Can an 11-DOF redundant robot improve CT-guided needle insertion dexterity?

What the extra degrees of freedom buy

The paper's answer is a workspace and dexterity gain, not a clinical outcome. The previous platform combined a 3-DOF linear base with a 5-DOF end-effector, which confined reachability to configurations in front of the scanner. The new system replaces that base with a 6-DOF commercial arm (Stäubli TX2-60L) and keeps a redesigned 5-DOF cable-driven end-effector, giving 11 joints in total.

The split of labor matters for reading the results. The arm supplies large-scale positioning and reorientation around the bore, while the end-effector supplies local precision for needle alignment and insertion. Redundancy here means the robot has more joints than the task strictly needs, so multiple joint configurations can produce the same needle pose; the controller's job is to choose among them.

Why the needle task is effectively rank five

The controller does not treat the needle pose as a full six-dimensional target. Position contributes three task dimensions and needle-axis direction contributes two, but rotation about the needle's own principal axis is left uncontrolled. The task Jacobian therefore has rank five even though the error vector is written in six dimensions.

That gap is what makes redundancy useful rather than merely awkward. With 11 joints and a rank-5 task, the null space has dimension six, so the robot can pursue secondary goals without disturbing the primary needle pose. The paper uses that freedom to maximize Yoshikawa manipulability, a scalar built from the determinant of the Jacobian times its transpose, as a null-space subgoal.

Weighting joints to separate coarse and fine motion

The weighted inverse kinematics controller penalizes joint motion through a positive-definite diagonal weight matrix, where a larger weight on a joint means a higher penalty for moving it. The authors treat this matrix as a hyperparameter and compare three policies: one emphasizing end-effector motion, one with equal weights, and one emphasizing base motion. The two asymmetric policies apply a factor-of-10 weight difference between the 6-DOF base and the 5-DOF end-effector.

This is the paper's two-stage priority scheme. Large-scale movement into the bore and fine in-bore needle adjustment place different demands on which joints should move, and the weight policy selects between them. The reported control settings were a gain of 0.02 for both the task and null-space terms and a damping factor of 10⁻⁴.

Enforcing joint limits without losing the subgoal

Null-space projection alone does not respect joint limits, so the paper adds a quadratic program at each control cycle. It finds the joint velocity closest to the redundancy-resolved command while enforcing velocity bounds and position bounds over one timestep. The commanded joint update is then the current configuration plus the timestep times that projected velocity.

The authors note that with the small damping used, the null-space projector's residual coupling into the primary task is negligible, so the manipulability subgoal is pursued without materially disturbing needle tracking. This is a design argument from the controller formulation, not a separately measured clinical result.

Reachability gains in simulated anatomies

The headline comparison is a simulation. Human bodies of average, large (+1.5σ), and very large (+3σ) size were generated for both females and males, placed in an 85 cm scanner bore, and sampled with 600 surface normals over the chest plus 36 additional orientations within a 30° adjustment at each normal. Reachability was counted with a 1 mm position tolerance and a 1° orientation tolerance.

Across these scenarios the 11-DOF system exceeded 97% reachability, with full coverage of the area of interest except a few outliers. The 8-DOF system showed clear non-reachable regions concentrated along both sides of the torso, where proximity to the bore constrained the robot, and its reachability degraded as anatomy size increased. Because this is an inverse-kinematics reachability test rather than a tissue or phantom insertion study, it measures geometric access, not targeting accuracy in patients.

Tracking accuracy across five biopsy subtasks

The weight policies were then tested on five representative trajectories drawn from clinical biopsy scenarios, in both simulation and real-world experiments. Tracking was characterized by mean position and orientation errors, plus rise time, settling time, and steady-state error for position. The abstract reports superior tracking accuracy and enhanced manipulability for the task-oriented policy relative to the alternatives.

The paper's framing is that no single weight policy is universally best; the matrix is a tunable hyperparameter that shifts effort between the base and the end-effector. Readers should treat the comparison as evidence that the weighting choice changes behavior in a controllable way, not as a claim that one fixed policy is optimal for every subtask.

What the teleoperation demo does and does not show

A teleoperation test used a Phantom Omni device to supply the target needle-tip pose. The workflow had three stages: moving the end-effector from its setup position outside the bore to inside, adjusting needle position and orientation within the bore, and inserting the needle into the target. The authors present this as a demonstration of the intended workflow, with the arm handling large movements and the end-effector handling fine positioning.

This is a feasibility demonstration of the control architecture, not a clinical trial. The paper's own stated next steps are haptic feedback to guide the operator during insertion, compensation for respiratory motion, and an optimization framework to find the best weight policy for a specific performance requirement. Those gaps define where the current evidence stops.

Where the 97% reachability claim applies

The most decision-relevant number in this paper is the reachability split, and its scope is narrow: simulated generated anatomies, an 85 cm bore, chest-surface targets with orientation perturbations, and inverse-kinematics collision-free solutions. Within that setting, the practical threshold is anatomy size. The 8-DOF design's weakness grew as bodies became larger and in-bore space shrank, while the 11-DOF system held above 97% across the same size range.

So the paper's priority is not a blanket claim that redundancy improves needle biopsy. It is that the base replacement and the added end-effector range mainly buy access at the torso sides and in large patients, exactly the cases where the previous linear-base design failed. The weighted controller and null-space subgoal then determine how that extra access is used during tracking and insertion.

Sources used in this answer