Data-driven inference of conic relations via saddle-point dynamics

Conic relations of an input-output system are important system properties that can be exploited in order to design robust controllers. Therefore, we study the problem of determining the minimal cone containing such an input-output system. While in applications the input-output relation itself is often undisclosed, input-output data tuples can be sampled. Therefore, we present an iterative sampling approach to determine conic relations of a linear time-invariant system from input-output data. This sampling approach is based on saddle-point dynamics, whose convergence properties are then investigated.