Akademik Veri Yönetim Sistemi
Neural Computing and Applications, cilt.35, sa.23, ss.17397-17413, 2023 (SCI-Expanded)
The anti-swing radial basis neuro-fuzzy LQR (RBNFLQR) controller for a multi-degree-of-freedom (DOF) rotary inverted pendulum is developed in this paper. One of the major challenges is to design an anti-swing RBNFLQR controller that has high precision, robustness, and vibration suppression to control the multi-DOF rotary inverted pendulum system. The study here demonstrates a novel RBNFLQR controller in which the positions and velocities of state variables multiplied by the LQR gains are tuned using the radial basis neural networks (RBNNs) architecture. The outputs of the RBNN are fuzzified by the fuzzy controller to obtain the desired torque of the pendulum systems. The RBNN based on the Bayesian regularization (BR) algorithm is able to self-adjust the LQR gains of the state variables. In order to stabilize the pendulums to zero positions more effectively, the tuned gains of LQR help to reduce the aggressiveness of the fuzzy control rules. The control performance of the anti-swing RBNFLQR controller was verified by simulation and experimental results in two, three, and four DOF rotary inverted pendulum systems. The proposed controller exhibits robustness to external disturbances and has much better vibration suppression capability. The present work provides a novel and effective framework to develop an anti-swing RBNFLQR controller for multi-DOF pendulum systems.