Neural network model reference decoupling control for single leg joint of hydraulic quadruped robot

Purpose To control one of the joints during the actual movement of the hydraulically driven quadruped robot, all the other joints in the leg need to be locked. Once the joints are unlocked, there is a coupling effect among the joints. Therefore, during the normal exercise of the robot, the movement of each joint is affected by the coupling of other joints. This brings great difficulties to the coordinated motion control of the multi-joints of the robot. Therefore, it is necessary to reduce the influence of the coupling of the hydraulically driven quadruped robot. Design/methodology/approach To solve the coupling problem with the joints of the hydraulic quadruped robot, based on the principle of mechanism dynamics and hydraulic control, the dynamic mathematical model of the single leg mechanism of the hydraulic quadruped robot is established. On this basis, the coupling dynamics model of the two joints of the thigh and the calf is derived. On the basis of the multivariable decoupling theory, a neural network (NN) model reference decoupling controller is designed. Findings The simulation and prototype experiment are carried out between the thigh joint and the calf joint of the hydraulic quadruped robot, and the results show that the proposed NN model reference decoupling control method is effective, and this method can reduce the cross-coupling between the thigh and the calf and improve the dynamic characteristics of the single joint of the leg. Practical implications The proposed method provides technical support for the mechanical–hydraulic cross-coupling among the joints of the hydraulic quadruped robot, achieving coordinated movement of multiple joints of the robot and promoting the performance and automation level of the hydraulic quadruped robot. Originality/value On the basis of the theory of multivariable decoupling, a new decoupling control method is proposed, in which the mechanical–hydraulic coupling is taken as the coupling behavior of the hydraulic foot robot. The method reduces the influence of coupling of system, improves the control precision, realizes the coordinated movement among multiple joints and promotes the popularization and use of the hydraulically driven quadruped robot.

Multivariable Decoupling Control Based on TC Control in the Diving and Floating Process of AUV

The diving plane motion of autonomous underwater vehicles (AUVs) is a complex multivariable nonlinear system with pitch-heave coupling. Tornambe’s controllers (TCs) can online estimate the dynamic uncertainty regardless of the type of disturbance. The TCs for depth and pitch angle are adopted to achieve multivariable decoupling control by introducing virtual control inputs. The methodology can obtain fast and non-overshoot control of depth and pitch angle in the diving and floating process. Simulation results demonstrated the effectiveness of the proposed control scheme.