Input to State

"Input-to-state" (I/S) research focuses on understanding and guaranteeing the stability of dynamical systems, particularly those modeled using neural networks, in the presence of external inputs and disturbances. Current research emphasizes developing models and control algorithms (e.g., model predictive control, coupled oscillator networks, and various recurrent neural networks like GRUs and LSTMs) that possess provable I/S stability properties, such as input-to-state stability (ISS) and its variants. This work is crucial for ensuring the safety and reliability of AI-driven systems in applications like robotics, autonomous vehicles, and process control, where robustness to uncertainties is paramount.