摘要
This addresses errors in [1]. The following references to (10) should have been to (1) and are corrected below. Section II.A, paragraph 1: The system matrices A, B d , B a , G, J, C, B n and B s in (1) are with appropriate dimension. Section II.A, Assumption 1: The continuous-time linear stochastic jump diffusion system in (1) is observable, i.e., the pair (A,C) is observable. Section II.A, Assumption 2, paragraph 2: In the linear stochastic jump-diffusion system in (1), the diffusion term Gx(t)dw(t) denotes the continuous but non-differentiable random fluctuation and the jumping term Jx(t)dp(t) denotes the discontinuous random fluctuation. Section II.A, lead-in text to (9): Then, the augmented system is proposed to include the smoothed signal models (5) and (8) of malicious attack signals on system and sensor into (1): Section II.A, Remark 3: If we estimate the state x(t) in (1) by Luenberger type filter directly, the estimated {hat{x}(t) could be corrupted by attacked signals f a (t) and f s (t). Section II.A, Theorem 1: For the continuous-time linear stochastic jump diffusion system in (1), if the system (A,C) is observable, i.e., Section II.B, paragraph 1: In order to optimally estimate state and malicious attack signals in (1), based on the augmented system in (9) without considering v(t), the following the stochastic H 2 filtering performance of the Luenberger-type filter in (15) is introduced [5] and [6]: Section II.B, Remark 4, lead-in text to (18) In order to achieve the stochastic optimal H∞ robust filtering performance and the stochastic optimal H2 filtering performance simultaneously, the MO H 2 /H∞ SEF design of stochastic system with external disturbance and malicious attack signals in (1) is formulated as follows: The following references to (13) should have been to (4) and are corrected below. Section II.A, Remark 2, paragraph 1: Since the extrapolation error is considered in (4), the smoothed signal model in (4) is able to extrapolate the value of malicious function at future sample point by the current and previous values of malicious function as possible. Section II.B, Remark 4, paragraph 1: Since the extrapolation errors τ a (t) and τ s (t) in (4) and (7) are unavailable and it will influence on the estimation performance, the worst-case effect of extrapolation errors τ a (t) and τ s (t)in (17) should be attenuated as small as possible.