Model-Free design of control systems over wireless fading channels

Wireless control systems replace traditional wired communication with wireless networks to exchange information between actuators, plants and sensors in a system. The noise channels renders ideal policies suboptimal, their performance is moreover directly dependent on the way which resources are allocated loops. Proper design of policy resource allocation based both plant states fading then critical achieve good performance. resulting problem co-designing control-aware communication-aware controllers, however, challenging due its infinite dimensionality, existence system constraints need for explicit knowledge network models. To overcome those challenges, we rely constrained reinforcement learning algorithms propose model-free approach systems. We demonstrate near optimality stability using near-universal parametrizations present practical algorithm learn co-design policy. Numerical experiments show strong learned over baseline solutions.