Data Driven Control

Data-driven control uses machine learning to design controllers without relying on explicit system models, aiming to improve performance and adaptability in complex systems. Current research emphasizes developing robust and explainable methods, exploring architectures like neural networks, decision trees, and extended dynamic mode decomposition (EDMD), often incorporating techniques like reinforcement learning and constrained optimization to address challenges such as safety and scalability. This approach holds significant promise for applications across robotics, energy management, and other domains requiring efficient and adaptable control strategies in the face of uncertainty and high dimensionality.