Hybrid IRBM-BPNN Approach for Error Parameter Estimation of SINS on Aircraft

Area of Science:

• * Navigation Systems Engineering
• * Artificial Intelligence in Aerospace
• * Signal Processing and Estimation Theory

Background:

• * Strap-down inertial navigation systems (SINS) are crucial for aircraft navigation but are susceptible to instrument and alignment errors.
• * Accurate estimation of these errors is essential for improving overall navigation precision.
• * Existing methods often struggle to achieve optimal accuracy in complex error forecasting scenarios.

Purpose of the Study:

• * To develop a novel hybrid approach for estimating SINS error parameters.
• * To enhance the navigation accuracy of aircraft by improving SINS error forecasting.
• * To introduce an improved restricted Boltzmann machine BP neural network (IRBM-BPNN) for this purpose.

Main Methods:

• * Analysis of SINS error generation mechanisms and establishment of error models.
• * Utilizing an unsupervised Restricted Boltzmann Machine (RBM) to initialize a BP neural network (BPNN).
• * Developing an IRBM-BPNN model integrating SINS/GPS/CNS data, with enhanced structure and pulse signal input.

Main Results:

• * Simulation experiments demonstrated the feasibility and effectiveness of the IRBM-BPNN method for SINS error parameter forecasting.
• * The hybrid RBM-BPNN approach significantly improved forecast accuracy compared to standalone BPNN.
• * The enhanced IRBM-BPNN structure yielded optimal forecast accuracy for SINS error parameters and navigation accuracy.

Conclusions:

• * The proposed IRBM-BPNN method offers a superior solution for SINS error parameter estimation.
• * Combining RBM and BPNN, along with structural improvements, effectively boosts forecast accuracy.
• * This approach leads to optimal aircraft navigation accuracy, outperforming traditional methods.