A Novel KGP Algorithm for Improving INS/GPS Integrated Navigation Positioning Accuracy.

Area of Science:

• Navigation Systems
• Data Fusion
• Machine Learning

Background:

• Multi-source sensor data fusion enhances vehicle navigation accuracy.
• Neural network-based inertial navigation systems (INS) and GPS integration suffer from reduced accuracy due to system nonlinearity.
• Existing methods struggle with nonlinearities and unstable GPS signals, impacting positioning precision.

Purpose of the Study:

• To propose a novel data fusion method, KF-GDBT-PSO, for improving vehicle navigation accuracy.
• To address the weakening generalization abilities of traditional navigation systems in nonlinear environments.
• To enhance positioning accuracy and continuity, especially during GPS signal fluctuations.

Main Methods:

• Developed a Kalman Filter-integrated Gradient Boosting Decision Tree (KF-GBDT) model for INS error compensation.
• Optimized Gradient Boosting Decision Tree (GBDT) prediction accuracy using Particle Swarm Optimization (PSO).
• Applied the KF-GDBT-PSO method to handle vehicle positioning data nonlinearity and correct INS errors during unstable GPS conditions.

Main Results:

• The proposed KF-GDBT-PSO method effectively addresses nonlinearity issues in vehicle navigation.
• The system demonstrated improved positioning accuracy and continuity, particularly when GPS signals were unstable.
• Achieved a 28.20-59.89% increase in positioning accuracy compared to multi-layer perceptron neural networks and random forest regression.

Conclusions:

• The KF-GDBT-PSO fusion method significantly enhances vehicle navigation system performance.
• This approach offers a robust solution for nonlinear systems and unreliable GPS signals.
• The method provides a substantial improvement in positioning accuracy, outperforming existing machine learning regression techniques.