<p>The notable increase in computational power and the advancement of sophisticated numerical techniques have significantly elevated the role of data-driven approaches in system modeling. In the analysis of nonlinear systems, attention is increasingly shifting from traditional modeling and black-box frameworks toward methods that identify interpretable and structured models directly from real-time experimental data. This work presents a novel recursive least squares (RLS) algorithm for the online identification of nonlinear equations in state space as the system evolves. Unlike neural network-based or input-output models, the proposed approach preserves the explicit structure of the state space, facilitating physical interpretability and seamless integration into state estimation and control design. To promote simplicity, practical applicability, and model robustness, sparse identification techniques are integrated into the RLS framework. Numerical simulations demonstrate the algorithm’s accuracy, noise robustness, and computational efficiency, confirming its effectiveness for identifying nonlinear systems in real time and its relevance to modern control applications.</p>