Integrated Bio-inspired Synergistic Framework with Self-Sensing Control for Anthropomorphic Coordination of TCP-Driven Hands

Abstract Twisted and coiled polymer (TCP) actuators have recently been used in a broader range of robotic systems, largely because they combine muscle-like compliance with relatively high power density. For anthropomorphic robotic hands, this combination is useful in a very direct sense: the actuator must fit into a compact structure while still producing compliant finger motion. However, the coupled electro-thermal-mechanical response of TCP materials makes internal-state estimation difficult. It also complicates coordinated multi-finger actuation, especially when additional sensors cannot be easily embedded in a soft hand without increasing size, wiring complexity, or structural redundancy. In this work, we develop a synergistic control framework with temperature self-sensing for regulating multiple TCP actuator arrays in a soft dexterous hand. Electrical and mechanical signals are collected through an integrated sensing architecture, after which principal component analysis (PCA) is used to establish a low-dimensional relationship between input power commands and finger bending trajectories. To make the actuator response more tractable, the total contraction force is divided into a temperature-dependent component and a mechanical component. On this basis, a predictive thermal regulation model is constructed, allowing the TCP arrays to be controlled without relying on external temperature sensors. The framework was evaluated in Feix grasp-maintenance and grasp-transition experiments. Across three representative grasp-maintenance tasks with five repeated trials per condition, self-sensing control reduced the mean fingertip dispersion by 37.5-42.1% in the X direction and by 8.1-14.7% in the Y direction compared with open-loop actuation. The corresponding planar fingertip dispersion decreased by 16.3-36.5%. During grasp transitions, Dz changed by 1.4–30.7% across the four transition tasks, with pronounced reductions in Feix-09–10 and Feix-26–27–28, whereas Feix-12–13–14 showed only a small reduction. These results indicate that temperature self-sensing improves both trajectory repeatability and hand-level coordination in the TCP-driven dexterous hand.