Kinematic singularities occur when the Jacobian matrix of a 6-axis articulated robot loses full rank (det(J) = 0), collapsing a degree of freedom and causing inverse kinematics algorithms to demand infinite angular joint velocity. This technical paper presents practical singularity detection and damped least squares (DLS) avoidance algorithms.
1. The Three Classical 6-Axis Robotic Singularities
Wrist, Shoulder, and Elbow alignment geometries
In industrial manipulators, inverse kinematics maps Cartesian end-effector velocities (v, ω) to joint angular velocities (q̇) via the manipulator Jacobian J(q). When det(J(q)) approaches zero, computing q̇ = J⁻¹ · ẋ demands near-infinite joint velocity, triggering servo overcurrent trips or violent physical vibrations.
The three fundamental singular configurations are:
1. Wrist Singularity: Occurs when Joint 4 and Joint 6 axes become collinear (Joint 5 angle = 0°). The robot loses the ability to rotate about the normal to the wrist plane without spinning Joints 4 and 6 at extreme counter-opposed speeds.
2. Shoulder Singularity: Occurs when the wrist center point falls on the vertical centerline of Joint 1. The robot cannot translate sideways in Cartesian space without an instantaneous 180° swing of the heavy base axis.
3. Elbow Singularity: Occurs when the arm is fully extended or fully folded back (wrist center aligns with the upper and lower arm plane). The robot loses radial extension capability.
Axes 4 and 6 align collinearly. Controller cannot resolve planar rotation without infinite angular joint velocity. Solution: Apply Damped Least Squares (DLS).
Wrist center aligns directly with vertical centerline of Axis 1. Sideways translation in Cartesian space demands 180° instant base spin.
Arm reaches maximum radial extension or folds completely flat against upper link. The arm loses 1 translational degree of freedom.
Where ẋ is Cartesian 6-DoF velocity vector and q̇ is 6-axis joint velocity vector.
Where λ is dynamic damping factor. As det(J) -> 0, λ increases smoothly, trading minor trajectory tracking deviation for finite bounded joint velocities.
During offline robot cell programming (e.g., in laser cutting or welding paths), tilt the workpiece or tool mounting angle by 5° to 10° relative to the base axis. This simple engineering adjustment keeps Joint 5 away from 0°, eliminating wrist singularity crossings entirely.
2. Real-Time Dynamic Singular Region Blending
Transitioning smoothly from Cartesian to joint-space interpolation
INDUS robot controllers continuously evaluate the manipulability index w = √(det(J · Jᵀ)). When w drops below a calibrated threshold (w_crit), the trajectory generator automatically engages Damped Least Squares (DLS) or smoothly blends the motion from Cartesian linear (MOVL) to joint interpolation (MOVJ).
This ensures automated continuous welding torches and glue dispensing heads maintain smooth flow without stuttering or tripping servo drive overspeed safety interlocks.
Need Custom Calculations for Your Machine?
Our technical team provides kinematic sizing, finite element torsional analysis, and fieldbus latency audits for your exact payload and cycle parameters.