Neural Network control with stability and performance guarantees 

This workshop deals with novel design methods allowing us to ensure closed-loop guarantees, such as stability and performance, in control approaches based on Neural Network (NN) policies. These guarantees are crucial as NNs become more and more integrated into control engineering, raising concerns about their reliability and predictability during and after training. The workshop first delves into innovative methodologies for designing NN controllers with stability guarantees through Linear Matrix Inequalities (LMI), highlighting the opportunities and challenges in their implementation. The second part of the workshop focuses on scalable methods and unconstrained optimization approaches for learning NN controllers that are guaranteed to maintain closed-loop properties during the training phase, with a focus on various kinds of closed-loop stability and distributed scenarios. The envisioned outcome of the workshop is for participants to gain insights into merging advanced machine learning techniques with traditional control systems. A final session and discussion will highlight open challenges and promising research directions in the field.