Database-Driven Capability Maps for Pareto-Based Configuration Selection in Tendon-Driven Continuum Robots

Abstract

Tendon-Driven Continuum Robots (TDCRs) are increasingly used in minimally invasive surgery and cluttered environments because of their compliance and dexterity. However, dense workspace characterization and configuration selection remain challenging because inverse kinematics must be solved in a highly redundant, nonlinear space under tendon-actuation constraints. Existing capability map frameworks do not directly address configuration-level tendon feasibility or dexterity-aware selection among feasible alternatives. This paper presents a database-driven CAP framework that combines piecewise constant curvature kinematics with particle swarm optimization to construct an offline library of configurations consistent with the prototype tendon-routing and command-mapping model over 180 workspace voxels and 54,000 voxel-orientation samples. For each feasible sample, pose residuals and Jacobian-based dexterity descriptors are computed offline and stored in the database. At runtime, the method retrieves feasible entries for a queried voxel and ranks them through a Pareto-based trade-off between combined pose residual and inverse manipulability. The output is a dexterity-aware representative configuration for the queried local neighbourhood, rather than a newly optimized exact-pose solution. Motion-capture validation on a custom two-segment TDCR shows millimetre-order position errors in the evaluated ROI.

Keywords: Tendon-driven continuum robot, Capability map, Voxel-level configuration selection, Piecewise constant curvature, Particle swarm optimization, Manipulability