Bi, Kaiming; Chen, Yuyang;Wu, Chih-Hang (John); Ben-Arieh, David Learning-based impulse control with event-triggered conditions for an epidemic dynamic

Abstract:This paper investigates the epidemic control problem of the infectious disease epidemic system. The main objective of this paper is to design an epidemic control mechanism based on data instead of the subjective empirical method. Because the uncertainty of future epidemics prevents accurate scheduling for control triggers before an epidemic occurs, epidemic intervention should be triggered by disease prevalence. In this paper, the event-triggered control (ETC) determines the most appropriate timing to implement the control, and the Learning-Based Impulse Control (LBIC) mechanism is used to determine the optimal control level. In the design of LBIC, neural networks such as convolutional neural networks, recurrent neural networks, and fully connected neural networks are trained to learn the relationship between prevalence data and historical control strategies. This paper shows that the dynamic epidemic system is stable under ETC with and without periodicity. Also, numerical simulation experiments and comparisons have proved the validity, optimality, and robustness of the proposed epidemic control method.

Engineering,Medicine