how to fix very low frame publish rate of turtlebot
Hi all,
I'm trying to use moveit it to control a turtlebot arm mounted on turtlebot. When I run my moveit.launch file, there's always a transform error:
[ERROR] [1543219008.768123467]: Transform error: Lookup would require extrapolation into the future. Requested time 1543219008.716465048 but the latest data is at time 1543219008.605108938, when looking up transform from frame [wheel_left_link] to frame [camera_rgb_optical_frame]
[ERROR] [1543219008.768212737]: Transform cache was not updated. Self-filtering may fail.
[ERROR] [1543219009.818338395]: Transform error: Lookup would require extrapolation into the future. Requested time 1543219009.718358101 but the latest data is at time 1543219009.712928056, when looking up transform from frame [arm_wrist_flex_servo_link] to frame [camera_rgb_optical_frame]
[ERROR] [1543219009.818410744]: Transform cache was not updated. Self-filtering may fail.
[ERROR] [1543219011.771677221]: Transform error: Lookup would require extrapolation into the future. Requested time 1543219011.720434783 but the latest data is at time 1543219011.650256141, when looking up transform from frame [wheel_left_link] to frame [camera_rgb_optical_frame]
[ERROR] [1543219011.771738074]: Transform cache was not updated. Self-filtering may fail.
I then used tf view_frames to check the rate of each frame is pubshed at, and I found there are few frames are published at a very low rate:
Is this low publishing rate causing the transform error? Why those tf frames are published at different rate while they were published by the same node? How can I fix this?
eidt 1:
CPU usage:
top - 08:03:29 up 5:35, 5 users, load average: 0.15, 0.20, 0.14
Tasks: 239 total, 1 running, 194 sleeping, 0 stopped, 0 zombie
%Cpu(s): 3.4 us, 1.3 sy, 0.0 ni, 95.0 id, 0.1 wa, 0.0 hi, 0.2 si, 0.0 st
KiB Mem : 8057792 total, 6119848 free, 1028348 used, 909596 buff/cache
KiB Swap: 8278012 total, 8278012 free, 0 used. 6590016 avail Mem
PID USER PR NI VIRT RES SHR S %CPU %MEM TIME+ COMMAND
18363 river 20 0 1231972 42708 23304 S 12.3 0.5 0:32.60 nodelet
19677 river 20 0 1680908 131768 54924 S 5.6 1.6 0:26.04 move_group
17799 river 20 0 953512 21844 16036 S 1.7 0.3 0:05.06 nodelet
17877 river 20 0 1355884 74276 11640 S 1.7 0.9 0:05.22 python
17883 river 20 0 753564 51512 7220 S 1.0 0.6 0:02.52 python
17797 river 20 0 460608 15508 13452 S 0.7 0.2 0:01.63 robot_state_pub
18640 river 20 0 578952 13440 12068 S 0.7 0.2 0:00.32 nodelet
8 root 20 0 0 0 0 I 0.3 0.0 0:48.82 rcu_sched
928 root 20 0 337240 46872 26520 S 0.3 0.6 0:23.62 Xorg
17753 river 20 0 324964 55484 7684 S 0.3 0.7 0:00.69 roslaunch
17833 river 20 0 578952 13408 12048 S 0.3 0.2 0:00.43 nodelet
18346 river 20 0 ...
My first guess would be that your maxing out the CPU of the on-board computer. Can you run the linux command 'top' on the turtlebot while everything is running to test to hypothesis.
Please see the edit for the 'top' results. It seems not very bad to me. Thank you!
Can you please post it as text using copy and paste. Images can't be searched, and I can't read it on my phone at all.
Yes of course. Please see the updated edit. Hope the format works. Thank you!
How many cores does the on board computer have? The 12.3% looks suspiciously like a maxed core of an 8 core system.
Oh... Yes the on board pc does have a 8 cores CPU. It's a 7th i7 of a NUC. Is there anything I can do to solve this problem? I couldn't find many discussions regarding multi core moveit. Thank you!
It seems very strange that some of the joints TFs are being published at a normal high rate while others are very slow. Can you think of anything common to all the slow ones? Such as a different hardware connection, or something similar.
Yes they actually have something in common. Those other links being published at high rate are all fixed links, while those few being published at low rates are attached to active joints. I tried to increase the publish frequency of the robot state publisher in the turtlebot bringup launch, but...
... the maximum rate I can get is 20.819 Hz, which is the same as the rate of the /odom being published. HOWEVER, I just found this afternoon that even I keep getting those transform errors, moveit seems working actually. I can plan and execute trajectories for the arm and I also get the octomap.