Bioinspired Computation for Identifying Joint Compliance in Biomimetic Flexible Manipulators.

High-precision robotics is frequently compromised by joint compliance, a factor often over-simplified by traditional rigid-body modeling. This research investigates the structural dynamics of a two-link manipulator, addressing critical discrepancies between experimental data and conventional models. Much like biological musculoskeletal systems, joint flexibility fundamentally influences the dynamic response of articulated structures. While traditional rigid-joint models accurately capture mode shapes, they yield excessive natural frequency prediction errors with peaks reaching 72%. To bridge this gap, a refined Flexible-Joint Finite Element Model (FJFEM) is developed to mimic adaptive joint compliance. This model is integrated with a bio-inspired computational framework (a Double-Stage Genetic Algorithm Framework (DSGAF)) to identify configuration-dependent joint stiffness across the operational workspace, where experimental frequencies f1 and f2 shift nonlinearly from 25.5 Hz to 44 Hz and 92.2 Hz to 51 Hz, respectively. Experimental validation demonstrates that this evolutionary strategy reduces frequency tracking errors to less than 3.5% across all positions, achieving an average identification routine runtime of 1.8 s. By capturing nonlinear compliance behavior, this framework provides a robust foundation for the design, online calibration, and vibration control of advanced flexible robotic systems.