Dual-loop control of lower-limb exoskeletons using adaptive admittance and nonlinear extended state observer-based active disturbance rejection

Lower-limb exoskeleton systems are characterized by strong human–robot coupling, dynamic uncertainties, and multiple concurrent disturbances, which challenge traditional control methods to simultaneously ensure system stability, precise trajectory tracking, and human–robot interaction compliance. To address these challenges, this paper proposes a dual-loop control method tailored for diverse gait rhythms. This method consists of an outer-loop double-layer adaptive admittance control and an inner-loop nonlinear extended state observer-based active disturbance rejection control (NLESO-ADRC). The outer loop establishes baseline admittance parameters based on different velocity levels and performs online adjustments of virtual damping and stiffness by incorporating real-time human–robot interaction torques and admittance corrections, thereby generating smoother and more compliant reference trajectories. The inner loop treats unmodeled dynamics, parameter uncertainties, and friction errors as total system disturbances, which are estimated and compensated for in real-time by the NLESO, thus enhancing joint trajectory-tracking precision and disturbance rejection capabilities. To validate the effectiveness of the proposed approach, various control combinations were constructed based on gait experiment data, and comparative analyses were conducted from multiple perspectives. The results demonstrate that the proposed method not only maintains high trajectory-tracking accuracy and disturbance rejection capabilities across different walking speeds but also achieves compliant regulation during human–robot interaction.