Localization Techniques Ch. 9 other material

1
Localization Techniques(Ch. 9 other material)

• Prof. Sunggu Lee
• EE Dept., POSTECH

2
Introduction

• Localization (synonym location discovery)
• Has many uses in many application fields
• Navigation, context-aware ubiquitous computing,
  power control in smart buildings, routing in
  MANETs, sensor networks, etc.
• The need for location discovery research
• GPS-based location discovery cannot be used
  inside buildings
• Some applications (such as sensor networks)
  require very low cost devices GPS too expensive
• Different location accuracy requirements -
  approximate locations may only be required in
  some applications

3
Location Discovery Overview

• Meaning of location
• the position of an object in physical space
  with respect to a specific frame of reference
  Basagni 2004, p. 233
• GPS provides latitude, longitude and altitude
  info
• In 2-D space, a (0, 0) origin and x and y
  coordinates are sufficient
• Phases of location discovery
• Measurement phase
• Produces a set of angle and/or distance
  measurements from a set of anchor points
• Combining phase
• Uses an aggregrate set of measurements to produce
  final location estimates
• Accuracy of location estimates is dependent on
  algorithm used and accuracy of measurement
  techniques used

4
Measurement Technologies

• Most common measurement methods
• Received signal strength, time-based, directional
• Types of signals radio, acoustic or optical
• Inertial techniques also sometimes used
• Start from known location and use velocity info
• Signal strength based methods
• Uses signal attenuation with distance
• Cannot be used for accurate distance measurements
• large variations in signal attenuation in
  different environments
• Multipath and shadowing effects (refer to Fig.
  2.1, p. 51)
• Signal strength based methods typical used to
  determine proximity (close to signal source) or
  used in conjunction with other methods

5

• Time-based methods
• Measures distance by recording the time of flight
  (ToF) of a signal
• Requires accurate clock synchronization of
  receiver and transmitter or round-trip
  measurements
• Alternative method use two signals with
  different propagation speeds
• Measure difference between arrival times of two
  signals
• Example system Active Bat from ATT Cambridge
  research labs
• Directional methods
• Can use angle of arrival (AoA) or direction of
  arrival (DoA) information to compute locations

6
Geometric Algorithms for Location Discovery

• Refer to Figure 9.1, p. 257
• If distances are used, gt 3 noncollinear
  measurements to landmarks are required
• However, note that even 2 measurements are
  sufficient if there are other sets of
  interdependent measurements (refer to paper)
• If angular measurements are provided, two
  reference points are sufficient
• With respect to a local frame of reference,
  trigonometric relationships can be used

7
Sources of Measurement Errors

• Multipath fading and shadowing
• With radio signals, up to 30-40 dB variation over
  distances on order of ½ wavelength
• Scattering near receiver affects angle-of-arrival
  (AoA) measurements
• With time of flight (ToF) measurements, multipath
  fading results in peak correlation shifts
• Nonline-of-sight
• Affects both AoA and ToF measurements
• Multiple-access interference
• Transducer calibration errors
• Exacerbated (increased) with the use of low-cost
  radios
• Fluctuations in signal propagation speeds
• Common in acoustic measurements

8
Multipath Fading Effects
Figure 2.1 of Basagni 2004
9
Received Signal Strength (RSS) Fluctuations
Fig. 9.11 of text
Fig. 9.10 of text
10
Dedicated and Nondedicated Localization

• Dedicated localization
• Uses a set of sensors installed at known
  locations within an area of interest
• Location of a mobile host (MH) is determined by
  combining the info from sensors that detect the
  MH
• Example active badge Ref 1 of text system
  uses IR as the measured medium
• Nondedicated localization
• Mostly based on triangulation or trilateration
• Terms are different text is misleading in this
  regard
• Material in text is mostly concerned with using
  Received Signal Strength (RSS) and overcoming
  difficulties posed by distance measurement
  inaccuracies based on RSS
• Additional material in this ppt file and in the
  reference paper

11
Radio Map

• Build a signal strength model and estimate the
  location of a MH based on the model
• Refer to Figure 9.4 of text
• Nearest Neighbors
• Nearest neighbor in signal space (NNSS) technique
  Ref 2 in text
• Calculates distances based on each RSS recorded
  in the radio map uses least squares estimation
• Smallest polygon
• Probabilistic methods

12
Atomic Multilateration

• Use distance measurements to a set of landmarks
  (anchors, beacons)
• Once the location of one node is determined in
  this manner, it can be used as a landmark for
  later position estimates
• Results in a sequential, iterated approach
• Given 3 or more measurements, position estimates
  must be checked for consistency
• Since many measurements will be imprecise, use
  minimum least squares estimation or other method
  to derive location estimates
• Radio map (Sec. 9.2.2), nearest neighbors,
  smallest polygon, probabilistic

13
Iterated Trilateration

• Trilateration ordering
• Order vertices v1, v2, v3, , vn such that v1, v2
  and v3 are completely connected and each vi (i gt
  3) is adjacent to at least 3 vj (j lt i)
• Iterated trilateration
• An initial set of 3 nodes used to define a
  coordinate system (local frame of reference)
• At each stage of algorithm, there are localized
  nodes and unlocalized nodes
• Iteratively compute locations for nodes with
  measurements to three or more previously
  localized nodes
• Iterated trilateration is sub-optimal
• There are many localizable networks that cannot
  be localized in this manner

14
Bilateration

• Bilaternation ordering
• Order vertices v1, v2, v3, , vn such that v1 and
  v2 are connected and each vi (i gt 2) is adjacent
  to at least 2 vj (j lt i)
• Bilateration graph
• A graph with a bilateration ordering
• Example a wheel graph (Figure 2 in paper)
• Authors of paper claim that an iterative
  localization algorithm based on bilateration can
  localize many more networks (especially sparse
  networks) than trilateration-based algorithms

15
Sweeps Algorithm

• Example iterated procedure (basic shell Sweeps)
  shown in Figure 3 of paper
• Note that several nodes have several (finite)
  possible locations
• The number of possible locations for a node
  increases exponentially if each consecutive node
  only has measurements to two previously computed
  nodes
• By changing the ordering, the number of possible
  locations for each consecutive node can be
  dramatically reduced
• In the shell Sweeps phase, reorder nodes so
  that at each stage, all nodes having distance
  measurements to at least two already swept nodes
  are placed earlier in the ordering than other
  nodes. paper, page 4
• Sweeps algorithm (Fig. 4) finitely localizes all
  nodes in bilateration networks
• Very promising simulation results shown in
  Section 5 of paper

16
Ad Hoc Techniques for Location Discovery

• Desirable to develop ad hoc location discovery
  methods that work well in the dynamic
  environments provided by mobile ad hoc networks
• Motivations for development of ad hoc
  localization techniques
• Randomly deployed nodes (e.g., sensor networks)
• Rapid infrastructure installation
• Localization in the presence of obstacles in
  highly dynamic environments
• Desirable characteristics of location discovery
  algorithm
• Should be computationally lightweight
• Should be executable in a fully distributed
  manner
• Should be able to execute even in the presence of
  faults

17
Existing Ad Hoc Localization Approaches

• GPS-less low-cost outdoor localization system for
  very small devices
• Assumes a set of predeployed location-aware
  reference nodes
• A node with an unknown position localizes itself
  using proximity (closeness) sensor info
• Accuracy of location depends on density of
  reference nodes and their transmission ranges
• Best results obtained with a mesh pattern
• Convex position estimation in wireless sensor
  networks
• Concept of a convex hull of a set of points
• Localization works best when anchor nodes are
  located on the convex hull of a set of nodes
• Algorithm computes locations of nodes by
  performing computation at a central point in the
  network
• Location estimation formulated as linear program
  problem

18

• Convex optimization considers either radial
  constraints or angular constraints
• E.g, overlapping sensing range circles
• According to simulation results Basagni 2004,
  when the anchor nodes are on the perimeter of the
  network, positions accuracies between 0.72 to
  0.64R are possible (R is transmission radius)
• GPS-free positioning in mobile ad hoc networks
• Uses radio time-of-flight (ToF) measurements
• Even with measurements errors, mobile nodes with
  speeds up to 20 m/s can be supported with
  adequate location accuracies for location-aided
  routing
• Algorithm forms local coordinate systems for each
  node and then merges them to construct a global
  coordinate system

19

• Locating tiny sensors in time and space
• Uses two types of nodes
• One type is small, cheap and computationally
  limited
• Second type is larger, faster and more expensive
• Act as bases for the smaller nodes
• Localization subsystem has three components
• Wideband acoustic ranging system
• Uses fine-grained time synchronization and ToF
  measurements of acoustic signals between pairs of
  nodes
• Wavelengths range from 0.01 to 1 m
• Received signal is compared to locally generated
  signal and correlation performed
• Example of correlation shown in Figure 8.6 (p.
  244)
• Local coordinate system algorithm
• Location service

20

• Self-localization method for wireless sensor
  networks
• sensor node positions and orientations can be
  estimated using signals from acoustic sources
  with unknown locations. Basagni 2004, p. 245
• Sensor nodes detect acoustic signals and
  propagate ToA and DoA info to central unit
• Central unit fuses all such info using a maximum
  likelihood estimator to determine location and
  orientation of sensor nodes
• Ad hoc positioning system
• Estimates locations by considering distances to
  landmarks
• Three alternative propagation methods used
• DV-hop, DV-distance and Euclidean

21

• In DV-hop method, hop-distances are used
• Landmarks compute average hope distances between
  themselves
• Nodes use average hop distances to each of the
  landmarks to estimate distances to landmarks
• Corrections are propagated in the network using
  controlled flooding
• In DV-distance method, received radio signal
  strength is used to measure distances
• Euclidean method uses true distance measurements
  to a landmark
• Iterative trilateration or bilateration type of
  approach

22

• Robust positioning algorithms for distributed ad
  hoc wireless sensor networks
• Uses inter-node distance measurements and a set
  of anchor nodes
• Works in two phases startup and refinement
• In startup phase, hop info (to each anchor node)
  is used (similar to previous approach)
• In refinement phase, estimated distances between
  one-hop neighbors are used to compute more
  accurate location estimates
• Works in an iterative manner
• Two main problems of this approach are error
  propagation and hard network topologies
• Ad hoc localization system
• Works on the basis of collaborative
  multilateration
• Nodes can estimate their locations even when
  beacon nodes are found multiple hops away
• Works with both centralized and distributed models

23
Future Research

• Physical layer controlling measurement error in
  different environments
• Fusion of measurements from orthogonal sensing
  methods
• Sensor transducer calibration at the network
  level
• Integration of localization systems and protocols
  with other protocols and applications
• Better scalable localization algorithms that work
  well even with imprecise measurements
• Context-aware ubiquitous computation and
  communication based on estimated node locations
• Security and privacy with respect to localization
  info