Cellular Neural Networks Based Time-Series Approximation for Real Time Systems' Modeling-and-Identification and Behavior Forecast in Transportation: Motivation, Problem Formulation, and Some Research Avenues

This paper discusses the potential of using "time series approximation" for mathematical modeling, online system identification and forecasting of the dynamical behavior of scenarios in the field of traffic and transportation. The tremendous attention devoted to both modeling and forecasting (in transportation) is justified whereby some challenges and unsolved research issues are discussed. Due to the time-varying dynamics experienced by transportation related systems/scenarios, an appropriate identification process is necessary and should be applied to determine the parameter settings of the corresponding mathematical models in real time. The concept of a simulation and computing platform design based on the cellular neural network (CNN) paradigm will be presented. Then the capability to study the spatio-temporal and time-varying dynamics exhibited by time-varying transportation systems/scenarios will be demonstrated. In the essence, we develop a concept that uses the CNN model as a universal mathematical system-model and/or system-model approximator.