| ROS Resources: Documentation | Support | Discussion Forum | Index | Service Status | ros @ Robotics Stack Exchange |
I am using a Universal Robots UR5 mounted to a cylindrical post using fixed joints (one to world, the other to the UR5).
When I drive it in Gazebo using Moveit! the model shakes/wobbles. It eventually settles down....
I am wondering if there is something that I'm doing wrong to induce this? Or if there are things I need to adjust to eliminate this?
Hi b3l33 The gains of the controller and the inertia of the model have default values. This cause the the arm has a lot of inertia and the controller is very slow and underdamped. The origin of the oscilations is the gravity, acting as a perturbation on the system. Other efect is a very large position error. When the arm stops, the hand is very down the goal state. You can see it in rviz.
I want to develope a best modell next week, but now you can use the following;
(This values are tested for hydro-dev branch of ros-industrial/universal_robot package.) In order to improve the behaviour, you should increase the gain of the controller, reduce the inertia of the links and increase the damping in the joins. You can change all p gain from 1.000 to 10.000 in the file,
ur_gazebo/controller/arm_controller_ur5.yam
shoulder_pan_joint: {p: 10000.0, i: 0.0, d: 0.0, i_clamp: 0.0, i_clamp_min: 0.0, i_clamp_max: 0.0}
shoulder_lift_joint: {p: 10000.0, i: 0.0, d: 0.0, i_clamp: 0.0, i_clamp_min: 0.0, i_clamp_max: 0.0}
elbow_joint: {p: 10000.0, i: 0.0, d: 0.0, i_clamp: 0.0, i_clamp_min: 0.0, i_clamp_max: 0.0}
Reduce inertia in all links to inertia to ixx="0.002" ixy="0.0" ixz="0.0" iyy="0.02" iyz="0.0" izz="0.002" in files
ur_description/ur5_joint_limited.urdf.xacro and ur_description/ur5_joint_limited.urdf.xacro
<inertia ixx="0.002" ixy="0.0" ixz="0.0" iyy="0.02" iyz="0.0" izz="0.002"/>
And increase the damping in all joints to 0.5 in files
ur_description/ur5_joint_limited.urdf.xacro and ur_description/ur5_joint_limited.urdf.xacro
<dynamics damping="0.5" friction="0.0"/>
Update: I tried the suggested settings. The wobbling appears to be much improved but still looks like it is vibrating. Execution is much faster and smoother. Wobbling is less violent - much better damped. This will prove to be much more usable! Thank you - this was helpful and informative.
Another problem, is the arm apears in rviz in strange positions. Maybe is caused because you use demo.launch and a gazebo launch. First launch fake_controller . In order to avoid this situation, your launch file should include <arg name="fake_execution" value="false">
Check /joint_states topic is only published by gazebo using: rostopic info /joint_states If there are two publishers, and you use demo.launch, then comment or remove <node name="joint_state_publisher" pkg="joint_state_publisher" type="joint_state_publisher"> <rosparam param="/source_list">[/move_group/fake_controller_joint_states]</rosparam> </node> Enjoy a rviz graphics without weird positions of arm.
Again THANK YOU! I get it now... Yes - this fixed the issue of "ghost state" :) I owe you one... or two. Here's a clip of the corrected/improved configuration: http://youtu.be/k5rzlfS1RvA
My UR5 is also wobbling despite the updated gains and parameters. I am using hydro under precise (12.04LTS).
I have cloned repo universal_robot from ROS-I and am on branch hydro-devel (current head: f39b2e75 and I also tested 8a15a0bc dating back to 23/05, which is the commit of hydro-devel at the time you wrote the previous messages).
I have modified the following files according to the previous message (with a sed):
As my setup does not work (actually it is a bit better with the updated gains but really bad compared to the results of b3l33 shown in his videos), I have a few questions:
Any idea of what else I could check to make things work as good as in b3l33's videos?
Also would you be kind enough to diff my .yaml and .urdf files here against yours? Just to make sure my edits are right... I have also joined my setup files (how I fix the base link to the world) just in case...
Thanks,
Antoine.
Till end of weekend I will not have acces to my PC. I will chech your repo on monday. ;) jalfonso100
Hola Alfonso, this is a good start. Unfortunately I am away from my machine and for some reason I do not manage to connect to it. I will try again later but I may have to wait until the week end when I come back in order to test. I can already thank you, I'll give more details at the week end :)
I have made a pull request with the new param values for the UR5 (dynamic params and PID gains) in https://github.com/ros-industrial/universal_robot/pull/85. To find the params, did you just play with the values or did you use identified inertias/damping and a know computation method for PID gains?
Asked: 2014-04-24 13:33:57 -0500
Seen: 4,855 times
Last updated: Jul 26 '14
ROS Answers is licensed under Creative Commons Attribution 3.0 Content on this site is licensed under a Creative Commons Attribution Share Alike 3.0 license.
Has the cylindrical post a mass?
Yes. It has a mass of 50. I have since discovered that the manipulator exhibits the same behaviour if fixed directly to the ground as well. Please see the following youtube video for reference: http://youtu.be/RVYIK4C_mgk Any suggestions on what I may do to stabilize the robot's behaviour? Thank you!