Skip to content

Latest commit

 

History

History
62 lines (45 loc) · 6.41 KB

USAGE.md

File metadata and controls

62 lines (45 loc) · 6.41 KB
Raw

pick_ik : Usage

Using pick_ik as a Kinematics Plugin

As discussed in the MoveIt 2 documentation, you can use the MoveIt Setup Assistant or change the kinematics.yaml file for your robot setup to use pick_ik as the IK solver.

An example kinematics.yaml file might look as follows:

panda_arm:
  kinematics_solver: pick_ik/PickIkPlugin
  kinematics_solver_timeout: 0.05
  kinematics_solver_attempts: 3
  mode: global
  stop_optimization_on_valid_solution: true
  position_scale: 1.0
  rotation_scale: 0.5
  position_threshold: 0.001
  orientation_threshold: 0.01
  cost_threshold: 0.001
  minimal_displacement_weight: 0.0
  gd_step_size: 0.0001

As a sanity check, you could follow the MoveIt Quickstart in RViz tutorial and change the moveit_resources/panda_moveit_config/config/kinematics.yaml file to use a configuration like the one above.

Parameter Description

For an exhaustive list of parameters, refer to the parameters YAML file.

Some key parameters you may want to start with are:

  • mode: If you choose local, this solver will only do local gradient descent; if you choose global, it will also enable the evolutionary algorithm. Using the global solver will be less performant, but if you're having trouble getting out of local minima, this could help you. We recommend using local for things like relative motion / Cartesian interpolation / endpoint jogging, and global if you need to solve for goals with a far-away initial conditions.
  • stop_optimization_on_valid_solution: The default mode of pick_ik is to give you the first valid solution (which satisfies all thresholds) to make IK calls quick. Set this parameter to true if you rather want to use your complete computational budget (based on kinematics_solver_timeout and the maximum number of iterations of the solvers) to try to find a solution with a low cost value.
  • memetic_<property>: All the properties that only kick in if you use the global solver. The key one is memetic_num_threads, as we have enabled the evolutionary algorithm to solve on multiple threads.
  • cost_threshold: This solver works by setting up cost functions based on how far away your pose is, how much your joints move relative to the initial guess, and custom cost functions you can add. Optimization succeeds only if the cost is less than cost_threshold. Note that if you're adding custom cost functions, you may want to set this threshold fairly high and rely on position_threshold and orientation_threshold to be your deciding factors, whereas this is more of a guideline.
  • position_threshold/orientation_threshold: Optimization succeeds only if the pose difference is less than these thresholds in meters and radians respectively. A position_threshold of 0.001 would mean a 1 mm accuracy and an orientation_threshold of 0.01 would mean a 0.01 radian accuracy.
  • approximate_solution_position_threshold/approximate_solution_orientation_threshold: When using approximate IK solutions for applications such as endpoint servoing, pick_ik may sometimes return solutions that are significantly far from the goal frame. To prevent issues with such jumps in solutions, these parameters define maximum translational and rotation displacement. We recommend setting this to values around a few centimeters and a few degrees for most applications.
  • position_scale: If you want rotation-only IK, set this to 0.0. If you want to solve for a custom IKCostFn (which you provide in your setFromIK() call) set this and rotation_scale to 0.0. You can also use any value other value to weight the position goal; it's part of the cost function. Note that any checks using position_threshold will be ignored if you use position_scale = 0.0.
  • rotation_scale: If you want position-only IK, set this to 0.0. If you want to treat position and orientation equally, set this to 1.0. You can also use any value in between; it's part of the cost function. Note that any checks using orientation_threshold will be ignored if you use rotation_scale = 0.0.
  • minimal_displacement_weight: This is one of the standard cost functions that checks for the joint angle difference between the initial guess and the solution. If you're solving for far-away goals, leave it to zero or it will hike up your cost function for no reason. Have this to a small non-zero value (e.g., 0.001) if you're doing things like Cartesian interpolation along a path, or endpoint jogging for servoing.

You can test out this solver live in RViz, as this plugin uses the generate_parameter_library package to respond to parameter changes at every solve. This means that you can change values on the fly using the ROS 2 command-line interface, e.g.,

ros2 param set /rviz2 robot_description_kinematics.panda_arm.mode global

ros2 param set /rviz2 robot_description_kinematics.panda_arm.minimal_displacement_weight 0.001

Custom Cost Functions

The kinematics plugin allows you to pass in an additional argument of type IkCostFn, which can be passed in from common entrypoints such as RobotState::setFromIK(). See this page for a usage example. Keep in mind that you need to set position_scale = 0.0 and rotation_scale = 0.0 if you want to calculate the costs solely based on your IkCostFn. If these parameters are non-zero, the target pose will still be part of the overall cost function. Additionally, you might want to define values for the cost_threshold, approximate_solution_cost_threshold, and stop_optimization_on_valid_solution parameters to decide when pick_ik will stop optimizing for your cost function and which solutions it should accept.

Alternatively, consider adding your own cost functions to the pick_ik source code (specifically, in goal.hpp and goal.cpp) and submit a pull request with the new functionality you add.