| Abstract | Location-based services (LBSs) have long been identified as an important component
of emerging mobile services. While outdoor positioning has been strongly established,
systems dealing with indoor positioning in urban environment are still in a nascent stage. The
upcoming LBSs require positioning systems (PS) that are available ubiquitously, which
necessitates the integration of PS available in outdoor environment with PS in indoor
environment. Global Navigation Satellite Systems (GNSSs) such as GPS, GLONASS,
Galileo, and QZSS are some of the prominent systems that provide outdoor positioning.
Indoor Positioning Systems (IPSs), however, are undergoing rapid development, which can
be supplied using short-range wireless technologies such as Wi-Fi, Bluetooth, RFID, and
Infrared. Despite such technologies, the performance of existing IPSs is constrained by the
expensive infrastructure setup of the system while high accuracy availability could still be
limited.
The main objective of this thesis research is therefore to determine the feasibility of installing
indoor positioning system using short wave radio signals in a comparatively large area and
comparing this system to already developed IPSs. Software based integration of IPS and
GNSS which could provide ubiquitous positioning, would also be developed.
The key contributions of the thesis could include:
1. Development of received signal strength indicator (RSSI) collection method taking
into consideration the orientation in which the RSSI could be observed in 3D
environment;
2. Development of a novel Wi-Fi based IPS using cascading artificial neural network
(CANN) for positioning in 3D environment and optimization of the system developed
using genetic algorithm (GA);
3. Showcasing that GA is a feasible technique for optimizing Wi-Fi based IPSs;
4. Development of ubiquitous positioning system (UPS), a client server architecture
using ANN which promotes software based integration of GNSS and IPS; and
5. Demonstrating the result of extensive field study of the performance of UPS and IPS
in terms of accuracy and precision.
The developed client-server-based UPS demonstrated that a GA-optimized cascading ANN
presents a good choice for the development of an IPS, which, along with its integration with
GNSS, provides ubiquitous positioning. The proposed novel approach of optimizing the
cascading ANN using GA offers a considerable advantage over well-documented IPSs and
techniques, as it increases the accuracy of the technique in terms of mean and median
distance error. When a single ANN is compared with the GA-optimized cascading ANN, a
29% decrease in average distance error could be observed.
Optimization using the GA also led to increased precision of the system because with
optimized GA on the one hand, 87% of the distance error was within the range of 0-3 m. On
111
the other hand, with the GA not optimized only 51 % of the distance error was within that
same range.
The UPS displayed a mean distance error of 3.29 m when tested in covered area (CA),
open sky area (OSA), and half open sky area (HOSA).The forwarding block also showed
good results since 89% of the time, it succeeded in activating the positioning module with
better accuracy.
During the deployment and testing of the system, no issues arose regarding its scalability
or latency. Although the latency of the system was slightly higher in the beginning because of
the weight calculation by GA and selection of the positioning module, in the future, a study
could be conducted to quantify the latency of the system. As for scalability, when the study
area was expanded, only new fingerprinting data was added in the main repository at the
server end. The technique could therefore be reasonably scalable, and further study could be
carried out to evaluate in more detail the scalability of the system.
The main advantage of the presented concepts is that the system can be readily deployed
considering that Wi-Fi and GNSS modules are now available in almost all mobile devices.
Thus, on the client side being passive receiver, no additional hardware is required.
|