Exponentially extended dissipativity-based filtering of switched neural networks

This paper considers the filtering problems for discrete-time switched neural networks with time delay. A unified performance index named as exponentially extended dissipativity is proposed, which combines some existing performance indices in literature such as the extended dissipativity, exponential H_∞ performance, exponential l₂−l_∞ performance, and exponentially dissipativity. By introducing extra negative quadratic state terms, a vector Wirtinger-based summation inequality is proposed. Based on these ingredients, a unified filter existence criterion is presented to ensure the filtering error systems to be exponentially stable and exponentially extended dissipative. The desired exponentially extended dissipativity-based filters for switched neural networks are achieved by solving the proposed criterion. The advantages of the exponentially extended dissipativity-based filter design result are demonstrated by two illustrating examples.