Object grasping pipeline

Object grasping pipeline based on visual-servoing - 'kinova_vs_manipulation' package

How to run

• For MoveIt! visualization, run this (your local terminal must have correct ROS_HOSTNAME and ROS_IP, set by running 'export_mbot7'):

    moveit_kinova_rviz
  

• Launch the pipeline. This launches MoveIt, the object localizer, and makes a service available to trigger an object grasp:

    roslaunch kinova_vs_manipulation object-grasp.launch
  

• Grasp an object (using service call):

    rosservice call /kinova_vs_manipulation/object_grasp_sm "object_class: 'orange'
    reposition: false
    target_position:
      x: 0.0
      y: 0.0
   theta: 0.0"
  

• Grasp an object (using mbot_class):

    mbot_class
  
    # Grasp - lift and lower object:
    > mbot.kinova_manipulation.grasp_object('orange') 
  
    # Grasp and move the object to target position relative to table origin:
    > mbot.kinova_manipulation.grasp_object('cup', target_table_position=(0.3, 0.2))
  

• To grasp using smach, the kinova_manipulation_states module in the mbot_states package has useful smach states. For an example of their use, check the object_grasp_sm_service.py node in the kinova_vs_manipulation package. For usage in a competition benchmark, check the fbm6 package in mbot_competitions.

How to configure

Main ROS parameters are in the kinova_vs_manipulation/launch/object-grasp.launch file.

Details

The code for the ROS nodes is documented with top-of-file comments.

Main nodes:

• pregrasp_service advertises a service that selects a pregrasp pose relative to the target object, and moves the arm to it, making it visible by the camera
• visual_servoing advertises a service to start approaching the arm to the object based on camera-frame error
• joint_control reads the end-effector velocity topic published by visual_servoing, calculates corresponding joint velocities using inverse Jacobian, and publishes them to the arm
• object_grasp_sm_service advertises a service to perform the complete grasping pipeline, detecting the object and calling the services above