SMART: Situationally-Aware Multi-Agent Reinforcement Learning-Based Transmissions

In future wireless systems, latency of information needs to be minimized satisfy the requirements many mission-critical applications. Meanwhile, not all terminals carry equally-urgent packets given their distinct situations, e.g., status freshness. Leveraging this feature, we propose an on-demand Medium Access Control (MAC) scheme, whereby each terminal transmits with dynamically adjusted aggressiveness based on its situations which are modeled as Markov states. A Multi-Agent Reinforcement Learning (MARL) framework is utilized and agent trained a Deep Deterministic Policy Gradient (DDPG) network. notorious issue for MARL slow non-scalable convergence – address this, new Situationally-aware MARL-based Transmissions (SMART) scheme proposed. It shown that SMART can significantly shorten time converged performance also dramatically improved compared state-of-the-art DDPG-based schemes, at expense additional offline training stage. outperforms conventional MAC schemes significantly, Carrier Sensing Multiple (CSMA), in terms average peak Age Information (AoI). addition, has advantage versatility different Quality-of-Service (QoS) metrics hence various state space definitions tested extensive simulations, where shows robustness scalability considered scenarios.