1: \section{Introduction}
2: Sensor networks fundamentally depend on location information for
3: applications, which must bind spatial coordinates with sensed
4: data. In addition, location can be useful for routing,
5: power management, and other services.
6: A localization (or positioning) service enables sensor nodes to
7: acquire their spatial coordinates without having to program and deploy
8: each sensor to a precise location.
9:
10: In some cases position can be obtained by having Global Positioning
11: System (GPS) on each sensor node, however this can have high expense,
12: power cost, and put problematic constraints on deployment.
13: Instead of using GPS, most self-localization algorithms assume
14: only some sensor nodes can obtain their position information directly,
15: and they are called anchors or beacons (by contrast, some algorithms
16: construct relative coordinates only, and may use constraint solving
17: techniques to combine local maps \cite{MLRT04}).
18: Using anchor positions, positions of other sensor
19: nodes will then be calculated. Already there is extensive literature
20: on the general problem of localization in wireless sensor networks,
21: with solutions ranging from specialized ranging assumptions
22: to complex algorithms that minimize overall positioning error
23: \cite{LR03,NN01,SLR02,SPS02,KMSKA04,BY04,SKMKA05,ZC04,BP00,MSD05,GKLP04,HHBSA03,PS05}.
24:
25: The general problem of (self-) localization makes few assumptions about the
26: topology of the network beyond basic communication connectivity and
27: density (both sensor and anchor) constraints: the topology can be
28: arbitrary and sensor deployment could be some random
29: process. Presently, sensors are often expensive enough to justify more
30: careful deployment than a random one; an economical emplacement of
31: sensors for uniform coverage is a regular tiling \cite{DHL04}
32: or grid topology. In \cite{BKA05}, a large-scale wireless sensor
33: grid is implemented and evaluated. Some localization schemes \cite{SS04},
34: \cite{SKMKA05} are also based on the grid deployment.
35: The question driving our investigation is:
36: does the assumption of a grid topology reduce
37: the localization problem to a simple, practically implementable one?
38: In a wireless sensor grid, one has more position constraint
39: information than in a randomly placed network. With this
40: additional information, we hope for a fast localization algorithm
41: with small memory footprint, that could reasonably be run on the
42: current generation of sensor nodes.
43:
44: \textit{Contribution.}
45: In this paper, we investigate a very simple approach to
46: localization for wireless sensor networks in a grid topology.
47: The algorithm consumes little computing power and
48: does not depend on high precision in distance estimation. It
49: can be implemented in a distributed manner at different levels.
50: Despite its almost na\"{i}ve approach to assigning grid positions
51: based on table lookup from hop distance and despite significant radio
52: ranging errors, we find reasonable results
53: (the reader may look ahead to Figure \ref{result} for
54: illustrative results). Our experimental results were obtained
55: in rough and grassy terrain using RSSI for ranging,
56: where acoustic ranging --- the foundation for other
57: distributed localization research \cite{MLRT04} ---
58: has been difficult to use \cite{KMSKA04}.
59:
60: The rest of the paper is organized as follows. Section~\ref{prelim} describes the
61: grid topology and anchor placement. Section~\ref{sec-algorithm} introduces
62: a table lookup algorithm for localization and discusses different
63: implementation possibilities.
64: Section~\ref{sec-experiment} presents experimental results and compares our
65: algorithm to related results.
66: Section~\ref{sec-analysis} provides analysis of
67: the algorithm, and Section \ref{conclusion} contains conclusions and future work.
68: