EuRoC Forum

[dti]

The time left to solve task 2, as published on /time_budget seems to be following wall time. Thus if Gazebo cannot keep up (due to to high CPU load) there is in reality less time to solve the task.

Output of 'rosbag info' with a 600 second wall time run

Code: Select all

path:        2014-11-11T09-13-00_task_2.bag
version:     2.0
duration:    3:25s (205s)
start:       Jan 01 1970 01:00:37.05 (37.05)
end:         Jan 01 1970 01:04:02.68 (242.68)
size:        141.7 MB
messages:    803196
compression: none [177/177 chunks]
types:       rosgraph_msgs/Log  [acffd30cd6b6de30f120938c17c593fb]
             std_msgs/Bool      [8b94c1b53db61fb6aed406028ad6332a]
             std_msgs/Int32     [da5909fbe378aeaf85e547e830cc1bb7]
             std_msgs/String    [992ce8a1687cec8c8bd883ec73ca41d1]
             tf/tfMessage       [94810edda583a504dfda3829e70d7eec]
             tf2_msgs/TFMessage [94810edda583a504dfda3829e70d7eec]
topics:      active_task                  1 msg     : std_msgs/String   
             collision_human_robot    10282 msgs    : std_msgs/Bool     
             rosout                     960 msgs    : rosgraph_msgs/Log  (18 connections)
             target_frame               252 msgs    : std_msgs/String   
             tf                      791098 msgs    : tf/tfMessage       (11 connections)
             time_budget                603 msgs    : std_msgs/Int32


In this case, the real time factor as reported by Gazebo was around 0.3. As you can see this rosbag is only around 200 seconds long. Looking at the number of messages appearing on time_budget (=603 messages) time really is going too fast.

[Kapi.m]

Hi,

I can confirm this in task 1.2, as you can see in this output:

Code: Select all

[ INFO] [1415695427.586700258, 28.065000000]: com_client: Time remaining: 24
[ INFO] [1415695428.570983456, 28.362000000]: com_client: Time remaining: 23
[ INFO] [1415695429.565350058, 28.648000000]: com_client: Time remaining: 22
[ INFO] [1415695430.585212080, 28.960000000]: com_client: Time remaining: 21


The time between every time_budget message is 1s in wall time, but only 0.3s in simulated time...

Best,
Kapi

[michieletto]

Hi all,
we also confirm this issue.

Best.
Stefano

[admin]

Dear all,

The wall time was used intentionally to fix-limit the duration of the task.
What we can garanty is that each team will have the same amount of time (simulated and wall time).
The simulator runs on a standalone machine and the realtime factor is fixed to about 0.5.

For testing you can compensate for another realtime factor by adjusting the content of the variable task_time_budget_s in the file ipa325_com_server.py (number of seconds for each task)

The same holds true for task 1.

Best regards