Preview-Based Relative-Motion Control of an Insertion Tool for Neural-Thread Placement in Pulsating Tissue

arXiv:2608.08860v1 Announce Type: cross Abstract: Robotic neural-thread placement requires regulating the insertion-tool tip relative to tissue that moves with cardiac and respiratory pulsation. This paper develops a preview-based relative-motion controller that estimates latency-delayed periodic surface motion, predicts it over a short horizon, and uses offset-free model predictive control to regulate relative placement while limiting actuator effort and lateral relative velocity. In MuJoCo, the 1-DOF controller achieves 12.0m free-space and 1.9m contact RMS relative-placement error, versus 18.3/176.8m for delayed-feedback impedance and 286.1/275.5m for lab-frame PD, at the cost of higher peak contact force (3.43 versus 2.00mN) since offset-free tracking drives the tip fully to the commanded depth rather than yielding against the tissue. In 3 DOF, coupled preview reduces contact lateral shear from 1.34 to 0.50mm/s with 2.1m lateral RMS error. A feasibility-restored octagonal shear formulation keeps the QP solvable under degraded sensing by adding a bounded shared slack: at 10m RMS per-axis sensing noise, where a matched cost-only controller violates the 0.80mm/s budget in all 10 seeds (mean/maximum 0.988/1.175mm/s), the soft-octagon controller completes all 10 seeds with no fallback and no measured violation (0.653/0.712~mm/s), with its operating envelope characterized up to 15m RMS. A two-vertex Lyapunov certificate for the controller's actual finite-horizon error-feedback gain holds over -40%/{+}50% reflected-mass mismatch. The modeled tip is a rigid contact point, and the study is simulation-only: flexible-thread and carrier-needle mechanics, a validated transient-force constraint, biological damage thresholds, and hardware-realistic sensing and timing remain required before deployment.

Source: arXiv cs.RO | 2026-08-11