Abstract
This paper addresses query scheduling for goal-oriented semanticcommunication in pull-based status update systems. We consider a system wheremultiple sensing agents (SAs) observe a source characterized by variousattributes and provide updates to multiple actuation agents (AAs), which actupon the received information to fulfill their heterogeneous goals at theendpoint. A hub serves as an intermediary, querying the SAs for updates onobserved attributes and maintaining a knowledge base, which is then broadcastto the AAs. The AAs leverage the knowledge to perform their actionseffectively. To quantify the semantic value of updates, we introduce a grade ofeffectiveness (GoE) metric. Furthermore, we integrate cumulative perspectivetheory (CPT) into the long-term effectiveness analysis to account for riskawareness and loss aversion in the system. Leveraging this framework, wecompute effect-aware scheduling policies aimed at maximizing the expecteddiscounted sum of CPT-based total GoE provided by the transmitted updates whilecomplying with a given query cost constraint. To achieve this, we propose amodel-based solution based on dynamic programming and model-free solutionsemploying state-of-the-art deep reinforcement learning (DRL) algorithms. Ourfindings demonstrate that effect-aware scheduling significantly enhances theeffectiveness of communicated updates compared to benchmark scheduling methods,particularly in settings with stringent cost constraints where optimal queryscheduling is vital for system performance and overall effectiveness.