[Firebug-cvs] firebug/web spie_2004.tex,1.7,1.8 spie_reviews.tex,1.3,1.4

[<do...@us...>]

Update of /cvsroot/firebug/firebug/web
In directory sc8-pr-cvs1:/tmp/cvs-serv10040a

Modified Files:
	spie_2004.tex spie_reviews.tex 
Log Message:
Copied main body of reviews into paper.  The reviews
file should be left as is, but the review in the main paper 
need to be condensed.

Index: spie_2004.tex
===================================================================
RCS file: /cvsroot/firebug/firebug/web/spie_2004.tex,v
retrieving revision 1.7
retrieving revision 1.8
diff -C2 -d -r1.7 -r1.8
*** spie_2004.tex	21 Jul 2003 23:40:18 -0000	1.7
--- spie_2004.tex	22 Jul 2003 21:02:38 -0000	1.8
***************
*** 99,107 ****
  \subsection{Previous work with outdoor sensors}
  
! \input{spie_reviews}
  
  \paragraph{Mehta and El Zarki}~\cite{mehta:v2002} ????
  
  
  \subsection{Pr\'ecis}
  
--- 99,296 ----
  \subsection{Previous work with outdoor sensors}
  
! %\input{spie_reviews}
! 
! 
! 
! \paragraph{Ganesan et al.}~\cite{ganesan:d2002} 
! have done the first large scale study 
! for sensor networks to explore the behavior of such a 
! distributed system. Their data provides insight into 
! designing protocols and algorithms for scalable distributed 
! sensor networks. They differentiate how each of the layers 
! in the protocol stack, link, medium access, routing and 
! application is affected.
! 
! Their study involves using 150 motes laid out in a parking lot. 
! Link characteristics and statistics as well as aggregate 
! statistics for packet loss, and packet forwarding are gathered. 
! At the medium access layer they use timing information to 
! identify metrics that capture both end-to-end properties of a 
! basic flood propagation, and local properties such as contention 
! and collision. At the node level, they maintain the number of 
! hops to the base-station and the number of children each node 
! has to take into account the clustering dynamics.
! 
! What is observed at scale is termed as non-uniform flood 
! propagation. This is shown by the existence of  backward links, 
! long links, assymetrice links, nodes that are missed out in the 
! flood process and even clustering behavior wherein a few nodes 
! have many children.
! 
! 
! \paragraph{Bulusu et al.}~\citeyear{bulusu:n2002,bulusu:n2001}
! have developed techniques for 
! RF-based localiazation in sensor
! networks. Their work can be divided into two 
! distinct problem areas, namely localization, 
! and beacon placement. They profess a GPS-less 
! environments with beacon nodes that are 
! deployed with their absolute location/position 
! stored within them and sensor nodes that localize 
! themselves based on proximity to a subset of beacons.
! 
! Some of the salient characteristics of their approach are: 
! 
! \begin{itemize}
! 
! \item an idealized radio model with perfect spherical 
! radio propagation and identical transmission range for all radios, 
! 
! \item  a localization algorithm that relies on a 
! connectivity metric threshold to decide
! which beacons to use in computation of a nodes 
! position estimate, which is the centroid
! of the beacons selected.
! 
! \item  measurement based adaptive beacon placement, 
! which can adjust the density of beacons
! and in turn increasing system lifetime, 
! and reduced excessive channel interference and 
! contention.
! \end{itemize}
! 
! As per their results the spherical radio model 
! correlates upto 90% with real conditions in 
! an outdoor uncluttered environment.
! 
! 
! 
! \paragraph{Simic et al.}~\cite{simic:sn2003}
! describe distributed computation 
! and estimation of a scalar field in a sensor 
! network. The scalar field could be anything 
! from temperature to amount or intensity of 
! light. They provide an algorithm and precise 
! theoretical analysis of it.
! 
! The basic idea of their algorithm is as follows. 
! Each node communicates with its neighbors
! and computes the maximal difference quotient 
! of the sensed scalar field. The estimate
! of the gradient at each node is taken to be 
! the vector in the corresponding direction with
! norm equal to the maximal difference quotient. 
! The method amounts to approximate differentiation 
! of the function defined by the scalar field, 
! given its value on a set of random points. 
! They analyze the accuracy and complexity of 
! the algorithm from a probabilistic point of 
! view. The estimated probability that the 
! error is small and converges to one, as the 
! number of nodes goes to infinity, is shown.
! 
!  
! \paragraph{West et al.}~\citeyear{west:b2001}
! are designing parts of the sensor 
! network architecture with microclimate
! monitoring as their application for focus. 
! Their necessities for an architecture stem 
! from preliminary testing. To understand the 
! performance of low-power transceivers, they 
! have taken propagation measurements using 
! commercially available equipment in the local
! Ponderosa pine forests. An interesting 
! observation made in this environment was that if
! the antenna was placed at a height of less 
! than 1 meter, the range severely degraded.
! 
! Some of the suggestions that they make to 
! expand current sensor network implementations 
! and architectures are:
! \begin{itemize}
! \item  Distributed source coding of spatio-temporally 
! correlated vector process.
! \item  Multi-hop protocols with inter-layer 
! interaction, as interaction between layers 
! may prove benfecial in a more compact and power efficient system design.
! \item  Coded macrodiversity in energy-limited 
! multi-hop nets where a transmission using a 
! basic radio is heard by multiple neighboring nodes.
! \end{itemize}
! 
! 
! \paragraph{Ramadurai and Sichitui}~\citeyear{ramadurai:v2003,sichitiu:ml2003}
! develop distributed algorithms 
! for outdoor localization in sensor networks. 
! The method is based on radio-frequency (RF) signal strength
! measurements, which tend to have a certain degree 
! of inaccuracy. They define two classes of nodes, 
! namely unknown and beacon nodes. The "beacon" nodes have known
! absolute (using GPS) positions, and the "unknown" 
! nodes do not know their positions.
! The beacon nodes peridically inject packets which 
! are used to form position estimates
! at the unknown nodes. Once an unknown nodes has 
! a rough idea of its position, it can
! assist other unknown nodes in estimating 
! their position.
! 
! They employ two techniques for associating signal 
! strength measurements to distance. Both techniques 
! are grounded through preliminary empirical data. 
! The first uses bounded values that associate an 
! RSSI reading to a distance range, plus adding power level
! variability gives it increased accuracy and granularity. 
! The second uses probabilistic position estimation, 
! where any node receiving a beacon packet will 
! estimate itself to be located on a surface that 
! has a probability distribution dictated by a mean and 
! standard deviation corresponding to the signal 
! strength received.
! 
! The measurements for both these approaches were 
! collected outdoors with very little interference, 
! however their data also indicates that even in 
! heavily wooded areas the signal propagation is 
! approximately circular, and hence the signal strength is 
! linearly proportional to the distance.
! 
! 
! \paragraph{Mainwaring et al.}~\citeyear{mainwaring:a2002}
! have had success in deploying 
! large scale sensor networks as part of
! a habitat monitoring project at the Great Duck 
! Island. It is reflective of the domain of applications 
! for which sensor networks are going to be used. 
! It serves as a collection of requirements, costraints 
! and guidelines that serve as a basis for a general 
! sensor network architecture. 
! 
! They present a tierd architecture, where at the 
! lowest level are sensor nodes, which perform general 
! purpose computing and are organized into sensor 
! patches. Individual sensor nodes transmit their data 
! through a patch to a sensor network gateway which links
! the sensor patch through a local transit network to 
! a remote basestation. The transit network consists 
! of one or more hops of wireless links. The current 
! system consists of 32 nodes on a small island off 
! the coast of Maine streaming live data onto the web.
! 
! The main tasks at hand that the network spends 
! resources for are 1) data sampling and collection, 
! 2) communications, 3) network retasking and 4) 
! health and status monitoring.  Keeping these in mind, 
! energy required for each task is projected and taken 
! into account to estimate how long the sensor nodes 
! will survive.
! 
! 
  
  \paragraph{Mehta and El Zarki}~\cite{mehta:v2002} ????
  
  
+ 
+ 
  \subsection{Pr\'ecis}
  
***************
*** 117,144 ****
  One paragraph summary of AM here.
  
! 
! Tiny Active Messages ~\cite{hill:j2000} attempt to preserve the
! basic concepts of integrating communication with computation
! and matching communication primitives to hardware
! capabilities. Each message contains the name of a 
! user-level handler to be invoked on a target node upon arrival 
! and a data payload to pass in as arguments. The handler function 
! serves the dual purpose of extracting the message from the network
! and either integrating the data into the computation
! or sending a response message. The basic paradigm of typed messages
! causing handlers to be invoked upon arrival matches up
! well with the event based programming model supported
! by TinyOS (the operating system) and demanded by the underlying 
! sensor hardware. The low overhead associated with event based 
! notification is complementary to the limited resources of networked
! sensors. Applications do not need to waste resources
! while waiting for messages to arrive. Additionally,
! the overlap of computational work with application
! level communication is essential. Execution contexts and
! stack space must never be wasted because applications are
! blocked, waiting for communication. Essentially, the active
! messages communication model can be viewed as a
! distributed eventing model where networked nodes send
! each other events. 
  
  
--- 306,333 ----
  One paragraph summary of AM here.
  
! 
! Tiny Active Messages ~\cite{hill:j2000} attempt to preserve the
! basic concepts of integrating communication with computation
! and matching communication primitives to hardware
! capabilities. Each message contains the name of a
! user-level handler to be invoked on a target node upon arrival
! and a data payload to pass in as arguments. The handler function
! serves the dual purpose of extracting the message from the network
! and either integrating the data into the computation
! or sending a response message. The basic paradigm of typed messages
! causing handlers to be invoked upon arrival matches up
! well with the event based programming model supported
! by TinyOS (the operating system) and demanded by the underlying 
! sensor hardware. The low overhead associated with event based 
! notification is complementary to the limited resources of networked
! sensors. Applications do not need to waste resources
! while waiting for messages to arrive. Additionally,
! the overlap of computational work with application
! level communication is essential. Execution contexts and
! stack space must never be wasted because applications are
! blocked, waiting for communication. Essentially, the active
! messages communication model can be viewed as a
! distributed eventing model where networked nodes send
! each other events. 
  
  

Index: spie_reviews.tex
===================================================================
RCS file: /cvsroot/firebug/firebug/web/spie_reviews.tex,v
retrieving revision 1.3
retrieving revision 1.4
diff -C2 -d -r1.3 -r1.4
*** spie_reviews.tex	21 Jul 2003 23:40:56 -0000	1.3
--- spie_reviews.tex	22 Jul 2003 21:02:38 -0000	1.4
***************
*** 189,190 ****
--- 189,194 ----
  into account to estimate how long the sensor nodes 
  will survive.
+ 
+ 
+ 
+ \paragraph{Mehta and El Zarki}~\cite{mehta:v2002} ????