Please use this identifier to cite or link to this item: http://hdl.handle.net/10609/146651
Title: Estudio y evaluación de algoritmos de procesamiento de datos inalámbricos para la estimación de la localización y ocupación en interiores
Author: Molina Abril, Ginés
Tutor: Torres Sospedra, Joaquín
Others: Lozano Bagén, Antonio
Martínez Sala, Alejandro
Keywords: Indoor Positioning Systems (IPS)
Wi-Fi Fingerprinting
MLOps
K-means Clustering
K-nearest neighbors (k-NN)
Support vector machine (SVM)
Continuous Training (CT)
ETL Pipeline
IoT infrastructure
Concept drift
MLFlow
Multinomial logistic regression
Issue Date: 5-Jun-2022
Abstract: Humans spend most of their time indoors, and GPS has its limitations. Indoor positioning systems, or IPS, have been developed to address these limitations. These systems can leverage existing Wi-Fi or Bluetooth Low Energy (BLE) infrastructures to provide service at very low cost, with low power consumption, greater flexibility and greater device compatibility. These systems can also function as heterogeneous IoT systems using both techniques and obtain great results, enhancing their advantages to mitigate the problems associated with any indoor wireless system such as interference. It is proposed the study and evaluation of algorithms of location (position and area level) and estimation applied to different scenarios using the methods described above. The final objective is the development of an end-to-end production tool capable of predicting the location of a device through its fingerprint. The technical debt is the effect that a decision in the design stages can have in terms of generating a greater maintenance or work effort in the future. For this reason, and following the MLOps approach it will take in consideration the decisions of design that are aimed at offering the possibility of scaling the number of system components, greater versatility and monitoring tools that allows the evolution of the different model architectures and, what is known as Continuous Training (CT). This concept is coupled with those already known as Continuous Integration (CI) and Continuous Delivery (CD). The proposed system should be easily integrated in constant data flow analysis through IoT systems and may evolve to more complex architectures such as Transfer Learning or Incremental Learning architectures.
Language: Spanish
URI: http://hdl.handle.net/10609/146651
Appears in Collections:Bachelor thesis, research projects, etc.

Files in This Item:
File Description SizeFormat 
ginesmoliTFM0622.pdf7,45 MBAdobe PDFThumbnail
View/Open
Share:
Export:
View statistics

Items in repository are protected by copyright, with all rights reserved, unless otherwise indicated.