Amit Saha:                                                                                
                                                                                
Title: Range-Free Localization Schemes for Large Scale Sensor Networks
Authors: He, Huang, Blum, Stankovic, Abdelzaher
MobiCom 2003
----------------------------------------------------------
                                                                                
Abstract:
---------
The authors propose a range-free localization scheme and compare it
against existing range-free techniques and show that their technique
provides better accuracy.  Range-free techniques are those techniques
which do not require estimation of distance from signal strength.
                                                                                
Comments:
---------
The APIT technique seems to be a simple approximate technique and since
sensor networks probably do not need very accurate location estimates,
this technique might be good enough for many sensor applications.
                                                                                
The title of the paper is a bit misleading since the title seems to
suggest that they are going to propose several range-free localization
schemes. They actually propose just one and compare against other 3
existing ones. Comparing with at least one or two existing techniques
would anyway be necessary for this work to be accepted unless this is the
first technique (which it is not). The effect of accuracy on routing
performance is also not relevant since such results have obviously been
shown in other papers whose entire job is to do routing with location
information. It seems that section 7 (Localization Error Impacts) is has
just been tagged on to the paper.

Because of the large overhead of Amorphous and DV-HOP the difference
betweent the overhead of the Centroid technique and the proposed APIT
technique is less visible.  Also, with density going up the overhead of
APIT is going to be higher. The computational overhead should also have
been compared since a computationally intensive technique would kill the
sensor nodes. However, for stationary sensor nodes, this might not be a
problem since positions need to be calculated once and maybe have to be
recalculated very infrequently.


*****************************************************************************

Dan Sandler:
                                                                                
In wireless sensornet scenarios, individual sensor nodes may need to be
able to determine their absolute position in order to make their data
meaningful.  Solutions to this location problem fall into two broad
categories: range-based (involving direct measurement of inter-node
distance or direct conversion of delay to distance) and range-free
(estimation of location using known locations of landmarks).  The
range-free APIT technique developed by the authors relies on an array
of well-distributed landmark nodes to emit beacons audible by all
sensors.  An individual sensor selects an arbitrary triangle composed
of three landmarks (whose beacons it can hear); if the sensor node
determines it is inside the triangle, it can approximate its location
from the locations of the landmarks.  This approximation is repeated
and averaged for all available triangles, until the desired accuracy is
reached.  The authors compare APIT to other range-free techniques and
demonstrate its accuracy, even in the presence of realistic
channel-fading and asymmetry.
                                                                                
The range-free localization technique described relies on two
properties of the sensornet topology: (1) Each sensor must be in radio
range of many beacons (at least 3, but probably many more for
accuracy); (2) Each sensor must be within the convex hull of the
landmark graph, to ensure it is within some beacon triangles.  There
are few real sensor networks to point to to validate the realism of
these requirements, but it seems plausible that sensornets might exist
in situations where (1) and (2) are difficult to satisfy (e.g. the
region to be sensed does not take the form of a "walled garden").


***********************************************************************************

Yan-jun Sun: 
                                                                                
Abstract:
                                                                                
The authors present a range-free locaolization scheme employing
the location and signal strength information received from the
sparsely distributed anchors. Performance comparison with other
well-know range-free localization algorithms are conducted in
realistic settings, and investigate the effect from the location
errors to the protocols and application on top of then.
                                                                                
Assumption:
1. TOA(GPS), TDOA, AOA, RSSI based schemes are not cost efficient.
2. Coarse location estimation is good enough
3. signal strength decreases monotomically as the distance increases
                                                                                
Algorithm's main idea:
1. "subdivision based" location estimation, using triangles instead
of polygon to increase the precision
2. use signal strength to find smallest triangle
                                                                                
--


***************************************************************************