Skip to content

Latest commit

 

History

History
36 lines (26 loc) · 3.67 KB

Readme.md

File metadata and controls

36 lines (26 loc) · 3.67 KB
Raw

fourr_kazerounian_controller

This repository collects the source files and documentation of the move_group interface used to control the move_group node implementing the 4R planar manipulator from Kazerounian & Wang.

The MoveGroup class exposes a set of default capabilities that can be used for planning and other needs, that are:

  • MoveGroupCartesianPathService: Computing straight line Cartesian paths with collision checking via a ROS service
  • MoveGroupExecuteService: Allow execution of previously computed paths via a ROS service
  • MoveGroupExecuteTrajectoryAction: Allow execution of previously computed trajectories via a ROS action
  • MoveGroupKinematicsService: Compute forward and inverse kinematics via ROS services
  • MoveGroupMoveAction: Compute motion plans via a ROS action
  • MoveGroupPlanService: Compute motion plans via a ROS service
  • MoveGroupQueryPlannersService: Allow querying of available planners (loaded from the motion planning plugin) via a ROS service
  • MoveGroupStateValidationService: Provide a ROS service that allows for testing state validity
  • MoveGroupGetPlanningSceneService: Provide a ROS service that allows for querying the planning scene
  • ApplyPlanningSceneService: Provide a ROS service for blocking updates to the planning scene
  • ClearOctomapService: Provide a ROS service that allows for clearing the octomap

Be aware that the capabilities above are not immediately visible through the MoveGroupInterface implemented in this package, but they are called by the move_group node executing in the scope of the controlled package.

MoveGroupCartesianPathService and MoveGroupMoveAction capabilities are those mainly used for planning in the move_group_interface_tutorial. Both of them are also used in this package for demonstration, together with the MoveGroupDPRedundancyResolutionService developed in this project.

Planning schemes used in this package

  • Planning with globally-optimal redundancy resolution, by using the plugin developed in this project, using dynamic programming.
  • Planning to a pose goal (commented out): a goal is set in the workspace and a plan is asked to OMPL by using the KDL kinematic solver. The planner fails in finding the trajectory as KDL has poor performances with <6 DOF systems. This scheme exploits the MoveGroupMoveAction capability.
  • Planning to a joint space goal: a goal in the joint space is set and a plan is found by using the MoveGroupMoveAction (masked by the plan function in the code).
  • Planning a cartesian path: a (fine) list of waypoints is loaded from the Parameter Server with timing information in order to create a WorkspaceTrajectory. The waypoints describe a circumference in the xy plane. These are passed to the MoveGroupCartesianPathService to compute a finer set of control points that are kinematically inverted to find a continuous trajectory in the joint space. Remember that step resolution must be smaller than the maximum distance between waypoints, otherwise the trajectory cannot be achieved.

Available trajectories

These trajectories can be configured in the package launch file.

  • ws_trajectory_matlab.yaml: this is the trajectory exported from the MATLAB experiments. It is the one used with the optimal redundancy resolution and cannot be used for computing cartesian path through the default MoveGroupCartesianPathService capability because of the overlap between waypoints
  • ws_trajectory.yaml: this is a trajectory with meaningless timestamps, that can be used for computing cartesian paths through the MoveGroupCartesianPathService capability
  • ws_trajectory_yaw_0.yaml: same as above but with yaw=0.