This model was developed in Python using Pandas and scikit-learn libraries.

Proposed System

Here, in this figure, a traditional method called signal fingerprinting is applied, consisting of two phases: training and matching. The matching algorithm is used during the matching phase. The system demonstrates the sequence of tasks involved in generating and matching fingerprints and the process of populating the fingerprint database with geotagged signatures. In order to develop a model that can estimate the user position coordinates based on the Received Signal Strength (RSSI), I have compared a wide range of regression algorithms.

Data Collection Processing and Visualization

Data preprocessing included shuffling and normalization of attributes to prepare the dataset for algorithmic evaluation. The data used in this study contains about 2557 lines with six different measures from six different Radio Base Stations (RBSs). It has two columns that represent the user's position. Also, two additional datasets are used: one provides more information about the RBSs, such as their positions and the power of the radiated signal, and another is used as a test dataset.

Heatmap showing the correlation between user positions and RSSI values from training dataset

Scatter plot of correlation between the user's position and the RSSI values from all Radio Base Station from training datase

Results

Algorithmic Performance Comparison

Evaluation of K Nearest Neighbor (KNN) (k=3) Model

Distance Error Histogram

Actual vs. Predicted Distance

Cross-Validation

The algorithm uses Leave-One-Out cross-validation to assess the model's accuracy. This cross-validation filters out poorly predicted points and refines the model accordingly.

For Further Development

[1] Open this folder as project folder in any Python IDE
[2] The above mentioned folder has datasates included
[3] Here Python 3.10 was used as based interpreter
[4] To perform simulation, first, create a virtual environment in a blank folder then install the required libraries
[5] Pycharm Professional Edition is Highly Recommended

File Information

Accuracy_Percentage

For calculating percentage result from mean error

Fingerprint_Algorithm

Fingerprint algorithm used for localization

KNN_Evaluation

For evaluating the performance of K Nearest Neighbor (KNN) Model

Leaveoneout_Cross_Validation

This cross-validation filters out poorly predicted points and refines the model accordingly

LTE_ML_Model

The new model is based on this model

Revised_6G_ML_Model

The new model for localization in 6G indoor environment

Test_Model

This was the initial model

Traditional_Models

Traditional models like Sui, OkamuraHata, Ericsson are included in this.

Tranditional_Models_With_ML

Traditional models with machine learning algorithms

References

[1] M. Z. Islam Sifat, M. Mostafa Amir Faisal, M. M. Hossain and M. A. Islam, "An Improved Positioning System For 6G Cellular Network," 2023 26th International Conference on Computer and Information Technology (ICCIT), Cox's Bazar, Bangladesh, 2023, pp. 1-6, doi: https://10.1109/ICCIT60459.2023.10441118
[2] VO, Quoc Duy; DE, Pradipta. A survey of fingerprint-based outdoor localization. IEEE Communications Surveys & Tutorials, v. 18, n. 1, p. 491-506, 2016, doi: https://doi.org/10.1109/COMST.2015.2448632
[3] Pandas. https://pandas.pydata.org
[4] scikit-lean. http://scikit-learn.org