Image Processing Reference
In-Depth Information
node selection process with different cost values. These cost values are described by the following
cost equations:
•
Application performance
: he minimum requirements for network performance are cal-
culated from the needed reliability of monitored data.
F
R
}
where
F
R
stands for the allowable set of possible sensor node combinations,
S
i
represents
the available sensor nodes, and
R
= {
S
i
∶∀
j
∈
JR
(
S
i
,
j
)≥
r
j
is their reliability.
•
Network costs
: Defines a subset of sensor nodes that meet the network constraints. The
network feasible set
F
N
(
S
i
)
={
S
i
∶
N
(
S
i
)≤
n
}
where
N
(
S
i
)
represents the total cost and
n
the maximal data rate the network can support.
•
Application performance and network costs are combined to the overall feasible set:
F
=
F
N
.
•
Energy: Describes the energy dissipation of the network:
C
P
F
R
∩
(
S
i
)=
∑
S
i
C
P
(
S
j
)
where
s
j
∈
C
P
(
S
j
)
is the power cost to node
S
j
.
It is up to the application to decide how these equations are weighted. his decision-making process
is completely hidden from the application. hus, the development process is simplified significantly.
MiLAN uses two strategies to achieve the objective to balance QoS and energy costs:
•
Turning off nodes with redundant information
•
Using of energy-efficient routing
The MiLAN middleware is located between network and application layer. It can interface a great
variety of underlying network protocols, such as Bluetooth and .. MiLAN uses an API to abstract
from network layer but gives the application access to low-level network components. A set of
commands identifies and configures the network layer.
12.3.6 EnviroTrack
EnviroTrack is a TinyOS-based application developed at the University of Virginia that solves a fun-
damental distributed computing problem, environmental tracking of mobile entities []. [].Therefore,
EnviroTrack provides a convenient way to program sensor network applications that track activi-
ties in their physical environment. he programming model of EnviroTrack integrates objects living
in physical time and space into the computational environment of the application through vir-
tual objects, called tracking objects. A tracking object is represented by a group of sensor nodes
in its vicinity and is addressed by context labels. If an object moves in the physical environment,
then the corresponding virtual object moves too because it is not bound to a dedicated sen-
sor node. Regarding the tracking of objects, EnviroTrack does not assume cooperation from the
tracked entity.
Before a physical object or phenomenon can be tracked, the programmer has to specify its activities
and corresponding actions. his specification enables the system to discover and tag those activities
and to instantiate tracking objects. For example, to track an object warmer than
○
C, the program-
mer specifies a Boolean function, temperature
○
C, a critical number or mass of sensor nodes,
whichfulilstheBooleanfunctionwithinacertaintime(afactthatisotenreferredtoasfreshness
of information). These parameters of a tracking object are called aggregate state. All sensor nodes
matching this aggregate state join a group. The network abstraction layer assigns a context label to
this group. Using this label, different groups can be addressed independent of the set of nodes cur-
rently assigned to it. If the tracked object moves, nodes join or leave the group because of the changed
aggregate state but the label resides persistent. This group management enables context-specific
computation.
>
