Optimization of multi-cell edge caching strategies based on trial and error multi-armed bandit algorithm

With the rapid development of edge computing technology, edge caching has become a key technology to improve network performance and reduce transmission delay, which has important application value in 5G and IoT ecosystem. However, how to effectively optimize edge caching strategies in a heterogeneous network environment with multiple users and base stations, especially when facing dynamically changing network loads and user demand patterns, is still an urgent challenge to be solved. In this paper, we propose an innovative TnE-MBA (Trial and Error Multi-Armed Bandit) algorithmic framework, which utilizes the exploration-exploitation balancing mechanism of the MAB algorithm to efficiently learn the optimal caching decisions in uncertain environments while avoiding the performance loss caused by blind exploration. In our design, each base station acts as an agent that makes caching decisions independently based on historical observations and the current network state, and continuously adjusts its strategy through system feedback. We comprehensively evaluate the proposed algorithm through large-scale simulation experiments, and the results show that the TnE-MAB algorithm significantly improves the system caching hit rate (by 28.94%) and significantly reduces the end-to-end transmission delay (by 32.14%) compared with the existing methods under high load and rapidly changing network environments. These results demonstrate the research value of the proposed method in multi-cell edge caching scenarios.