Real-Time Multi-Contact Estimation for Cable-Driven Serial Manipulator

Abstract

The cable-driven serial manipulator (CDSM) combines the flexibility of continuum robots with the high load-bearing capacity of rigid robotic arms. To ensure safe interaction with the surrounding environment and humans while enabling dexterous operations, tactile perception plays a crucial role as a sensory feedback source. This paper proposes a real-time multi-contact estimation method with dynamic compensation to address the challenges posed by nonlinear modeling errors, including friction effects from sliding cables, and coupled multi-contact force interactions. First, a unified joint-space perception model is developed to decouple external interactions from the dynamics of the driving system. Second, a directional-reversal zone integrator is designed to handle drastic state variations during motion reversals, complemented by a Gated Recurrent Unit (GRU) network to mitigate the effects of unmodeled dynamics. Third, a position-continuous optimizer is introduced to reduce solution instability caused by the coupling between external forces and contact positions. Experimental validation on a CDSM prototype demonstrates a 63% reduction in positional root mean square error (RMSE) with the synergistic combination of dynamic compensation and the enhanced solver. The implemented real-time control framework demonstrates an accuracy of 4.0 mm/1.3 N in static scenarios and 3.7 mm/2.5 N precision during dynamic operations, sustained at a 50 Hz control frequency.