OPTIMAL CONTROL OF INVERTED PENDULUM BASED ON TWO PID AND LQR ARRANGEMENT AND IMPROVED BP NEURAL NETWORK

BP Neural Network has a longer training time and a slow convergence. To deal with the defects of BP Neural Network a modified BP algorithm is proposed in the paper. The algorithm is applied for the control of Inverted Pendulum, a highly non linear system inherently being open loop unstable. Levenberg-Marquardt algorithm is used for the training purpose. The training samples are being collected by a two PID and one LQR arrangement. The simulation results prove that the modified BP algorithm for inverted pendulum control gives better efficiency, lesser training time and faster convergence.