Affordable Access

Protected Region Radio Map Estimation

Authors
  • Tivald, Jonathan
Publication Date
Jan 01, 2022
Source
eScholarship - University of California
Keywords
Language
English
License
Unknown
External links

Abstract

Passive radio frequency (RF) sensors and receivers are highly vulnerable to unintended radio interference from deployment of active RF transmitters in nearby areas of service. Often, these RF receivers may also be susceptible to overloading damages. High likelihood scenarios of overloading damages include ultra-sensitive receivers that cannot afford front-end protection, or receivers deployed while powered down without the ability to measure the environment before powering on. It is often costly to measure RF signal strength and assess potential interference over wide urban/suburban areas among various building structures and complex terrains. Moreover, these passive RF sensors and receivers are sometimes deployed in locations that are difficult to access and to measure radio signal strength from new RF transmissions or those under planning. Consequently, it is important in the service planning stage to estimate a wide area radio map from only limited RF measurement at locations of convenience. We propose that a network of cheap and robust RF receivers may be sparsely deployed in a geographical region to estimate a completed radio map. After receiving power measurements from the sparse network of RXs, several different estimation methods may be applied to reconstruct the region’s radio map. These estimation methods may be in the form of kernels, random processes, basis functions, and Machine Learning (ML) algorithms. We aim to provide a certain level of confidence in multiple estimation methods that may be used for estimating a completed radio map. Many of the interpolation methods produced favorable results when estimating a radio map. The Inverse Distance Weighting (IDW) algorithm performed the best overall due to being one of the most accurate estimators, having the fastest processing time, and robust performance with system parameter selection. Overall, the Machine Learning (ML) algorithms processed much faster than the average interpolation method, but performed worse on average. Iterative Shrinkage and Thresholding Algorithm (ISTA) Net performed the best due to estimating the most accurate radio maps.

Report this publication

Statistics

Seen <100 times