LOS/Multipath/NLOS Classifiers using Machine learning and Raytracing - A preliminary study to identify and address the Multipath error

Aug 5, 2026

Upload by Levi

This thesis explores applying machine learning algorithms to mitigate one of the key localization challenges in GNSS: multipath errors. Data is collected with a receiver mounted on an excavator placed in front of a building, used as a multipath source. Ray-tracing simulation in MATLAB supports supervised learning by providing labels for the recorded measurements. A novel feature based on the difference in range acceleration between code and carrier signals is proposed, showing promising separability relative to ray-tracing results. The study evaluates SVM performance (~75% average recall and precision), feature learning via an autoencoder with 1D-CNN layers, and a GRU-based deep neural network to convert multiclass into multi-label multi-output classification.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages117

Reading Time4 h 55 min

AccessFree to download

Abstract

Methodology and Data Collection

Feature Engineering and Supervised Learning

Model Experiments (SVM, Autoencoder-1D-CNN, DNN-GRU)

Problem Transformation: Multiclass to Multi-label Multi-output

Keywords

LOS/Multipath/NLOS Classiﬁers using Machine learning and Raytracing A preliminary study to identify and address the Mulitpath error Master’s thesis in Complex Adaptive System KARTHIK UPENDRA DEPARTMENT OF SPACE, EARTH and ENVIRONMENT CHALMERS UNIVERSITY OF TECHNOLOGY Gothenburg, Sweden 2024 [www.chalmers.se](www.chalmers.se) Master’s thesis 2024 LOS/Multipath/NLOS Classiﬁers using Machine learning and Raytracing A preliminary study to identify and address the Mulitpath error KARTHIK UPENDRA Department of Space Earth and Environment Division of Onsala Space Observatory Research Unit of Space Geodesy and Geodynamics Chalmers University of Technology Gothenburg, Sweden 2024 LOS/Multipath/NLOS Classiﬁers using Machine learning and Raytracing A preliminary study to identify and address the Mulitpath error KARTHIK UPENDRA © KARTHIK UPENDRA, 2024 . Supervisor: Jan Johansson, Department of Space Earth and Environment Examiner: Jan Johansson, Department of Space Earth and Environment Master’s Thesis 2024 Jan Johansson, Department of Space Earth and Environment Division of Onsala Space Observatory Research Unit of Space Geodesy and Geodynamics Chalmers University of Technology SE-412 96 Gothenburg Telephone +46 31 772 1000 Cover: Discritised Azimuth-Elevation space around the receiver. Diagram not to scale. Typeset in LATEX Printed by Chalmers Reproservice Gothenburg, Sweden 2024 LOS/Multipath/NLOS Classiﬁers using Machine learning and Raytracing A preliminary study to identify and address the Mulitpath error KARTHIK UPENDRA Department of Space Earth and Environment Chalmers University of Technology Abstract This thesis explores the application of machine learning algorithms to address oneof the challenges in localization in GNSS called the Multipath errors. The approach involves data collection via a receiver mounted on an excavator which is placed in front of the building, which acts as one of the sources for mulitpath error. In order to perform supervised machine learning, wireless communication tool box within Matlab is used for Raytracing simulation to label the data. Drawing inspiration from existing literature, we introduce a novel feature, the ’diﬀerence in range acceleration between code and carrier signals,’ which exhibits promising distribution and metrics when analyzed with respect to ray-tracing results. The support vector machine (SVM) achieves an average class-wise recall and precision of approximately 75% on the recorded measurments. Additionally, we explore the use of an Autoencoder with a 1D-CNN layer to extract new features aimed at enhancing classiﬁcation performance. By conducting three diﬀerent simulations with varying data sorting methods, we demonstrate how sorting the data and the machine learning algorithm can inﬂuence the learned features which in turn impacts classiﬁcation performance. Lastly, in order to take advantage of the insights gained from diﬀerent sorting methods, we transform the problem from a Multiclass to Multi-label-Multi-output classiﬁcation problem, wherein we utilize a deep neural network architecture with GRU units to classify the signals. Although complicated in terms of data restructuring and handling, the network demonstrated robust performance, achieving more than 85% average class-wise recall and precision and exceeding 96% for signals labeled LOS. Keywords: Machine learning, Multipath error, GNSS, Convolutional neural network (CNN), Recurrent neural network (RNN), Gated recurrent unit (GRU), 1D CNN, Autoencoders, Raytracing, Deep neural network (DNN) . Acknowledgements I express my sincere thanks to Professor Jan Johansson for his invaluable guidance and belief in my capabilities throughout this thesis. Special appreciation to my family for their unwavering support. I also acknowledge ChatGPT for its role in structuring and reﬁning my thoughts, contributing to the clarity of this work. Additionally, I appreciate the prompt ass

FAQs

Which models and techniques are evaluated for classifying LOS and NLOS-related signals?

The work evaluates an SVM classifier, an autoencoder with a 1D-CNN layer for feature extraction, and a deep neural network using GRU units, including a conversion from multiclass to multi-label multi-output classification.

Aug 5, 2026

Upload by Levi