Principle:ARISE Initiative Robosuite IK Solver Utilities
| Knowledge Sources | |
|---|---|
| Domains | Robotics, Kinematics, Control Systems |
| Last Updated | 2026-02-15 07:00 GMT |
Overview
A nullspace-aware damped least-squares inverse kinematics solver for robot arms that computes joint velocity commands from Cartesian end-effector targets, supporting multiple end-effector sites, configurable damping, and joint velocity limits.
Description
Inverse kinematics (IK) is the problem of finding joint configurations that place a robot's end-effector at a desired position and orientation. In simulation-based robotics, IK solvers are used both as low-level controllers (converting Cartesian commands to joint velocities in real time) and as planning utilities (computing feasible joint configurations for target poses). A damped least-squares IK solver provides a robust, numerically stable approach that avoids the singularity problems inherent in the pseudoinverse method.
The solver operates on the manipulator Jacobian, which relates joint velocities to end-effector velocities. Given a desired end-effector twist (linear and angular velocity toward the target), the solver computes joint velocities using damped least-squares inversion of the Jacobian. The damping factor acts as a regularizer that prevents excessively large joint velocities near singular configurations, trading off tracking accuracy for numerical stability. Joint velocity limits are enforced by clipping the computed velocities to configurable maximum values.
For robots with more degrees of freedom than end-effector constraints (redundant manipulators), the solver supports nullspace optimization. Per-joint gain weights allow secondary objectives -- such as joint centering or preferred posture maintenance -- to be projected into the nullspace of the primary task. The solver also handles multiple end-effector sites simultaneously (for bimanual robots or robots with additional reference frames), stacking the Jacobians and twists to solve for all sites in a single computation. Input actions can be specified in absolute coordinates, relative to the current pose, or relative to a reference frame such as the mobile base.
Usage
Use this principle when implementing Cartesian-space control for robot arms in simulation. It is appropriate for operational-space controllers, teleoperation interfaces that provide Cartesian targets, and any scenario where the controller must convert desired end-effector motions to joint commands while respecting velocity limits and handling kinematic redundancy.
Theoretical Basis
Damped least-squares IK solves for joint velocities dq given a desired task-space twist:
twist = [v_x, v_y, v_z, w_x, w_y, w_z] (desired end-effector velocity)
J = Jacobian(q) -- 6 x n_joints, from mujoco.mj_jac
dq = J^T (J J^T + lambda^2 I)^{-1} twist
where lambda is the damping factor. This is equivalent to solving the regularized normal equations and avoids singularity issues.
Orientation error is computed via quaternion difference:
q_error = q_target * conjugate(q_current)
w_error = 2 * q_error[1:4] / dt (angular velocity from quaternion error)
Multi-site stacking for N end-effector sites:
J_stacked = | J_1 | twist_stacked = | twist_1 |
| J_2 | | twist_2 |
| ... | | ... |
| J_N | | twist_N |
dq = J_stacked^T (J_stacked J_stacked^T + lambda^2 I)^{-1} twist_stacked
Joint velocity clamping:
dq = clip(dq, -max_dq, max_dq)
q_new = q_current + dq * integration_dt
Nullspace projection for redundancy resolution:
N = I - J^+ J (nullspace projector)
dq_null = N @ (gains * (q_rest - q_current))
dq_total = dq_primary + dq_null
where gains are per-joint nullspace weights from a configuration dictionary.