ASiT-CRNN: A Method for Sound Event Detection with Fine-Tuning of Self-Supervised Pre-Trained ASiT-based Model

This paper presents a practical predefined time control approach for nonlinear multi-agent systems with input saturation. A novel predefined time command filter, constructed using a hyperbolic tangent function, effectively addresses the issues of computational explosion problems and control singularity commonly observed in traditional backstepping designs. To enhance control precision and robustness, a filter compensation signal is introduced to mitigate errors caused by the command filter. Furthermore, a new injective and differentiable function is developed by integrating the properties of horizontal and oblique asymptotes, providing an effective solution for managing input saturation. Leveraging this foundation, a predefined time command filtering controller is designed by combining adaptive backstepping control with the multi-dimensional Taylor network technique, which significantly simplifies the controller design process. Utilizing practically predefined time stable theory, the boundedness of all closed-loop system signals is rigorously established, and it is proven that synchronization errors converge to a small neighborhood around the origin within the predefined time. Finally, simulation results validate the proposed method’s effectiveness and practical applicability.

NONLINEAR DYNAMICS