Home | Tutorials | Wiki | Issues
Ask Your Question
0

DRCSIM: performance gone? [closed]

asked 2013-03-13 15:33:35 -0600

cga gravatar image

updated 2013-03-13 20:08:12 -0600

On my dual CPU Intel(R) Xeon(R) CPU E5-2687W 0 @ 3.10GHz I am now getting a Real Time Factor of 0.06 for the following:

source /usr/share/drcsim/setup.sh
roslaunch atlas_utils vrc_task_2.launch

I rebooted, and downloaded drcsim from bitbucket at 4pm EDT, and still performance is bad. This is gazebo-1.5 and the current drcsim-2.2.?

The only update I did today was libbullet.

The CPU load (below) is actually less than when I run the qual_task_4, which almost runs at real time (3rd column in ps augx printout):

cga      11078  0.5  0.0 291264 19224 pts/0    Sl+  16:27   0:00 /usr/bin/python /opt/ros/fuerte/bin/roslaunch atlas_utils vrc_task_2.launch
cga      11094  0.6  0.0 570332 14488 ?        Ssl  16:27   0:00 /usr/bin/python /opt/ros/fuerte/bin/rosmaster --core -p 11311 __log:=/home/cga/.ros/log/7a4e2c74-8c1c-11e2-99ac-90b11c69ff24/master.log
cga      11107  0.3  0.0 315416  6592 ?        Ssl  16:27   0:00 /opt/ros/fuerte/share/rosout/bin/rosout __name:=rosout __log:=/home/cga/.ros/log/7a4e2c74-8c1c-11e2-99ac-90b11c69ff24/rosout-1.log
cga      11121  0.0  0.0   4404   620 ?        Ss   16:27   0:00 /bin/sh /usr/share/drcsim-2.2/ros/atlas_utils/scripts/run_gazebo vrc_task_2.world __name:=gazebo __log:=/home/cga/.ros/log/7a4e2c74-8c1c-11e2-99ac-90b11c69ff24/gazebo-2.log
cga      11122  0.0  0.0 421640 34764 ?        Sl   16:27   0:00 gazebo -s libgazebo_ros_api_plugin.so vrc_task_2.world
cga      11123  0.4  0.0 431964 16308 ?        Ssl  16:27   0:00 python /usr/share/sandia-hand-5.1/ros/sandia_hand_teleop/simple_grasp.py __name:=simple_grasp_left __log:=/home/cga/.ros/log/7a4e2c74-8c1c-11e2-99ac-90b11c69ff24/sandia_hands-l_hand-simple_
cga      11125  124  0.3 7198048 440040 ?      Sl   16:27   2:56 gazebo -s libgazebo_ros_api_plugin.so vrc_task_2.world
cga      11126 24.1  0.3 5983816 426092 ?      Sl   16:27   0:34 gazebo -s libgazebo_ros_api_plugin.so vrc_task_2.world
cga      11127  0.4  0.0 431960 16304 ?        Ssl  16:27   0:00 python /usr/share/sandia-hand-5.1/ros/sandia_hand_teleop/simple_grasp.py __name:=simple_grasp_right __log:=/home/cga/.ros/log/7a4e2c74-8c1c-11e2-99ac-90b11c69ff24/sandia_hands-r_hand-simple
cga      11130  1.6  0.0 503132 11568 ?        Ssl  16:27   0:02 /opt/ros/fuerte/stacks/robot_model/robot_state_publisher/bin/state_publisher joint_states:=/atlas/joint_states __name:=atlas_robot_state_publisher __log:=/home/cga/.ros/log/7a4e2c74-8c1c-11
cga      11145  1.4  0.0 503136 11568 ?        Ssl  16:27   0:02 /opt/ros/fuerte/stacks/robot_model/robot_state_publisher/bin/state_publisher joint_states:=/multisense_sl/joint_states __name:=multisense_sl_robot_state_publisher __log:=/home/cga/.ros/log/
cga      11165  1.5  0.0 543696  9852 ?        Ssl  16:27   0:02 /opt/ros/fuerte/stacks/geometry_experimental/tf2_ros/bin/buffer_server __name:=tf2_buffer_server __log:=/home/cga/.ros/log/7a4e2c74-8c1c-11e2-99ac-90b11c69ff24/tf2_buffer_server-7.log
cga      11199  0.5  0.0 2956832 11968 ?       Ssl  16:27   0:00 /opt/ros/fuerte/stacks/image_pipeline/stereo_image_proc/bin/stereo_image_proc __name:=stereo_proc __log:=/home/cga/.ros/log/7a4e2c74-8c1c-11e2-99ac-90b11c69ff24/multisense_sl-camera-stereo_

roslaunch atlas_utils vrc_task_1.launch

gets RTF = 0.45 (worse than with previous versions).

roslaunch atlas_utils vrc_task_3.launch

gets RTF = 0.31 (worse than with previous versions).

qual_task_1 is > 0.9
qual_task_2 is > 0.9
qual_task_3 is 0.85
qual_task_4 is > 0 ...
(more)
edit retag flag offensive reopen merge delete

Closed for the following reason question is not relevant or outdated by nkoenig
close date 2013-07-30 19:49:00.945728

Comments

"So the rubble is killing us, and the hands aren't helping."

You wanna know what really gets you killed? Try subscribing to the lidar scan in vrc world 2 and watch the RTF drop all the way from 1.3 to 0.02 (no rubble, no hands).

ThomasK gravatar imageThomasK ( 2013-03-13 20:45:01 -0600 )edit

1 Answer

Sort by ┬╗ oldest newest most voted
1

answered 2013-03-13 18:13:08 -0600

ThomasK gravatar image

What real-time factor did you used to get for the vrc tasks? I usually get 0.02-0.1 on an Intel® Xeon(R) CPU E5-1620 0 @ 3.60GHz ... so your numbers look pretty normal to me. :)

The problem with vrc task 2 is all the dynamically simulated rubble which slows it down incredibly. The RTF you're getting for tasks 3 and 1 are also in the ranges that I've always been getting.

The qual tasks run much faster as they don't have any complex environments (e.g. no huge hose that has to be simulated and no 50 pieces of rubble).

edit flag offensive delete link more

Question Tools

Stats

Asked: 2013-03-13 15:33:35 -0600

Seen: 285 times

Last updated: Mar 13 '13