Abstract:
Cell-Free Massive MIMO (CF-mMIMO) networks have emerged as a groundbreaking
advancement in wireless communication. This study proposes a novel hierarchical
multi-layered approach for the optimization of edge caching in CF-mMIMO networks
using the Deep Q-Learning (DQL) Framework. The objective of the proposed framework
is to balance service quality and minimize response times, backhaul traffic, and
power consumption. A substantial dataset encompassing user queries, response times,
content sizes, and caching choices is utilized to evaluate the proposed approach. The
proposed framework achieves an impressive 95% Cache Hit Ratio (CHR), indicating
efficient cache usage and reduced reliance on remote services. By leveraging the DQL
models, the efficiency of edge caching in CF-mMIMO networks is substantially enhanced.
This enhancement leads to consistently high cache hit ratios, expedited content
delivery, and improved network efficiency. Furthermore, the model’s energy efficiency
contributes to the network’s long-term sustainability. Rigorous testing across diverse
scenarios confirms the model’s versatility across rural, suburban, urban, and stadium
settings. By comparing different caching eviction policies, we establish LFU as optimal
for cache efficiency, considering the cache hit ratio and processing time. Overall,
this research underscores the substantial advantages of the proposed deep Q-learning
model for edge caching in cell-free Massive MIMO networks, significantly impacting
cache hit ratio, processing time, and energy efficiency across a range of practical
scenarios.
Keywords— Deep Reinforcement Learning (DRL), Network optimization, Content caching,
Intelligent decision-making Wireless networks.
