DRL-Based Distributed Joint Sub-Band Allocation and Power Control for eXtended Reality Over In-Body Subnetworks

In-body Subnetworks (IBS) are set to become key enablers for standalone eXtended Reality (XR) applications, allowing users to leverage smartphones and XR display devices (XRDD) to create immersive virtual and augmented environments. However, supporting reliable XR services over IBS poses significant challenges due to dense deployment, resource constraints, and stringent communication requirements. To address these challenges, we propose a resource allocation algorithm specifically designed for transmitting XR video frames over IBS. We formulate the problem as a reliability maximization task, modeled as a Partially Observable Markov Decision Process (POMDP). A discrete version of the Soft Actor-Critic (SAC) algorithm is employed to derive a near-optimal policy for joint power control and sub-band allocation. Our approach utilizes a multi-agent framework, where each IBS acts as an agent responsible for resource allocation decisions. Our results demonstrate that, despite relying on partial information of the network, the proposed algorithm surpasses distributed benchmark and performs comparable to centralized algorithms, achieving satisfaction rates of 55% and 98% for the case of 8000 IBSs per square kilometer and packet success rate (PSR) thresholds of 99% and 95%, respectively.