Unknown Dynamic

Research on unknown dynamic systems focuses on learning and predicting the behavior of systems whose governing equations are not explicitly known, aiming to build accurate predictive models and effective control strategies. Current efforts concentrate on developing data-driven methods, employing architectures like neural ordinary differential equations, Gaussian processes, Koopman operators, and transformers, often incorporating physics-informed constraints or control barrier functions for safety and robustness. This research is crucial for advancing control in robotics, autonomous systems, and various scientific domains where precise models are unavailable or computationally intractable, enabling safer and more efficient operation in complex environments.