Automated High-Precision Control of Twisted and Coiled Polymers Under Parameter Variability

Twisted and coiled polymer (TCP) artificial muscles offer high energy density and large deformation, but achieving reliable closed-loop control remains difficult due to time-temperature-dependent parameter drift, structural variability, and labor-intensive control parameter tuning, which accelerate actuator wear through repeated adjustments. This work proposes an automated two-stage IGAPSO (improved genetic particle swarm optimization)-based framework that integrates neural-network-assisted offline optimization with rapid online fine-tuning on physical actuators, minimizing manual calibration while ensuring robust performance. Experiments demonstrate rapid and precise tuning of TCP control parameters: in step force control, convergence occurs within eight iterations with a 0.14 s settling time and 0.02% steady-state error; in step displacement control, tuning completes in three iterations with a 0.55 s settling time and 0.14% error. Zero-overshoot responses are realized by adjusting the optimization objective. The method further achieves robust waveform tracking (mean RMSEs of 1.90 mN force and 0.47 mm displacement), enables secondary retuning after long-term placement within seven iterations, and transfers effectively to TCPs with different spring indices. This work provides an efficient and generalizable strategy for accurate, robust, and adaptive TCP control, paving the way for broader integration in robotic systems.