Flexible-joint manipulators suffer from severe performance degradation due to the coupling of joint elasticity and varying loads. To address this, we propose a composite finite-time active disturbance rejection control (CFT-ADRC) strategy utilizing a frequency-domain separation architecture. A recursive least squares (RLS) algorithm identifies slow-varying load parameters, while an extended state observer (ESO) compensates for high-frequency unmodeled dynamics and external disturbances, effectively preventing loop interference. A finite-time control law guarantees rapid tracking error convergence. Comprehensive simulations confirm that this approach significantly outperforms standard ADRC and neural network-based methods (RBFNN-ASMC). Under 50% load variations, it achieves an RMS tracking error of 2×10−3 rad and maintains robust stability during 200% instantaneous load mutations. The strategy presents a strong theoretical framework for future hardware implementation while maintaining an optimal balance of precision, robustness, and computational simplicity.
Shao et al. (Sun,) studied this question.