J. Liu et al. / Fuzzy Sets and Systems 273 (2015) 124–140 127
Suppose the time-varying delay in the network communication is d
k
and d
k
∈[0,
¯
d), where
¯
d is a positive real
number. Therefore, the sampled sensor measurements y(t
0
h), y(t
1
h), y(t
2
h), ··· will arrive at the filter side at the
instants t
0
h +d
0
, t
1
h +d
1
, t
2
h +d
2
, ···, respectively. ˆy(t) in Eq. (4) can be described as [5,14]
ˆy(t) =
r
i=1
h
i
C
i
x(t
k
h), t ∈[t
k
h +d
t
k
,t
k+1
h +d
t
k+1
] (5)
where h is the sampling period, t
k
∈{1, 2, 3, ···}. d
t
k
and d
t
k+1
are the network induced delays at the transmission
instant t
k
h and t
k+1
h, respectively.
Considering the possible sensor f
ailure, (5) can be rewritten as
ˆy(t) =
r
i=1
h
i
ΞC
i
x(t
k
h) =
r
i=1
m
l=1
h
i
Ξ
l
E
l
C
i
x(t
k
h), t ∈[t
k
h +d
t
k
,t
k+1
h +d
t
k+1
] (6)
where Ξ = diag{Ξ
1
, Ξ
2
, ···, Ξ
m
} with Ξ
i
(i = 1, 2, ···, m) being m unrelated random variables taking values on
the interval [0, θ ], θ ≥ 1 and E
l
= diag{0, ···, 0
l−1
, 1, 0, ···, 0
m−l
}. The mathematical expectation and variance of Ξ
i
(i =1, 2, ···,m) are
¯
Ξ
i
and δ
2
i
, respectively.
¯
Ξ
i
and δ
2
i
can determined the failure rate and the distortion degree
of the ith sensor.
Define
¯
Ξ =diag{
¯
Ξ
1
,
¯
Ξ
2
, ···,
¯
Ξ
m
}, we can easily derive
¯
Ξ =
m
l=1
¯
Ξ
l
E
l
. For a matrix Θ>0, we can get
⎧
⎪
⎨
⎪
⎩
E{Ξ −
¯
Ξ}=0
E
(Ξ −
¯
Ξ)
T
Θ(Ξ −
¯
Ξ)
=
m
l=1
δ
2
i
E
T
l
ΘE
l
Remark 2. When sensors have faults, the output signal may be larger or smaller than what it should be. Considering
this case, we assume the variables Ξ
i
(i = 1, 2, ···, m) take values in the interval [0, θ], θ ≥ 1. When Ξ
i
∈{0, 1}, it
means the sensor i has completely failure or not. Ξ
i
=1 means the sensor i works normally, Ξ
i
=0 means signal sent
by sensor i is lost during transmission. Moreover, 0 <Ξ
i
< 1 and Ξ
i
> 1 means the case of data distortion happen,
that is, the signal at the filter is smaller or greater than it actually is.
As is well kno
wn, the widely used periodic sampling mechanism may lead to transmit many unnecessary signals,
which reduces bandwidth utilization. In order to reduce the load of network transmission and save the network band-
width, there is a great need to introduce an event triggered mechanism which decides whether the newly sampled data
should be send out to the filter
. As is shown in Fig. 1, similar to [11], we introduce an event generator between the
sensor and the filter. The sensor measurements are sampled regularly by the sampler of the smart sensor with period h,
which will be given in sequel. Whether or not the ne
wly sampled sensor measurements will be sent out to the filter
are determined by the following judgment algorithm:
E
¯
Ξy
(k +j)h
−E
¯
Ξy(kh)
T
Ω
E
¯
Ξy
(k +j)h
−E
¯
Ξy(kh)
≤ρ
E
¯
Ξy
(k +j)h
T
ΩE
¯
Ξy
(k +j)h
(7)
where Ω is a symmetric positive definite matrix, j =1, 2, ···, and ρ ∈[0, 1). Only when the current sampled sensor
measurements y((k + j)h) and the latest transmitted sensor measurements y(kh) variate the specified threshold (7),
the current sampled sensor measurements y((k + j)h) can be transmitted by the event generator and sent into the
filter
.
Remark 3. From e
vent-triggered algorithm (7), it is easily seen that the sensor measurement are sampled at time kh
by sampler with a given period h, the next sensor measurement is at time (k +1)h. Suppose that the release times are
t
0
h, t
1
h, t
2
h, ···, it is easily seen that s
i
h = t
i+1
h −t
i
h denotes the release period of event generator in (7), s
i
h means
that the sampling between the two conjoint transmitted instant.