Localization

Channel Impulse Response based Fingerprinting Localization

Objective

  • Investigate more accurate indoor localization based on fingerprinting.
Issues
  • Noise and interference.
  • EM scattering.
Approach
  • Extract Channel Impulse Response (CIR) features continuously from physical layer.
  • Derive special Euclidean-type similarity metric to find best match in the database of fingerprints.

SparseTrack: Dead-reckoning with Sparse Range Corrections

Objective

  • Develop hybrid localization based on dead-reckoning (DR) with automatic range corrections in sparse reference infrastructure
 
Issues
  • Error accumulation with DR techniques.
  • Costly finger-printing techniques or limited resolution
Approach
  • Pervasiveness of multi-sensor mobile devices for localization with DR technique.
  • Overcome cumulative DR error with automatic range correction from sparse localization infrastructure.
  • Augment with map information for greater accuracy.

Cluster-based Localization in Ad Hoc Networks

Objective
  • Develop a cluster-based localization technique that leverages on cluster structure and availability to limited anchor points.
Issues
  • Generally not possible for all nodes to know their locations.
  • With known accurate anchors, to what extent can this be used to determine position of other nodes
Approach
  • Nodes organize into clusters.
  • Some nodes are anchors.
  • 2-phase location estimation:
  • Head nodes estimate member nodes’ positions based on uploaded info from member nodes
  • Estimates shared with cluster heads and member nodes for refinement

Multimodal Location Sensing Fusion

Objective
  • Develop a scalable system that fuses multiple modalities of location sensing to provide the most accurate and relevant location information.

Issues

  • Disparate indoor and outdoor localization.
  • Inconsistent location metadata formats.
  • Inconsistent coordinate systems.
  • Adaptable context at scale.
Approach
  • Develop layered approach to location information analysis and modeling.
  • Adopt a Bayesian filtering approach to extract desired location context based on availability of multiple location inputs.