Skip to content

davetcoleman/baxter_cpp

Repository files navigation

Baxter_cpp

A C++ version of the Baxter SDK that works along side the Rethink SDK. Is inteded to be used with a ros_control implementation of the Baxter controllers that is run on the robot (via SSH).

Contains pick and place MoveIt! code for Baxter, as well as some other tools.

On going development continues in the development branch and contributors are strongly encouraged to send pull requests and use this code. Seriously, please help me improve this code. Attempts are made to keep the indigo-devel branch stable, but at the end of the day I am an open source graduate student contributor ;-)

Features

  • Actuated fingers using a custom robot state publisher and URDF
  • Baxter ros_control integration on Baxter's internal PC using baxter_ssh
  • Baxter pick and place with MoveIt!
    • Generate grasps for blocks on a table
    • Execute a pick and place routine
    • Works on hardware and in an Rviz visualization
    • Other tools for testing trajectories
  • Integrated Asus Xtion Pro depth sensor (Kinect sensor)
    • Displays in MoveIt!

Prerequisites

  • A Baxter with dual parallel electric grippers with SDK v1.0.0 installed
  • (Optional) Asus Xtion Pro Camera
  • ROS Indigo on Ubuntu 14.04

Baxter Installation

  • Install wstool package

    sudo apt-get install python-wstool
    
  • Create a catkin workspace if you don't already have one (we recommend a separate one for Baxter)

    mkdir -p ~/ros/ws_baxter/src
    cd ~/ros/ws_baxter/src
    wstool init .
    
  • Install Rethink's Baxter SDK as documented below, if you have not already. Note: These instructions can also be found at Installing the Research SDK

    wstool merge https://raw.githubusercontent.com/RethinkRobotics/baxter/master/baxter_sdk.rosinstall
    
  • Install these baxter_cpp packages: Note: replaces Rethink's baxter_common with a version that has parallel electric end effectors modeled

    wstool merge --merge-replace -y https://raw.github.com/davetcoleman/baxter_cpp/indigo-devel/baxter_cpp/baxter.rosinstall
    
  • Install ros_control and other low-level components internall on Baxter using the new SSH access:

    Follow instructions on README of baxter_ssh repo

  • Download the Baxter packages:

    wstool update
    
  • Setup ROS if you haven't already (you can probably skip this):

    sudo apt-get update
    sudo apt-get dist-upgrade
    source /opt/ros/indigo/setup.bash
    rosdep update
    
  • Install dependencies and build

    cd ..
    rosdep install --from-paths . --ignore-src --rosdistro indigo -y
    catkin_make
    
  • Add Baxter setup.bash to your .bashrc (recommended)

    echo 'source ~/ros/ws_baxter/devel/setup.bash' >> ~/.bashrc
    

Customize for your robot

Every Baxter is factory calibrated for the mouting points of the arms because they are welded on. Therefore, you might want to customize the baxter_description/urdf/baxter.urdf file to your robot's custom values. To do so:

  • Start up your Baxter without launching any ROS nodes on your dev machine
  • While connected to Baxter, run the command:
    rosparam get -p /robot_description | tail -n +2 > my.baxter.urdf
    
  • Within my.baxter.urdf find the lines that say:
    <joint name="left_torso_arm_mount" type="fixed">
    
    and
    <joint name="right_torso_arm_mount" type="fixed">
    
  • Copy the following <origin> line to the corresponding location in baxter.urdf located in baxter_description/urdf/baxter.urdf

Bringup Baxter

Hardware

  • Ensure you have the correct ROS_MASTER_URI exported, this depends on your robot serial number. CU Boulder's is:

    export ROS_MASTER_URI=http://011305P0009.local:11311
    

    You might also need to set the ROS hostname environment variable if you have not already done so and you have communication issues:

    export ROS_HOSTNAME=128.138.244.72  # REPLACE WITH YOUR COMPUTER'S IP ADDRESS
    
  • Bringup ros_control controllers on Baxter

    Follow instructions on baxter_ssh

Rviz Visualization

This only shows a virtual Baxter in Rviz without any physics simulator. Instead it uses a ros_control hardware interface that simply loops back to itself. Good for testing MoveIt!.

  • Ensure you have the correct ROS_MASTER_URI exported:

    export ROS_MASTER_URI=http://localhost:11311
    
  • Start visualization:

    roslaunch baxter_control baxter_visualization.launch
    

    Note: you will not see anything when you launch this, only the /robot/joint_states topic starts being published. Proceed to the MoveIt! section, below.

Gazebo Simulation

This is still in beta. First install the baxter_gazebo stuff. I haven't documented this yet.

roslaunch baxter_gazebo baxter_world.launch

Start MoveIt!

Works with simulation, hardware or visualization:

  • Start MoveIt!'s move_group Server:

    roslaunch baxter_moveit_config baxter_moveit.launch
    
  • Start Rviz with MoveIt! configured:

    roslaunch baxter_moveit_config moveit_rviz.launch
    

Block pick and place demo

Picks small blocks located on a table in front of Baxter and places them to Baxter's left. Assumes perfect perception (doesn't have perception) as defined in custom_environment.h.

roslaunch baxter_pick_place block_pick_place.launch

Hardware Control Modes

Follow instructions on README of baxter_ssh repo

Test Programs

Send Baxter to random poses using motion planning and obstacle avoidance of a hard-coded planning scene

roslaunch baxter_pick_place random_planning.launch

Programmed Buttons

Duplicated from README of baxter_ssh

End Effector Cuff

  • Rectangular button: close end effector
  • Circular button: open end effector

Back of Robot Shoulders

  • Left shoulder button: enable Baxter
  • Right shoulder button: disable Baxter

License

BSD (New BSD License)

Contributors

  • @davetcoleman
  • @brawner
  • @jon-weisz

Please help - see Contribute page.

About

Additional Baxter packages for MoveIt, written entirely in C++

Resources

Stars

Watchers

Forks

Packages

No packages published

Languages