Parallel grid search-enhanced support vector regression for small unmanned helicopter modeling

Purpose Modeling small unmanned helicopters is particularly challenging due to their nonlinear, strongly coupled and complex dynamic behaviors. Therefore, this paper proposes a modeling method to improve the accuracy of the small unmanned helicopter model. Design/methodology/approach This paper introduces a novel modeling methodology based on support vector regression (ϵ-SVR), a robust regression technique tailored for continuous numerical prediction. By incorporating the ϵ-insensitive loss function, the ϵ-SVR approach enhances noise resilience and achieves superior generalization, thereby ensuring high-precision regression results. To further refine the parameter optimization process, a grid search algorithm is augmented with parallel computing, which significantly improves both optimization efficiency and model generalization. Findings Flight test data from a self-developed small unmanned helicopter are utilized to train the model, resulting in an accurate mathematical representation. Experimental findings validate that the parallel grid search-optimized ϵ-SVR model delivers remarkable prediction accuracy and generalization across diverse flight scenarios. Originality/value In this paper, a new modeling method called support vector regression (ϵ-SVR) is applied to the modeling of small-scale unmanned helicopter, which enhances the ability to recover from noise and ensures high precision regression results. Combined with the parallel computing ability of the grid search algorithm, the parameter optimization process is further refined, and the optimization efficiency and the generalization ability of the model are significantly improved.