DRL-based Resource Allocation in Remote State Estimation

Remote state estimation where sensors send their measurements of distributed dynamic plants to a remote estimator over shared wireless resources is essential for mission-critical applications Industry 4.0. Existing algorithms on radio resource allocation systems assumed oversimplified communications models and can only work small-scale settings. In this work, we consider with practical the orthogonal multiple-access non-orthogonal schemes. We derive necessary sufficient conditions under which be stabilized. The are described in terms transmission power budget, channel statistics, plants’ parameters. For each scheme, formulate novel problem as decision-making achieving minimum overall long-term average mean-square error. Both quality states taken into account decision making. systematically investigated problems different schemes large discrete, hybrid discrete-and-continuous, continuous action spaces, respectively. propose action-space compression methods develop advanced deep reinforcement learning solve problems. Numerical results show that our effectively provide much better scalability than literature.