Dear Annika,

     I apologize for taking so long to supply you with the information you
requested.  Here is some information for the SpectrumWare web page.  I will
send more in early January.

     Steve Garland




Project Summary

The SpectrumWare project seeks to develop flexible and adaptive communications
and computation systems that adapt easily to unanticipated user and application
requirements, as well as to dynamically changing communication channel
conditions.

Project Description

Wireless communication and multimedia require more diverse network support than
is provided by today's wired networks or communications hardware.  Current
networks and hardware are optimized for fixed applications (e.g., reliable data
transfer or GSM cellular telephony).  They are generally designed to provide a
guaranteed level of service under anticipated worst-conditions.  As a result,
they waste resources most of the time (e.g., in the average case), and they
perform poorly much of the time (e.g., because they do not allocate resources
as they are needed).

SpectrumWare provides flexibility in system design by moving the
hardware/software boundary closer to the antenna and by performing real-time
signal processing and other computations in portable application-level
software.  The scope of the SpectrumWare project includes:
-- novel signal-processing algorithms that support digital communication over
   dynamically changing channels
-- protocols for "radioactive networks" that enable network devices to respond
   to current operating conditions and application requirements, downloding and
   installing new software as needed from server-based libaries
-- an application program interface that enables network devices to trade off
   transmission rate against power consumption

Image

There's one at the top of the spectrumware web page listed below.

Demos, movies, and other examples

Video clip of SpectrumWare applications (FM, cellular telephone, black and
white television receivers).
http://nms.lcs.mit.edu/~garland/spectrumWareDemo.mpg 

Presentations

To follow.

Publications

Perhaps John Guttag can suggest which ones are most relevant here.

Links

The SpectrumWare Project (http://nms.lcs.mit.edu/projects/spectrumware)


For your information, following the the brief project report I sent to Shuji
Kubota last month.


1) Application program interface (API) for sensing channel conditions and
   adjusting transmission to reflect application requirements

     We have developed a prototype API for an adaptable physical layer in a
wireless communication system.  The physical layer uses this API to provide
information (e.g., power consumption, latency, available bandwidth, and bit
error rate of one or more channels) to applications; applications use it to
request changes in the physical layer (e.g., a lower bit error rate).  A
controller module uses a set of high level rules to determine how best to
fulfill such a request (e.g., by increasing the amount of forward error
correction, changing the modulation format, using a different channel, or some
combination of these actions).

2) Protocols for "radioactive networks"

     We are developing protocols that controllers for adaptable physical layers
in a transmitter and receiver can use to negotiate and coordinate the kinds of
changes in transmission format mentioned in item (1) above.

3) Physical layer algorithms for digital communication over dynamically
   changing channels

     For transmission, we have developed a digital modulation technique, direct
waveform synthesis, that uses a precomputed table of samples to map discrete
data for transmission into segments of a digitally modulated waveform.  This
technique is twenty times faster than conventional modulation techniques.  It
is useful in both hardware and software implementations.

     For reception, we have developed algorithms for channel separation (i.e.,
isolating the desired signal) and detection (i.e., recovering the original
transmitted data from the channel waveform).

     Channel separation involves isolating a desired narrow-band signal from a
wideband input signal.  Typically, it requires processing proportional to the
bandwidth of the input signal.  Our approach uses a frequency shifting filter,
with processing proportional to the bandwidth of the output signal, and random
subsampling of the input signal, with processing proportional to the required
output signal quality.  Using this approach, wideband receivers can be scaled
to wider input bandwidths.

     We have extended our earlier work on matched filters to develop a
multi-threshold detector that provides a more effective balance between
computation and the output confidence levels needed for efficient overall
system performance.  This newer detector is five to ten times more efficient
than the full matched filter detector, depending on the required output quality
and input noise levels.