Non-chain inverse kinematics solver for MoveIt!
C++ Other
Switch branches/tags
Nothing to show
Clone or download
Permalink
Failed to load latest commit information.
whole_body_kinematics_plugin
.gitignore
CMakeLists.txt
README.md
package.xml
whole_body_kinematics_plugin_description.xml

README.md

MoveIt! Whole Body IK

Whole body (non-chain) inverse kinematics solver for MoveIt! for any robot. Uses a weighted least-norm solution combined with the gradient projection method of the null space to avoid joint limits. Calculates the psuedo inverse using Lapack's singular value decomposition (SVD) implementation. Uses some componentes from the Kinematic Dynamics Library (KDL) from OROCOS. Build to work with MoveIt!

Developed by Dave Coleman while at the University of Tokyo JSK lab in collaboration with Shunichi Nozawa and Kei Okada.

Video Demos

YouTube

Limitations

  • Does not yet support mimic joints
  • Does not yet support locking joints externally

Setup

kinematics.yaml

Your kinematics.yaml should have each group name listed with the correct solver and the desired tips (end effectors) that you wish to solve for:

whole_body_fixed:
  kinematics_solver: whole_body_kinematics_plugin/WholeBodyKinematicsPlugin
  kinematics_solver_ik_links:
    - LARM_LINK6
    - RARM_LINK6
    - LLEG_LINK5
    - RLEG_LINK5

Optional Parameters

Additionally you can customize the performance with the following parameters:

kinematics_solver_max_solver_iterations:   # iterations for newton raphson method before trying a different seed state
kinematics_solver_timeout:                 # amount of time allowed before giving up
kinematics_solver_verbose:                 # show lots of console output
kinematics_solver_debug_mode:              # similar to verbose but only shows matrix-related math debug output
kinematics_solver_visualize_search:        # publish to rviz every step of the solver
kinematics_solver_epsilon:                 # error tolerance between desired pose and current pose for termination condition
kinematics_solver_null_space_epsilon:      # threshold for when ee pose and solved pose are close enough that we should reduce influence of the null space projection
kinematics_solver_null_space_vel_gain:     # k, the amount the null space calculation affects the overall velocity gain
kinematics_solver_ee_pos_vel_limit:        # maximum allowed input positional velocity of the end effector before limiting, in meters
kinematics_solver_ee_rot_vel_limit:        # maximum allowed input rotational velocity of the end effector before limiting, in meters
kinematics_solver_joint_velocity_max_ratio:# the fraction of a joint's total range that it is allowed to move per iteration

SRDF

  • You must define planning groups for each tip listed in the ''ik_links'' section, above, that goes from BASE_LINK -> TIP_LINK
  • You must also define planning subgroups for any overlapping links. For example:
    • Group 1: BASE -> TORSO -> ARM1 -> HAND1 group (has overlapping links with ARM2)
    • Group 2: TORSO -> HAND1 group (no overlapping links with ARM2)

Assumptions

Note: this code currently assumes your joint model group is in this order:

TORSO ARM ARM LEG LEG

It also assumes your left & right arms each share N number of torso joints, and those are the only shared joints on the whole robot

Assumes all planning groups have their root at the base_link (root link)