Goal-Oriented Scheduling in Sensor Networks with Application Timing Awareness

Taking inspiration from linguistics, the communications theoretical community has recently shown a significant recent interest in pragmatic , or goal-oriented, communication. In this paper, we tackle problem of pragmatic communication with multiple clients different, and potentially conflicting, objectives. We capture goal-oriented aspect through metric Value Information (VoI), which considers estimation remote process as well timing constraints. However, most common definition VoI is simply Mean Square Error (MSE) whole system state, regardless relevance for specific client. Our work aims to overcome limitation by including different summary statistics, i.e., value functions separate clients, diversified query on client side, expressed fact that applications may request state at times. A query-aware Deep Reinforcement Learning (DRL) solution based statically defined can outperform naive approaches 15-20%.