Scholarlink

<p>This study explores the use of Machine Learning (ML) to reduce the GPS errors, which are common in dense environments with obstacles like mountains and buildings. The errors observed in GPS positioning are minimized by the application of Machine learning algorithms on the data collected by a GPS receiver mounted on a vehicle. In this study, we applied different machine learning algorithms on the GPS data of the NH48 stretch for 100&#xa0;km from Pune to Satara to detect and predict future geolocation positioning with respect to speed. This kind of data is generally referred to as spatio-temporal data, as it involves both spatial (latitude and longitude) and temporal (time and speed) components. The 100&#xa0;km route from Pune to Satara is selected because it influences vehicle movement and GPS performance due to its terrain and traffic density. It provides a realistic and challenging real-world trajectory that helps to evaluate positioning performance under diverse conditions. Out of selected route of total distance of 100&#xa0;km, approximately 22% of total distance is falls under Hilly or Ghat terrain, around 54% distance is open highway and remaining 24% roadway falls under semi-urban or urban sections. Such diverse conditions cause signal obstruction, multipath effects and shadowing, which are critical features taken under consideration for testing GPS inaccuracy. In the proposed approach, we have compared the performance of six machine learning algorithms; Linear Regression, KNN, Decision Trees, Random Forest, ANN, and RNN. These techniques enhance stability, prevent overfitting, and improve prediction accuracy. The RNN algorithm showed the best performance, with an accuracy of 99.89% and minimal loss over traditional GPS methods for critical applications like navigation and tracking.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Machine Learning Based Location Estimation for Analysing GPS Accuracy

  • 期刊论文

  • 发表日期：

    2026年7月17日

查看全文
  • Pankaj P. Tasgaonkar,
  • Rahul Dev Garg,
  • Pradeep Kumar Garg,
  • Poonam Bhosale

摘要

This study explores the use of Machine Learning (ML) to reduce the GPS errors, which are common in dense environments with obstacles like mountains and buildings. The errors observed in GPS positioning are minimized by the application of Machine learning algorithms on the data collected by a GPS receiver mounted on a vehicle. In this study, we applied different machine learning algorithms on the GPS data of the NH48 stretch for 100 km from Pune to Satara to detect and predict future geolocation positioning with respect to speed. This kind of data is generally referred to as spatio-temporal data, as it involves both spatial (latitude and longitude) and temporal (time and speed) components. The 100 km route from Pune to Satara is selected because it influences vehicle movement and GPS performance due to its terrain and traffic density. It provides a realistic and challenging real-world trajectory that helps to evaluate positioning performance under diverse conditions. Out of selected route of total distance of 100 km, approximately 22% of total distance is falls under Hilly or Ghat terrain, around 54% distance is open highway and remaining 24% roadway falls under semi-urban or urban sections. Such diverse conditions cause signal obstruction, multipath effects and shadowing, which are critical features taken under consideration for testing GPS inaccuracy. In the proposed approach, we have compared the performance of six machine learning algorithms; Linear Regression, KNN, Decision Trees, Random Forest, ANN, and RNN. These techniques enhance stability, prevent overfitting, and improve prediction accuracy. The RNN algorithm showed the best performance, with an accuracy of 99.89% and minimal loss over traditional GPS methods for critical applications like navigation and tracking.