Finding by Counting: A Probabilistic Packet Count Model for Indoor Localization in BLE Environments

TL;DR

This paper introduces a probabilistic packet reception model for BLE in indoor environments, utilizing a Bayesian approach and a novel Monte-Carlo algorithm to improve localization accuracy.

Contribution

It presents a new quadratic probabilistic model for BLE packet reception and a Bayesian parameter estimation method, enhancing indoor localization precision.

Findings

Average localization error of ~1.2m

53% improvement over baseline algorithms

Effective in various spatial configurations

Abstract

We propose a probabilistic packet reception model for Bluetooth Low Energy (BLE) packets in indoor spaces and we validate the model by using it for indoor localization. We expect indoor localization to play an important role in indoor public spaces in the future. We model the probability of reception of a packet as a generalized quadratic function of distance, beacon power and advertising frequency. Then, we use a Bayesian formulation to determine the coefficients of the packet loss model using empirical observations from our testbed. We develop a new sequential Monte-Carlo algorithm that uses our packet count model. The algorithm is general enough to accommodate different spatial configurations. We have good indoor localization experiments: our approach has an average error of ~1.2m, 53% lower than the baseline range-free Monte-Carlo localization algorithm.