LoRa Indoor Localization Based on Improved Neural Network for Firefighting Robot

Abstract

Trapped occupants’ safety is a critical problem in the fireground and a major issue is the lack of reliable indoor localization decision-making system for firefighting. State of the art methods have failed to provide an automatic, accurate and reliable solution that can facilitate the decision-making of incident commanders. This paper aims to develop a novel smart firefighting robot to achieve this goal, by combining artificial neural network with received signal strength indication of the new wireless communication approach named Long Range (LoRa). Our solution includes a new indoor localization algorithm that contains a process for optimizing the initial weights and thresholds of BP neural networks. The solution can improve the location accuracy of trapped occupants in fire. We fully implement the algorithm in a complete indoor localization system and conduct experiments in the space of 25 m $\times$ 25 m $\times$ 5 m that involved a firefighting robot and some trapped occupants. The localization results demonstrate that our solution greatly shortens the convergence time and reduces the average and minimum location error to 0.7 m and 0.2 m respectively in a 20 m $\times$ 15 m testing area.

This work is funded by the National Key Research and Development Plan of China (2017YFE0112200) and European Commission Marie Skłodowska-Curie SMOOTH project (H2020-MSCA-RISE-2016-734875).