Indoor positioning systems: Smart fusion of a variety of sensor readings

TL;DR

This paper presents a versatile sensor fusion system for indoor positioning, demonstrating that combining RF and IMU sensors yields robust localization across different scenarios.

Contribution

It introduces a reproducible measurement system to compare and fuse multiple sensors, highlighting the robustness of simple raw data fusion methods across varied environments.

Findings

Raw data fusion provides reasonable generalization performance.

RF and IMU sensor combination shows promising robustness.

Some techniques fail to generalize across scenarios.

Abstract

Robust and versatile localization techniques are key to the success of the next industrial revolution. Yet, it is uncertain which combination of sensors will be the most robust and valuable. Thus, we present a versatile and reproducible measurement system incorporating a manifold number of state-of-the art sensors to compare and fuse the raw input data. It is shown that some techniques show very good results on the same scenario and data-set, but fall apart on translating to a slightly different scenario. In general we show that the vanilla approach to fuse the raw data achieves reasonable results in the generalization domain, demonstrating that radiofrequency (RF) localization techniques in combination with an inertial measurement unit (IMU) could result in a very robust and promising candidate for solving this challenging task.