Multi-Sensor Fusion for Localization: Comparative Study of Kalman Filter and Machine Learning Models

Multi-sensor fusion is critical for achieving precise localization in autonomous driving and robotic navigation. This study compares the traditional Kalman Filter (KF) with machine learning models (Linear Regression, Decision Tree, and Random Forest) for integrating data from Inertial Measurement Units (IMU), Global Navigation Satellite Systems (GNSS), and Light Detection and Ranging (LIDAR). Using a dataset of 8734 samples, this study assess single-, dual-, and triple-sensor configurations, especially focusing on Z-axis prediction challenges due to its small magnitude and nonlinear characteristics. This paper compares models based on Z-axis prediction accuracy using MSE, MAE, and R<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> metrics, as well as the runtime efficiency. Experimental results demonstrate that Random Forest achieves the best performance, particularly in triple-sensor settings, and significantly outperforms Kalman Filter on Z-axis prediction.