TRAC-IK Kinematics Solver¶
TRAC-IK is an inverse kinematics solver developed by TRACLabs that combines two IK implementations via threading to achieve more reliable solutions than common available open source IK solvers. From their documentation:
(TRAC-IK) provides an alternative Inverse Kinematics solver to the popular inverse Jacobian methods in KDL. Specifically, KDL’s convergence algorithms are based on Newton’s method, which does not work well in the presence of joint limits — common for many robotic platforms. TRAC-IK concurrently runs two IK implementations. One is a simple extension to KDL’s Newton-based convergence algorithm that detects and mitigates local minima due to joint limits by random jumps. The second is an SQP (Sequential Quadratic Programming) nonlinear optimization approach which uses quasi-Newton methods that better handle joint limits. By default, the IK search returns immediately when either of these algorithms converges to an answer. Secondary constraints of distance and manipulability are also provided in order to receive back the “best” IK solution.
The package trac_ik_kinematics_plugin provides a KinematicsBase MoveIt! interface that can replace the default KDL solver. Currently mimic joints are not supported.
As of v1.4.5, trac_ik is part of the ROS Indigo/Jade binaries:
sudo apt-get install ros-melodic-trac-ik-kinematics-plugin
- Install trac_ik_kinematics_plugin and trac_ik_lib package or add to your catkin workspace.
- Find the MoveIt! kinematics.yaml file created for your robot.
- Replace kinematics_solver: kdl_kinematics_plugin/KDLKinematicsPlugin (or similar) with kinematics_solver: trac_ik_kinematics_plugin/TRAC_IKKinematicsPlugin
- Set parameters as desired:
- kinematics_solver_timeout (timeout in seconds, e.g., 0.005) and position_only_ik ARE supported.
- solve_type can be Speed, Distance, Manipulation1, Manipulation2 (see trac_ik_lib documentation for details). Default is Speed.
- kinematics_solver_attempts parameter is unneeded: unlike KDL, TRAC-IK solver already restarts when it gets stuck
- kinematics_solver_search_resolution is not applicable here.
- Note: The Cartesian error distance used to determine a valid solution is 1e-5, as that is what is hard-coded into MoveIt’s KDL plugin.
Open Source Feedback
See something that needs improvement? Please open a pull request on this GitHub page