Digital Signal Processing Reference
In-Depth Information
case study, this corresponds to less than 3000 bits per state. A software-based QoS
Module within the AP's network management layer maintains the list segments of
all associated nodes and processes the current state of each node. At the beginning of
each period, it executes the run-time phase of the control method and determines the
configuration for each node during that period. The scheduling period requirement
is determined by the rate at which the system state varies. Channel measurements
show coherence times of 166 ms for stationary objects and moving scatterers [37].
Given a video frame rate of 30 ms, it is clear that this requires a scheduling period
less than 30 ms. Since this timing requirement is rather low, a software module is
sufficient. Alternatively, the QoS module can be integrated in a light-weight RTOS
present in most embedded devices (as done in [89]). We expect the performance of
the control method be lower for mobile networks with faster state dynamics, when
it is difficult to feedback the system state timely. Faster adaptation schemes will
be needed that integrate the adaptation module in hardware close to the physical
layer [90].
6.5 Adapting to the Dynamic Context
We now show how a cognitive radio can save energy by optimally adapting to the
varying context. More specifically, the smart radio knows how to adapt its flexibil-
ity across many layers to achieve the best configuration in terms of QoS, energy or
resource use. The focus is on real-time streaming media applications with a reason-
able target
JFR
∗
set to 10
−
3
. In order to evaluate the relative performance of the
cognitive control method, we consider four comparative transmission strategies:
1.
Smart:
This is the optimal scheme considering the energy trade-off between
sleep and scaling, exploiting multi-user diversity. The node configurations are
based on the profiles of Sect.
6.4.1
.
2.
PHY-layer:
This scheme considers only physical layer scaling knobs. The
Energy-TXOP profiles are set to scale maximally as no sleep mode is available.
3.
MAC-layer:
In this scheme only sleeping is possible by the energy-aware MAC-
layer. The physical layer is fixed to the largest constellation and code rate, with
maximum transmit power. This approach is used by commercial 802.11 de-
vices [8]. However, the 2-frame buffering makes the proposed implementation
more efficient as it eliminates all idle time between transmissions.
4.
Fixed:
Similar to MAC-layer scheme but the transceiver remains in idle after
transmission. A basic scheme with no energy management and hence no dynamic
range is exploited here.
The Energy-TXOP profiles are computed for the control dimensions considered in
each scheme and used by the scheduling scheme implemented in the Network Simu-
lator ns-2 [91]. This simulator has been extended with transceiver energy and perfor-
mance models, and a slow fading channel model. Our simulation model implements
the functions of the 802.11e with beaconing, polling, TXOP assignment, uplink, and
downlink frame exchange, fragmentation, retransmission and variable super-frame
