separate rosbag processes with different priorities? (But if we see that 
even motor torque controller data is dropped, the problem is probably too deep 
for this to make a difference.)

Original comment by yoos117@gmail.com on 5 Feb 2013 at 5:48

GoogleCodeExporter commented 8 years ago 
I'm beginning to work on the "reduce the amount of data logged" option.

Original comment by jrvanwhy on 14 Feb 2013 at 9:50

GoogleCodeExporter commented 8 years ago 
Daniel has reported that reducing the amount of data logged appears to have 
solved this issue.

Original comment by jrvanwhy on 27 Mar 2013 at 12:00