English

A State Space Technique for Wildlife Position Estimation Using Non-Simultaneous Signal Strength Measurements

Signal Processing 2018-05-30 v1

Abstract

A novel state-space technique is presented to estimate the location and airborne movements of VHF tagged wildlife individuals with fixed VHF arrays. The approach combines a movement model (Ornstein- Uhlenbeck random process in the transverse (horizontal) plane and a Cox- Ingersoll-Ross process in the vertical direction) to ensure biologically-consistent trajectories in three-dimensions, and an observation model to account for the effect of range, altitude and bearing angle on the received signal strength. The observation model of received signals accounts for low-end saturation from receiver noise, high-end saturation from receiver non-linearities as well as a wireless multipath phenomena, which modulates the received signal according to range, the altitude and radiation characteristics of the Yagi array. A pattern function for the Yagi array is synthesized that facilitates linearization of the received signals and subsequent application of Kalman filtering. We first validate the model using a simulated trajectory and then estimate the space-time trajectory of a migrating VHF-tagged shorebird, which was tracked with a regional automated radio telemetry network. The algorithm accurately predicted the average movement trajectory given the system parameters and the initial conditions (average error << 1 km). The modeled shorebird track represents a first estimate in three-dimensional (3D) of a radio tagged bird using a fixed telemetry array, and was qualitatively reasonable, but exhibited some sensitivity in the vertical plane and to initial conditions.

Keywords

Cite

@article{arxiv.1805.11171,
  title  = {A State Space Technique for Wildlife Position Estimation Using Non-Simultaneous Signal Strength Measurements},
  author = {Ramakrishna Janaswamy and Pamela Loring and James McLaren},
  journal= {arXiv preprint arXiv:1805.11171},
  year   = {2018}
}

Comments

24 pages