Hybrid genetic algorithm-system identification approach to model force sensing resistors

Abstract

Force sensing resistor (FSR) is a passive component that is composed of polymer thick films that change resistance between its terminals due to force applied at its surface. FSRs inherently exhibit many nonlinear behaviors. This work employs a Genetic Algorithm agent to navigate the search space to identify the optimal modeling systems for five circular FSRs of comparable sizes. The Hybrid GA-System Identification allows the globally optimized models for the original systems to be identified without the need of a differentiable measure function or linearly separable parameters. The GA searches for the order of the linear model (zeros and poles), the input and output nonlinearities, and the order and the interval of these nonlinearities. Meanwhile, the system identification optimizes the locations of the poles and zeros as well as the parameters of the input and output nonlinearities. The synergy between the two agents allows the entire space to be evaluated for a global solution using the heuristic search advantage of the GA coupled with the fine-tuning of the parameters using the localized search advantage of the system identification. Results show that using the GA agent expedited the search process and allowed for reaching a globally optimized model.