Bioinspired PID Controller Based on Izhikevich Neurons Optimized by Differential Evolution for Neuromorphic Implementations

The well-known proportional-integral-derivative (PID) controllers are challenged by variations in the process dynamics, inadequate tuning and other difficulties associated with the process itself. Other controller structures, such as the biologically inspired spike neural networks (SNN), were developed and essayed against such challenges. The goal of the present work is to develop a bioinspired PID controller with a small SNN using the neuron model proposed by Izhikevich, and optimized by a Differential Evolution algorithm for neuromorphic implementations. The controller architecture is based on 3 neurons, each responding with a proportional, derivative or integrative output. The parameters of each neuron were optimized with either a step signal or a sinusoidal waveform of variable amplitude and frequency as reference. The results show that the SNN controller performed similarly to a traditional PID controller for the two applications. The best result found is represented by a mean square error of 1.605 for the step signal. Improvements might be achieved by further optimization steps, an aim for future research. Neuromorphic control applications, such as in the control of prostheses, appear as potential applications.