During my masters program I worked on learning based algorithms for large-scale networked control systems. We developed DIRA a scalable Deep reinforcement learning based Iterative Resource Allocation algorithm to schedule communication resources in a control sensitive manner [arxiv].

Modern networked control systems contain various types of sensitive data. Ideally, control methods should ensure continuous encryption of process data within a control loop. We showed how this can be achieved using partial homomorphic encryption, see [1], [2].