High-Precision RCM Motion for Ophthalmic Robotics via Hierarchical QP Control and Authority-Guided Recovery Planning
General-purpose manipulators can position ophthalmic instruments under a software Remote Centre of Motion (RCM). However, at some reachable configurations that are nonsingular for the tip task alone, the RCM and tip-tracking objectives conflict and the servo oscillates. We formalise this failure as a constraint-induced task singularity: the RCM-projected tip Jacobian loses rank while the unprojected tip Jacobian stays well conditioned. Thus, the effective tip authority collapses although the target remains reachable. We present a high-precision RCM servo architecture with built-in singularity recovery: for each commanded target, an authority-guided planner selects a well-conditioned arm configuration that reaches it and plans an RCM-consistent route, executed by a fixed hierarchical-QP controller. On a 200-scenario kinematic benchmark, the complete nominal-and-recovery protocol reaches 128/128 near-singular targets and 71/72 near-joint-limit targets (100% and 99%, respectively), with the software-RCM residual below 0.142 mm. On a physical 6-DOF arm it executes across four RCM entry points, with a worst-case kinematically reconstructed RCM residual of 178 μm and a median tip-regulation error of 32μm, within the 200μm control requirement adopted for the ophthalmic task.