Design and parameter identification of model-based control integrating hydraulic cylinders in robotic leg dynamics

Abstract

With the widespread application of model-based controllers in legged robots, the model accuracy is emphasized. In hydraulic robots, joints are typically closed-chain structures formed by the articulation of hydraulic cylinders. However, it is common practice to disregard the cylinder links in the dynamics of various advanced controllers, leading to inaccuracies in the models. To address this problem, this paper proposes a model-based control framework for the hydraulic legged robot, which integrates the cylinder links into the rigid body dynamics. The model establishment and controller design are separately carried out for the stance and swing phases. The model parameters in the controllers are obtained through identification. Firstly, the hydraulic parameters in nonlinear form are identified using a combination of the least squares and particle swarm optimization algorithm. Secondly, to obtain robust parameter estimates for floating base dynamics, we excite the friction of individual cylinders, the stance and swing phases of the leg system respectively. Subsequently, all excitation results are integrated, and physical consistency constraints are added to formulate a convex optimization problem. The proposed framework has been validated through a series of prototype experiments, demonstrating the benefits of employing a more accurate model in position tracking performance.