Image Processing Reference
In-Depth Information
higher delivery velocity. As a result, the authors infer that by controlling the transmission power it is
possible to control the communication delay under light workloads.
However, the side effect of increasing the transmitting power is that the overall network capacity
decreases due to increased contentions and interferences. Hence, the main idea of the RPAR protocol
is to achieve a good trade-off between delay, energy consumption, and network capacity by dynam-
ically adapting the nodes' transmitting power. That is, when deadlines are tight, the transmitting
power is increased so that delay decreases, while when deadlines are loose, the transmitting power is
lowered, so that energy consumption is reduced.
The RPAR protocol builds upon four components: a dynamic velocity assignment policy, a
forwarding policy, a delay estimator, and a neighborhood manager.
The velocity assignment policy is used to transform the deadline requirements of a packet into
velocity requirements. More specifically, given a source node
S
and a destination node
D
,the
forwarding velocity required to meet the deadline is
d
(
S
,
D
)
v
req
(
S
,
D
)=
(.)
slack
rec
(
S
)−(
t
head
(
S
)−
t
rec
(
S
))
where
d
(
S
,
D
)
is the Euclidean distance between
S
and
D
slack
rec
is the slack of the packet when it is received by
S
, i.e., the amount of time left before
the deadline expires
t
rec
(
S
)
are respectively the time instant when the packet is received and the time
when it becomes the head of the transmission queue
If the velocity at each hop remains greater than the relevant
ν
req
(
S
)
and
t
head
(
S
)
value, then the packet will meet
the end-to-end deadline. hus, the global problem of meeting an end-to-end deadline is mapped into
a local problem of maintaining the velocity requirement of the single hop.
The delay estimator is assigned the task of assessing the one-hop delay for each forwarding choice
maintained by the neighborhood manager. Each forwarding choice is a pair (
N
,
p
), where
N
is a
neighbor node and
p
is the selected transmitting power. In general, the one-hop delay depends on the
contention delay, delay
cont
−
, the transmission time of the packet and its acknowledgment, delay
tran
,
and the number of retransmissions,
R
(
S
)
(
S
,(
N
,
p
))
. As a result, its value can be expressed as
delay
(
S
,
(
N
,
p
))=(
delay
cont
(
S
)+
delay
tran
)⋅
R
(
S
,
(
N
,
p
))
(.)
Since the transmission time is known from packet size and network bandwidth, the parameters that
the delay estimator has to predict are contention delay and the number of retransmissions. A per-
node estimation of delay and a per-forwarding-choice estimation of retransmissions are performed
using the Jacobson algorithm [Jac].
The forwarding policy chooses the most adequate forwarding choice for each packet, that is, the
mosteicientchoiceamongthosethatreachtherequiredvelocity
v
prov
.Sothevelocityprovidedby
each (
N
,
p
)
pair has to be calculated before a forwarding decision can be made. The value can be
calculated as
d
(
S
,
D
)−
d
(
N
,
D
)
v
prov
(
S
,
D
,
(
N
,
p
))=
(.)
delay
(
S
,
(
N
,
p
))
where the term
d
is the progress made toward the destination when node
N
is
chosen as the next hop, while the denominator is the delay as provided by the delay estimator. In
order to choose the most energy-efficient forwarding choice, the energy consumption of each choice
(
S
,
D
)−
d
(
N
,
D
)
