Pasar al contenido principal

Resultados relacionados

Resultados totales (Incluyendo duplicados): 3
Encontrada(s) 1 página(s)

Dipòsit Digital de Documents de la UAB
Publicación Capítulo o parte de un libro (BookPart)

RFID-based Soil Moisture Sensor for Smart Agriculture, a Gaussian Mixture Model Approach

  • Benelmeki, Nedal M.|||0009-0003-1955-5006
  • Diaz Machado, Elvis|||0000-0002-6583-8547
  • del Rio Toledano, Javier
  • Morell, Antoni|||0000-0003-2249-8594
  • Lopez Vicario, Jose|||0000-0002-3574-4697
In this work, we present an RFID-based indirect soil moisture sensor based on the application of Machine Learning. More specifically, we suggest an unsupervised approach that does not require information about the real height and moisture levels. This approach can be of great interest in practical agricultural deployments, where the careful deployment of tags at specific depths within the soil is challenging. It allows an estimation of the posterior probability of moisture, based on the available Received Signal Strength Indicator (RSSI) and phase. The suggested method enables the RFID system to operate as a sensor by probabilistically quantifying measurement uncertainty, which is a key distinction from existing ethodologies. In this paper, we focus on two differentiated moisture cases to show the validity of our approach. Future research will extend the proposed methodology to a wider set of moisture levels.
Dipòsit Digital de Documents de la UAB
Publicación Capítulo o parte de un libro (BookPart)

Detección Robusta de Tags RFID usando Hidden Markov Models

  • del Rio Toledano, Javier
  • Lopez Vicario, Jose|||0000-0002-3574-4697
  • Morell, Antoni|||0000-0003-2249-8594
Hidden Markov Models (HMMs) have proven to be powerful tools for modeling sequential data across various applications. This paper explores their use in RFID-based tag detection, focusing on improving accuracy and robustness in environments affected by signal noise and interference. We propose a methodology that leverages the probabilistic nature of HMMs to classify and predict RFID tag readings, thereby enhancing detection reliability. The effectiveness of the approach is evaluated through simulations, demonstrating its potential for improving modern RFID reader performance. Our results indicate that HMMs can significantly improve detection precision, providing a more reliable solution for RFID applications. Additionally, we analyze the impact of model parameters and propose optimizations to enhance performance in dynamic scenarios. This research is a first step towards improved RFID systems by integrating probabilistic model.
Dipòsit Digital de Documents de la UAB
Publicación Capítulo o parte de un libro (BookPart)

Preliminary Study on RFID Signal Characteristics Above Water using Probabilistic Modeling with Sparse Gaussian Processes

  • Diaz Machado, Elvis|||0000-0002-6583-8547
  • Benelmeki, Nedal M.|||0009-0003-1955-5006
  • del Rio Toledano, Javier
  • Morell, Antoni|||0000-0003-2249-8594
  • Lopez Vicario, Jose|||0000-0002-3574-4697
This paper presents a preliminary investigation into the relationship between UHF RFID signal characteristics (Received Signal Strength Indicator- RSSI, and phase) and the vertical distance between the reader antenna and a water surface. Due to experimental constraints preventing direct soil moisture measurement, the study was adapted to use a 50liter container filled with water, varying the reader height and transmission power. Initial manual data inspection revealed an apparent linear relationship between the signal phase and reader height, as well as between phase and the volume of water added (affecting water level). We outline the modified experimental methodology and propose the use of Feedforward Neural Networks (FFNN) for baseline modeling and Sparse Gaussian Process Regression (SGPR) to model the observed relationships and quantify measurement uncertainties efficiently. This work serves as a foundational step towards developing robust RFID-based moisture sensing techniques, providing a methodology for controlled experimentation and data analysis using scalable probabilistic machine learning models.