- 1 Summary
- 2 Installation/Prerequisites
- 3 Overview
- 4 Video
- 5 Tutorial
- 6 Kinect
- 7 IKFast
- 8 Interfaces
- 9 Related Examples/Tutorials
- 10 Troubleshooting
This tutorial describes how to use the Baxter Research Robot with MoveIt! the standard ROS motion planning framework.
This tutorial requires a ROS distribution >= Groovy
Check out the Baxter MoveIt configuration package into your ROS workspace (follow Catkin or rosbuild tab instructions accordingly)]
$ git clone https://github.com/ros-planning/moveit_robots.git
The follow debian packages are required for using MoveIt! with Baxter
$ sudo apt-get update
= Install MoveIt! =
$ sudo apt-get install ros-groovy-moveit-full
catkin_make to make the new additions to your ROS workspace
MoveIt! motion planning framework provides capabilities including Kinematics (IK, FK, Jacobian), Motion Planning (OMPL, SBPL, CHOMP) integrated as MoveIt! plugins, Environment Representation (robot representation, environment representation, collision checking, contraint evaluation), execution using move_groups, benchmarking, warehouse database for storage (scenes, robot states, motion plans), a C++/Python API and more!
This tutorial will focus on the MoveIt! Rviz plugin as an introduction to some of the capabilities of MoveIt!
Baxter Planning Groups
Describes the joints considered during motion planning. These are specified in the SRDF SRDF for Baxter. Baxter's SRDF includes planning groups, additional collision checking information, and default configurations.
You can view the SRDF at any time:
Verify that the robot is enabled from an RSDK terminal session, ex:
Start the joint trajectory controller, ex:
Note:* Do not launch
baxter_moveit_config demo_sim.launch with the ROS_MASTER_URI set to the robot (ie. using the robot's ros master) as this starts a new joint_state_publisher/robot_state_publisher causing the robot's internal joint_state_publisher/robot_state_publisher be preempted, essentially ceasing Baxter's ability to move the arms until reboot. *Please use
baxter_moveit_config demo_baxter.launch to run this demonstration on Baxter.
The Rviz gui will then open showing Baxter with interactive markers:
Executing a simple motion plan
You will see the goal state for the motion planning, shown for each plan group in Orange.
You can then move Baxter's arms to a planning goal by clicking and dragging the arrows (Cartesian) and rings (orientation) of the interactive markers. Kinematic/collision checking is happening simultaneously and will not allow self collisions (links will turn red), or goal states out of the reachable workspace (inability to drag and endpoint/gripper will turn red).
With a valid goal state, and assuming that the robot is in a valid start state, you are now ready to plan and execute a trajectory from your start state to the goal state.
In the bottom left you will the motion planning frame with a tabbed interface to access different tools of the MoveIt! Rviz visualizer.
Please select the Planning tab in this frame.
Click the Plan_ tab under the Commands field to plan a trajectory from your start state to the goal. You will see this solution executed in Rviz. When you are happy with the planned trajectory, click _Execute to run this trajectory on the robot.
Alternatively, if you would like to immediately execute the first plan from your current position to the goal position use the Plan and Execute button.
Another useful field in the Planning_ tab is Query. You can specify here _Select Start State_ choosing
<same as goal>, choosing these options with the _Update_ button. Also, You may _Select Goal State by clicking that button. Here again you can choose
<same as goal> for the Goal State.
Note: It is dangerous to plan from a start state other than your current joint positions. Please update the Select Start State_ option under the Query field to reflect your current position before planning. This can be done by clicking the _Select Start State:_ text, from the drop down menu select
<current>, click _Update to complete this step. You will now be planning from your current position as the start state.
Introducing an environment representation for planning
Select the Scene Object tab from the Motion Planning frame.
We will now create a scene object in a text file to be imported into our environment.
$ mkdir baxter_scenes
$ gedit baxter_scenes/baxter_pillar.scene
Copy in the following scene describing a pillar
0.2 0.2 1
0.6 0.15 0
0 0 0 1
0 0 0 0
Save and exit.
You can now import this scene from the Scene Geometry field selecting Import From Text
Navigating to select
After opening this scene you will now see the pillar inserted into your environment.
Important: You must publish your current scene so that MoveIt! knows that it will need to plan around your modified environment. Do so by selecting the Context_ tab from the Motion Planning frame. Under this tab you must click the _Publish Current Scene Button under the Planning Library field.
Similar to our previous planning we can now drag our interactive markers moving the goal state to a location on the opposite side of the pillar.
Selecting the Planning tab in the motion planning frame you can now plan a new trajectory that will avoid collision with your environment object (the pillar). The shell in which you launched demo_baxter.launch will provide information regarding which planner will be used, how long it took to find a solution, path simplification/smoothing time, and more. This will also display if your planner was unsuccessful in finding an allowable solution. This is often caused by collision with the environment during testing of the execution or invalid start/goal states. In very constrained or difficult motions, you may have to plan several times to find an acceptable solution for execution.
Upon execution the robot will avoid this 'virtual' object tracking the commanded trajectory.
That's it, you have successfully commanded Baxter using MoveIt!
There is much more information, tutorials, API documentation and more on moveit.ros.org.
It's possible to integrate sensor data from the kinect with MoveIt! so that you can plan paths for Baxter in dynamic environments. This is new functionality, so you should first update to the latest MoveIt configuration package. At this time, there are no drivers available to integrate either linux or ROS with the Xbox One Kinect, or the K4W2, so these instructions are for the original Xbox kinect, or the K4W. To get the kinect working, we first need to install the OpenNI ros packages
In an RSDK shell, launch the openni drivers to check that they're working correctly.
In a separate RSDK shell, you can use image_view to look at the data
You can also just look at the rgb image
Integrating with MoveIt!
You'll probably first want to specify where you've placed your kinect relative to Baxter. We set up a static transform publisher in our MoveIt! launch files to do this for us. By default, the camera is set to position:
# x y z yaw pitch roll parent_frame # 1 0 0 0 0 0 /torso
If you're fine with this location for now, you can leave it as is. If not, you can edit the file
where you can update the location of your kinect by changing the arguments at
<node pkg="tf" type="static_transform_publisher" name="camera_link_broadcaster" args="1.0 0.0 0.0 0.0 0.0 0.0 /torso /camera_link 100" />
More information on this transform publisher can be found on ROS's wiki
After doing this, you can run MoveIt! using input from the kinect. You should not be running your own OpenNI server anymore, but you should have the joint trajectory action server running in an RSDK shell. In another RSDK shell, run the kinect demo:
Alternatively, you could run the regular MoveIt! demo and pass in a kinect argument
After launching rviz, you should be able to see the input data from your kinect sensor in the environment. Self-filtering should be performed for you to ensure that the kinect doesn't consider Baxter to be part of the planning environment. You can now do motion planning for Baxter using the kinect data.
The floating set of axes in the images shows the camera location
If you want to use the kinect data with our MoveIt! simulator, that's also possible. Be careful not to run the simulator in an RSDK shell, or with the ROS_MASTER_URI set to the robot
There are a couple of settings that you may want to adjust for your use case, which can be done through the kinect yaml file
Where you'll see the following information
sensors: - sensor_plugin: occupancy_map_monitor/PointCloudOctomapUpdater point_cloud_topic: /move_group/camera/depth/points max_range: 5.0 padding_offset: 0 padding_scale: 3.0 frame_subsample: 1 point_subsample: 1
Information about these settings can be found on the MoveIt! wiki
#ikfast IKFast provides analytical closed-form solutions for manipulator inverse kinematics. Part of the OpenRAVE robot architecture/software framework, IKFast has been used to provide closed-form solutions for Baxter's inverse kinematics.
Advantages include order of magnitude faster IK solutions and the ability explore the manipulator's null space!
Using Baxter's MoveIt! IKFast plugins
In the baxter_moveit_config package - modify the kinematics.yaml file to specify the use of the IKFast plugins.
You can now uncomment the use of the IKFastKinematicsPlugin, commenting out the default KDL solver.
The result should look like so:
left_arm: # kinematics_solver: kdl_kinematics_plugin/KDLKinematicsPlugin # uncomment to use ikfast kinematics plugin kinematics_solver: baxter_left_arm_kinematics/IKFastKinematicsPlugin kinematics_solver_attempts: 3 kinematics_solver_search_resolution: 0.005 kinematics_solver_timeout: 0.005 right_arm: # kinematics_solver: kdl_kinematics_plugin/KDLKinematicsPlugin # uncomment to use ikfast kinematics plugin kinematics_solver: baxter_right_arm_kinematics/IKFastKinematicsPlugin kinematics_solver_attempts: 3 kinematics_solver_search_resolution: 0.005 kinematics_solver_timeout: 0.005
You will now need to recompile and install the IKFast plugins.
catkin_make install --pkg baxter_ikfast_right_arm_plugin
catkin_make install --pkg baxter_ikfast_left_arm_plugin
Note: It is known that with 0.6 software the SDK will prevent users from executing: ``
catkin_make install`` across the entire workspace. This will be addressed with the upcoming release.
That's it! MoveIt! will now be using IKFast as it's kinematics solver!
Please try it out following the above tutorial.
See the API Reference page for details.
- Joint Trajectory Action Server - /robot/limb/right/follow_joint_trajectory [control_msgs/FollowJointTrajectoryAction]
- Joint Trajectory Action Server - /robot/limb/left/follow_joint_trajectory [control_msgs/FollowJointTrajectoryAction]
- JointTrajectoryActionServer class:
The arm is not executing the trajectory
Verify that the robot is enabled:
rosrun tools enable_robot.py -e
Verify that the trajectory controller has been started:
rosrun baxter_interface trajectory_controller.py
Arm not executing plan; "Unable to identify any set of controllers"
Pressed plan, solution was found and a trajectory execution service request was recieved after which I got the error:
Unable to identify any set of controllers that can actuate the specified joints: [left_e0 left_e1 left_s0 left_s1 left_w0 left_w1 left_w2 right_e0 right_e1 right_s0 right_s1 right_w0 right_w1 right_w2 ]
After which nothing occurred.
If you see the following warning error in the terminal output when roslaunch-ing demo_baxter.launch:
[FATAL] [1372432489.342541284]: Parameter '~moveit_controller_manager' not specified. This is needed to identify the plugin to use for interacting with controllers. No paths can be executed.
This indicates you need to make sure you have the appropriate controller manager plugins installed. Make sure to follow the Baxter MoveIt! Installation instructions and install the required debian packages for both moveit-full and any required plugins. (For the current version, yes - you do need to install
ros-groovy-pr2-moveit-plugins, even though it is not a pr2 - these plugins include default versions which enable the interface to the controllers.)