Multi-Agent Reinforcement Learning Based on Hybrid Action Representation for UAV Swarms’ Integrated Communication and Control

This paper investigates the integrated communication and control (ICAC) co-design for unmanned aerial vehicle (UAV) swarms using multi-agent reinforcement learning, which involves optimization with hybrid action spaces. In such settings, UAV trajectory and power control are naturally modeled as continuous variables, while communication resource allocation is typically discrete. However, existing methods for handling hybrid actions often ignore the interdependencies between discrete and continuous components, thereby limiting cooperative decision-making capabilities for ICAC among UAVs. To overcome this limitation, we propose a multi-agent reinforcement learning algorithm based on hybrid action representation (MAHyAR) to solve the dynamic optimization problem, which maintains an equilibrium between fair communication and UAVs’ energy consumption. Furthermore, we innovatively construct a hybrid action representation space to characterize the intrinsic dependencies between discrete and continuous actions. Experimental results demonstrate that our approach achieves an 11% increase in fair throughput and conserves energy consumption by up to 33%, compared to the baseline method.