Page 185 - Kaleidoscope Academic Conference Proceedings 2024

Innovation and Digital Transformation for a Sustainable World

habad, a complex space with furniture, equipment, and
computers. We placed a single access point at the centre
of the laboratory and mounted one receiver, an ESP32
module and a smartphone, on a tripod. We captured the
data in two ways: reference points, which are the training
dataset, and test points, which are the validation dataset
[29].
To collect the data, we recorded data values for 2
minutes in four directions (right, left, up, and down) for
each of the 32 reference points. For the test data points,
we recorded for 1 minute in four directions for 90 data
points. The entire data collection process took approxi-
mately 15 hours. After data analysis, it was observed that

Fig. 1: Laboratory floor plan

(a) Data Analysis of Testing Dataset of a Smartphone

Fig. 2: Dataset recording Smartphone mounted on the
tripod
(b) Data Analysis of Training Dataset of a Smartphone
Shadowing (LDPLS) model, which models RSS noise as Fig. 3: Data Analysis of Smartphone Datasets
a Gaussian distribution, is used to compute these proba-
bilities. Additionally, the Design Rule (DR) is employed the multi-path effect causes large fluctuations in the RSSI
as a decision-making framework to iteratively combine signals. This effect occurs when Wi-Fi signals bounce off
information, thereby enhancing the overall believability multiple surfaces before reaching the receiver. Various
of the localization estimate. objects, such as furniture, moving people, glass walls,
and lab equipment, can also interfere with Wi-Fi signals.
III. DATA PREPARATION
IV. THE PROPOSED ALGORITHM
In general, an indoor localization system based on
RSSI comprises three stages of operation: dataset prepa- In offline processing, preprocessed fingerprints are
ration, offline processing, and online processing. used to train the algorithm. The algorithm identifies
patterns in fingerprints and maps them to known labels
A. Data Preparation
to generalise the relation between the fingerprints and
In this section, we will discuss the data preparation. labels. We have proposed a three-step process to estimate
We recorded the data in the CordIoT lab at IIIT Alla- the Region of any object in Indoor localisation.

– 141 –