Laboratory Assets Tracking Using Bluetooth Low Energy Technology with Decision Tree Algorithm

Tan, Chee Keong (2023) Laboratory Assets Tracking Using Bluetooth Low Energy Technology with Decision Tree Algorithm. Final Year Project (Bachelor), Tunku Abdul Rahman University of Management and Technology.

Text Full Text -Tan Chee Keong.pdf Restricted to Registered users only Download (2MB)

Abstract

Indoor Positioning System (IPS) has gained great attention from researchers worldwide due to Global Positioning System (GPS) is not capable for accurate indoor positioning. The challenges of this research are the difficulty of tracking indoor objects, the implementation of algorithms and possibility of having low accuracy prediction. Their corresponding objectives are to design and develop an IPS for laboratory asset tracking by using Bluetooth Low Energy (BLE) technology, to perform location prediction through Kalman Filter (KF) and Decision Tree (DT) algorithms, and to evaluate and improve the accuracy of predicted results. The system architecture of this study started from fingerprinted map creation, followed by setup of BLE gateways, data collection, KF filtering and DT classification. The system performs the best with 2m x 2m grid fingerprinted map and the BLE gateways setup at ceiling level. KF parameters are set differently in training and testing phase. In the training phase, R is dynamic, Q and Δt are both 0, while in the testing phase, R is 3.0, Q is 0.1 and Δt is 0. Next, the final DT model used “Entropy” criterion, “Best” splitter and maximum depth of 7. The final system reaches an accuracy of 97.75% for the test dataset.