Active Beyond-Diagonal RIS Empowered Heterogeneous Edge Computing: A Distributional Reinforcement Learning Approach

arXiv cs.AI · 18 hours ago

Researchers developed a reinforcement learning approach called DSAC-T to optimize energy usage and latency in mobile edge computing systems that use active beyond-diagonal reconfigurable intelligent surfaces. The method achieved an 81.67% feasibility ratio and online decision time of 0.0267 seconds per scenario. The approach enables more efficient computation offloading decisions by modeling return distributions rather than just expected values in a complex system with coupled optimization variables.