Research on Flexible Joint MPC Optimization Control Method Based on Joint Drive Parameter Identification

With the widespread application of high-dynamic quadruped robots, the influence of flexible joint structures on their motion control performance has become increasingly significant. Conventional model predictive control (MPC) methods often neglect the stiffness changes of joint drives during motion, leading to unstable foot force distribution and exacerbated center of mass attitude fluctuations. To address this issue, this paper proposes a flexible joint model predictive control method (FJ-MPC) that integrates joint drive parameter identification and real-time optimization. This method constructs a joint flexibility model based on the load-angle-stiffness nonlinear characteristics of harmonic drives, and introduces a stiffness variation penalty term into the cost function to achieve smooth stiffness transitions. Simulation results show that the proposed method significantly reduces stiffness fluctuations and improves center of mass stability under diagonal gaits. This study provides new insights into high-precision flexible modeling and control for quadruped robots, offering significant theoretical value and engineering application potential.