This package provides the folding assembly functionality to the SARAFun project. It controls the relative pose between two rigid objects in contact, along two motion directions: a translational direction and a rotational axis, which are non-collinear. It also provides adaptive estimators that will estimate kinematic uncertainties on the task, such as the available motion directions and contact point location. Besides a dual-armed robot integrated with ROS, the package assumes the existance of at least one force and torque sensor.
- sarafun_msgs (link to dummy package)
- generic_control_toolbox
- abb_irb14000_support
- ROS Indigo
The file 'config_action_node.yaml' must be edited in order to set the reference frames used for the controller this incudes the name of the KDL end-effector frame and the gripping frame name in which the controller quantities are expressed. The controller is launched with
$ roslaunch folding_assembly_controller controller.launch
After launch, and actionlib server will be running which can be called as
$ rosrun actionlib axclient.py /folding_controller/<action_name>
where <action_name> is set in the configuration file. You should see the following GUI:
The init_t_error and init_k_error variables are useful for controller experiments.
goal_force sets the desired contact force along the estimated surface normal.
The controller operates on two different modes: pose control or velocity reference.
In pose control (use_pose_goal set to true), the robot will regulate the two rigid parts to a relative pose along the two motion directions. Alternatively (use_pose_goal set to false), the system will command velocities along the estimated motion directions.
A dynamic reconfigure server is provided for facilitating testing the system.
Notably, it allows setting the degree to which the arms share the manipulation task, through the alpha parameter under the ects group.
