Principal Investigator: Sujit Dey, PhD
Graduate Students: Clark
N. Taylor, Debashis Panigrahi,
Dong-Gi Lee,
Naomi Ramos
Project Summary
Motivation
During wireless communication, the operating conditions of the appliance
can vary drastically over time. One way of designing a wireless appliance
is to assume the worst-case conditions and requirements. However, by designing
a mobile multimedia radio that adapts to the varying conditions and requirements
of the communication system, we can design a much more efficient and practical
wireless appliance. To enable wireless appliances to adapt to current network,
appliance, and application constraints, a method of selecting appropriate
wireless communication and application parameters must be developed. We propose
to develop a Network Aware Reconfiguration System (NARS) that will select
the appliance components to be configured and the corresponding configuration
parameters for the current network, appliance, and application requirements.
Approach
In Figure 1, we present a functional overview of the proposed NARS. The
NARS will receive data indicating the current conditions and requirements
of the network, the application, and the appliance; and produce parameter
selections that appropriately adjust the adaptive appliance to the current
conditions. The functions of the NARS can be divided into 3 steps that determine
when, what, and how to adapt to current conditions and requirements. To determine
when to adapt to current conditions, the NARS must discover current network,
application, and appliance conditions. By noting any significant changes
in the network, application, or appliance conditions, the NARS can determine
when a change in configuration is needed. In that case, the second step is
invoked, with information about the change in conditions.
Figure 1. Functional View of Network Aware Reconfiguration System
(NARS)
To determine what should be adapted in our adaptive
wireless communication architecture, the second step is responsible for
evaluating changes in the needs and conditions of the network, application,
and appliance, and determining which task(s) and component(s) should be adjusted.
For example, let us consider a WCDMA network where the first step notices
a significant increase in the noise present in the wireless channel and passes
this information on to the second step. The second step is responsible for
determining which protocol task(s) should be reconfigured to adjust for the
higher noise in the channel. If the application is not capable of handling
the increase in noise, the NARS must decrease the effects of noise in the
channel by either increasing the transmission energy of the radio, or strengthening
the channel coding being used. In CDMA based systems, the transmission energy
is set by the base station in a closed-loop system, independent of the application
being executed. Therefore, in this case, step 2 would determine that the
parameters of the channel codec must be modified to strengthen the channel
codec's noise resiliance. After step 2 determines which protocol task need(s)
to be adapted, step 3 will execute a run-time reconfiguration algorithm,
whose aim is to determine exactly how the parameters of the selected task(s)
should be configured to optimally respond to the network, application, and
appliance condition changes. Besides being able to select the optimal parameters,
the run-time reconfiguration algorithms should have very low area, delay,
and energy overhead, so that the gains obtained in the ensuing dynamic reconfiguration
far outweighs the overhead of the reconfiguration.
Current Work
In the past, we have developed run-time reconfiguration algorithms to enable
dynamic image compression for efficient wireless communication of image data.
For a dynamic image codec, the run-time reconfiguration algorithms developed
determine the optimal JPEG parameters (quantization level and virtual block
size) to meet the current network (latency and bandwidth), application (image
quality) and appliance (energy) requirements. Our approach in developing
the run-time reconfiguration algorithms for image compression consists of
quantifying the results of parameters on image quality, latency, bandwidth,
and computation and communication energy. With the knowledge of the effects
of different parameters, we have developed functions that determine the optimal
image compression parameters for current network, appliance, and application
conditions. We have demonstrated that the cost of executing the run-time reconfiguration
algorithms was less than 1% of the time and energy required to compress an
image, while energy savings of more than 80% was possible. In the proposed
research, we will extend our previous approach to developing low-cost run-time
reconfiguration algorithms to other components and tasks of our adaptive
architecture, and implement steps 1 and 2 of the NARS.
Energy/Latency/Image Quality Trade-offs in Enabling Mobile
Multimedia Communication, 12th Tyrrhenian Workshop on Digitial Communications,
CNIT, September 2000.
Adaptive Image Compression for Enabling Mobile Multimedia Communication,
in Proceedings, 2001 IEEE International Conference on Communication, June
2001.
Adaptive and Energy Efficient Wavelet Image Compression For Mobile Multimedia
Data Services, in Proc. IEEE International Conference on Communication,
April 2002. paper
Network-Aware Image Data Shaping for Low-Latency and Energy-Efficient
Data Services over the Palm Wireless Network, in Proc. World Wireless
Congress (3G Wireless), San Francisco, May 2003. paper
For any comments or questions, please contact Clark Taylor
Last modified on 6 Feb 2002.