Optimizing Effectiveness-Aware Information Exchange for Goal-Oriented Semantic Communication

The traditional maximalist approach to designing communication systems is increasingly misaligned with today’s evolving connectivity demands. Cyber-physical systems require a shift toward real-time multimodal communication, autonomous decision-making, and efficient distributed processing. The goal-oriented semantic communication paradigm offers a minimalist alternative by prioritizing the relevance and impact of exchanged information, which enhances effectiveness while reducing resource consumption by acquiring, transmitting, and receiving only the most pertinent information according to the goal of data exchange. Consequently, this paradigm presents a forward-thinking solution to modern communication challenges, balancing effectiveness with efficiency. This thesis targets the challenges of incorporating goal-oriented effectiveness into status update systems, where a sensing side observes a source and transmits status updates to an actuation side. As a foundation, we introduce a framework for evaluating effectiveness at multiple granularity levels and propose a grade of effectiveness metric, which serves as a cornerstone throughout this thesis. We then explore broader challenges such as effect-aware filtering and encoding, the timing and scheduling of updates, and determining which information to exchange. We examine these challenges across three distinct parts. Within the first part, we study push-based update systems where the sensing side selectively transmits updates based on their estimated relevance and timeliness. It employs effect-aware filtering and timely source coding to transmit only high-impact updates and maximize effectiveness while minimizing costs. 

We first address the single-user scenario before extending our approach to multi-user systems with heterogeneous goals. In this context, we propose effect-aware filtering and encoding mechanisms to optimize the effectiveness of communicated updates among multiple sensing and actuation agents. In the second part, we analyze pull-based update systems, where updates are queried based on their relevance to predefined goals. For a single sensing-actuation pair, we design effect-aware query control policies that dynamically adapt to the source’s temporal evolution. This is extended to a multi-agent system where a central hub schedules queries and broadcasts updates to actuators. We develop effect-aware scheduling policies that maximize long-term effectiveness while accounting for risk awareness and cost constraints. Finally, in the last part of this thesis, we propose an integrated push-and-pull model, where both sensing and actuation agents play roles in decision-making regarding the exchange of information while balancing effectiveness and cost. We extend our analysis to a multi-agent multiple access scenario, where sensing agents independently observe a common source and transmit updates over a shared medium. We introduce a goal-oriented self-decision multiple access scheme to maximize long-term effectiveness, deriving activation probabilities and threshold criteria for effective update transmission.