Safe Deep Reinforcement Learning
My Ph.D. project aims to explore novel techniques in Safe Deep Reinforcement Learning for adaptive control to prevent fatal states in large-scale systems where some control actions may end up in undesired consequences and dangerous situations. The proposed applications for this research are water distribution networks and aquaponic systems, with the possibility of exploring other application domains, e.g. networked collaborative robots and video-games. This approach proposes novel cooperative in-network data processing techniques and provides robust data analytics, learning and control with human-in-the-loop.