Predicting Lateral Dynamics of CRJ700 Using Multilayer Perceptron and Support Vector Regression

Accurate estimation of lateral aerodynamic coefficients is essential for improving flight stability and control. This study explores machine learning techniques, specifically multilayer perceptron (MLP) and support vector regression (SVR), to predict the lateral aerodynamic coefficients of a Bombardier Regional Jet CRJ-700. The dataset, obtained from a Level D CRJ-700 Virtual Research Simulator (VRESIM), covers diverse flight conditions. Bayesian optimization was used for hyperparameter tuning. Model performance was validated by comparing predictions with experimental data within FAA tolerance limits. The results show that MLP and SVR achieve lateral prediction errors below 5%, demonstrating high accuracy in estimating lateral aerodynamic coefficients. These findings suggest that AI-based methods can provide reliable aerodynamic models for flight simulation and control system design, reducing reliance on traditional empirical methods.