Page 183 - Kaleidoscope Academic Conference Proceedings 2024

Dempster-Shafer Theory-based Indoor Region

Prediction with Single Wi-Fi Access Point for

Static Object Localization

Mr. Ritesh Kumar Dr. Vijay Kumar Chaurasiya
Department of Information Technology, IIIT Allahabad Department of Information Technology, IIIT Allahabad
Prayagraj, India Prayagraj, India
Pro2017003@iiita.ac.in vijayk@iiita.ac.in

Abstract—Indoor localisation has gained significant at- such as RFID, Bluetooth (BLE), Ultrawideband (UWB)
tention in recent years due to its applications in various [7], Acoustic signals [8], Wireless Local Area Networks
domains, including robotics, asset tracking, and location- (WLANs) [9], and Cellular-based visible light commu-
based services. Traditional indoor localisation systems rely
on several access points or infrastructure, which are nication (VLC) [10] and, etc.
sometimes costly and challenging to establish. This paper In this technology, two approaches are used for lo-
investigates the viability of using a single access point for calization [3], [6]: geometric, such as Trilateration [11],
the indoor localisation of static objects. Initially, a dataset is and fingerprinting [9].
recorded in the laboratory with ESP32 and a smartphone. However, Infrastructure-independent does not require
After that, the dataset is utilised to train an algorithm that
can infer the user’s location from current RSSI readings. any external hardware to be set up in the environment,
Furthermore, the RSSI single feature is converted into five such as dead reckoning [12], camera-based localization
statistical features: 25%,50%, and 75% of quartiles, as [13], etc. The dead reckoning algorithm is a widely used
well as min and max. The Dempster-Shafer Theory (DST) indoor positioning algorithm [14]. It utilizes the data
is used to classify the different regions. The localisation
algorithm used in this system maps the RSSI values to from the smartphone’s sensors to calculate the user’s
appropriate locations inside the interior environment. The heading and step length, which are then used to esti-
algorithm shows better results than several other machine mate the user’s position. The algorithm’s accuracy relies
learning algorithms. heavily on the Inertial Measurement Unit (IMU) sensors
Keywords: RSSI, Fingerprinting, Dempster-Shafer The- present in the smartphone, namely the accelerometer,
ory, Localization gyroscope, and magnetometer. However, the accelerom-
eter tends to accumulate errors over time, the gyroscope
I. INTRODUCTION
sensor exhibits drift errors, and the magnetometer is
In the Internet of Things (IoT) era [1], indoor lo- susceptible to external disturbances or interference from
calization has garnered attention as a crucial aspect of nearby metals. In the literature, various methods are used
Industries, as it enables efficient monitoring of workers, to compensate for the error of the smartphone sensors.
machines, and logistics, leading to optimal resource uti- However, sensor fusion is widely adopted to reduce
lization [2]. As society has progressed, people’s lifestyles sensor errors, which improves the heading estimation.
have shifted towards spending more indoors - from This paper proposes a method for indoor localisation
their offices to homes or malls and waiting at airport
that utilises Single Access Point WiFi technology. While
lounges for travel [3]. Indoor localization has many
relying on the sensors’ data affects the localisation
applications in smart homes, from old-age care systems accuracy, WiFi can improve it by correcting the user’s
to emergencies, warehouses, hospitals for automation,
position indoors. WiFi is a widely available and afford-
and industries 5.0 [4], [5]. This has led to the utmost
able technology that measures signal strength through
requirement for an efficient indoor localization system. Received Signal Strength (RSSI). However, RSSI signals
Indoor localization technologies can be infrastructure-
have several issues, such as Multipath effects, Attenua-
dependent or independent [6]. Infrastructure-dependent tion, and Non-Line of Sight (NLoS), which can cause in-
technologies require a special type of hardware that
accuracies. The literature [3], [6] shows that fingerprint-
can sense the environment and send it to the receiver,
based localisation has significantly improved with the
help of machine learning and deep learning algorithms.
This work was supported by the Ministry of Education, Government
of India (GoI). The fingerprinting approach is popular and considered a

978-92-61-39091-4/CFP2268P @ITU 2024 – 139 – Kaleidoscope