Robotics StackExchange | Archived questions

Pose goal is easier to fail than Joint goal

I'm using ROS Kinetic with Ubuntu 16.04 I was trying to run a UR5 (6DOF robot arm) model with gripper in Rviz. I got 2 problems.

1) When using a pose goal instead of a joint goal, it's much easier to fail or get into some complicated path planning which took way longer than joint goal to finish. Although both will end up at same location for end effector. I wonder if there's some configuration file I can play with in the moveit package generated by moveitsetupassistant.

2) When using RVIZ to drag the arm to a desired pose, and then do plan&execute, it will show several solution and a message like this: ParallelPlan::solve(): Solution found by one or more threads in 0.046674 seconds. However, when using the my code in python, there is no such a message.

I attached my code in the attachment. Here is the output of drag in RVIZ directly:

[ INFO] [1563920423.622299456, 214.804000000]: Combined planning and execution request received for MoveGroup action. Forwarding to planning and execution pipeline.
 [ INFO] [1563920423.622413597, 214.804000000]: Planning attempt 1 of at most 1
 [ INFO] [1563920423.623414442, 214.805000000]: Planner configuration 'manipulator[RRTConnectkConfigDefault]' will use planner 'geometric::RRTConnect'. Additional configuration parameters will be set when the planner is constructed.
[ INFO] [1563920423.624040259, 214.805000000]: manipulator[RRTConnectkConfigDefault]: Starting planning with 1 states already in datastructure
[ INFO] [1563920423.656865860, 214.833000000]: manipulator[RRTConnectkConfigDefault]: Starting planning with 1 states already in datastructure
[ INFO] [1563920423.660062326, 214.835000000]: manipulator[RRTConnectkConfigDefault]: Starting planning with 1 states already in datastructure
[ INFO] [1563920423.671390394, 214.846000000]: manipulator[RRTConnectkConfigDefault]: Created 5 states (2 start + 3 goal)
[ INFO] [1563920423.675991732, 214.848000000]: manipulator[RRTConnectkConfigDefault]: Created 4 states (2 start + 2 goal)
[ INFO] [1563920423.697523406, 214.868000000]: manipulator[RRTConnectkConfigDefault]: Created 6 states (3 start + 3 goal)
[ INFO] [1563920423.702737269, 214.873000000]: manipulator[RRTConnectkConfigDefault]: Created 9 states (3 start + 6 goal)
[ INFO] [1563920423.703138043, 214.874000000]: ParallelPlan::solve(): Solution found by one or more threads in 0.046674 seconds
[ INFO] [1563920423.791421658, 214.952000000]: SimpleSetup: Path simplification took 0.066072 seconds and changed from 3 to 2 states
[ INFO] [1563920426.701524712, 217.784000000]: Completed trajectory execution with status SUCCEEDED ...

However, this is the output if I execute my code:

[ INFO] [1563920286.661937495, 82.204000000]: Combined planning and execution request received for MoveGroup action. Forwarding to planning and execution pipeline.
[ INFO] [1563920286.662151254, 82.204000000]: Planning attempt 1 of at most 1
[ INFO] [1563920286.663111198, 82.205000000]: Planner configuration 'manipulator' will use planner 'geometric::RRTConnect'. Additional configuration parameters will be set when the planner is constructed.
[ INFO] [1563920286.663499463, 82.205000000]: RRTConnect: Starting planning with 1 states already in datastructure
[ INFO] [1563920286.709997590, 82.251000000]: RRTConnect: Created 6 states (2 start + 4 goal)
[ INFO] [1563920286.710046071, 82.251000000]: Solution found in 0.046770 seconds
[ INFO] [1563920286.762445162, 82.301000000]: SimpleSetup: Path simplification took 0.052233 seconds and changed from 3 to 2 states
[ INFO] [1563920289.678544228, 85.144000000]: Completed trajectory execution with status SUCCEEDED ...

I used the code from this tutorial: http://docs.ros.org/kinetic/api/moveit_tutorials/html/doc/move_group_python_interface/move_group_python_interface_tutorial.html

I only did some modification for UR robot, didn't change configuration part in the code.

Asked by NormanYu on 2019-07-23 17:51:46 UTC

Comments

Answers

1) There are a lot of settings and planners you can play with out of the box in MoveIt. Is it really more difficult or less reliable to plan though? And is that true for all poses or is it just your impression? Systematic benchmarks to compare these issues and overall MoveIt performance are an open code sprint topic. You can apply for a GSoC grant if you are interested in those projects.

2) You did not post your code (there is no attachment), but I assume there is a setting for multi-threaded planning that you have not touched. You can check here what the Rviz plugin is doing to produce that terminal output.

Asked by fvd on 2020-03-03 00:16:35 UTC

Comments