Selecting Offline Reinforcement Learning Algorithms for Stochastic Network Control

Research evaluates offline reinforcement learning algorithms for wireless network control, finding Conservative Q-Learning produces more robust policies under stochastic conditions than sequence-based methods. The study provides practical guidance for AI-driven network management in O-RAN and 6G systems where online exploration is unsafe.

• →Conservative Q-Learning consistently outperforms sequence-based methods in stochastic wireless environments.
• →Offline RL is promising for next-generation wireless networks where online exploration poses safety risks.
• →Sequence-based methods can compete when sufficient high-return trajectory data is available.
• →The research addresses a knowledge gap in offline RL behavior under genuinely stochastic dynamics.
• →Findings offer practical algorithm selection guidance for O-RAN and 6G network control systems.