Lian Sang, Cung; Steinhagen, Bastian; Homburg, Jonas Dominik; Adams, Michael; Hesse, Marc; Rückert, Ulrich: Identification of NLOS and Multi-path Conditions in UWB Localization using Machine Learning Methods. In: Applied Sciences. Jg.10 H. 11. 2020
Inhalt
- Problem Description
- Related Works
- Conventional NLOS Identification Techniques in UWB
- Identification of the NLOS and MP Conditions in the Literature Based on Machine Learning Techniques
- Measurement Scenarios and Data Preparation
- Experimental Setup
- Data Collection Process
- Labeling the Measured Data and Dealing with the Class Imbalance Case
- Separation of the Training, Validation, and Test Dataset
- Feature Extraction
- Machine Learning Models for Identification of the LOS, NLOS, and MP Conditions
- Support Vector Machine Classifier for the UWB Localization System
- Random Forrest Classifier for the UWB Localization System
- Multi-Layer Perceptron Classifier for the UWB Localization System
- Section Summary
- Data Preprocessing and Feature Selection
- The Impact of Feature Extraction in the Evaluated Machine Learning Models
- The Impact of Feature Scaling in the Evaluated Machine Learning Models
- Evaluation Results
- Performance Comparison of the Three Classifiers Using the Macro-Averaging F1-Score as a Metric
- Result Representation of the Three Evaluated Classifiers Using the Confusion Matrix
- Comparative Analysis of the Two Test Scenarios for SVM Classifier
- Comparative Analysis of the Two Test Scenarios for the RF Classifier
- Comparative Analysis of the Two Test Scenarios for the MLP Classifier
- Summary of the Experimental Evaluation Results
- Discussions
- Conclusions
- References