Modification of Neuron PID Control in Case of Improper Learning Factors

Abstract

Some modifications of conventional neuron proportional-integral-differential controller (NPID) are presented in this paper to prevent its slow dynamic response and loss of control in case of improper learning factors. The quasi-step signal replaces the step signal as the reference signal to improve the dynamic characteristics. The control output of NPID is modified every step by multiplying a penalty factor called senior teacher signal to suppress further the overshoot and compress the settling time. The steady-state error from the modified NPID (MNPID) is reduced or removed by adjusting dynamically reference input signal while excluding the pseudo steady state. Lots of simulation experiments are done to prove the stability and convergence of the MNPID control algorithm.