An event-triggered approach to state estimation for neural networks with individual triggering thresholds

Abstract

In this paper, the state estimation problem is investigated for a class of discrete-time stochastic neural networks with event-triggered transmission (ETT) mechanism. Different from the traditional periodic communication mechanism, the ETT mechanism employed in this paper possesses the advantage of mitigating the network traffic resulting from the unnecessary sending and receiving data between the sensors and the estimators. Moreover, the proposed ETT mechanism is implemented individually for each neuron. That is, the measurement output of each neuron is transmitted to the estimator only when its corresponding triggering condition with the specified triggering threshold is satisfied, thereby reflecting the engineering practice in a more realistic way. The aim of the addressed problem is to design an estimator for the neural networks subject to the ETT mechanism such that the estimation error is asymptotically bounded in the mean square. Furthermore, the estimator gain matrix is explicitly parameterized in the sense of minimizing such an asymptotic error upper bound. Finally, a numerical example is presented to demonstrate the effectiveness of the results.