The field of wireless communication is witnessing a significant shift towards the adoption of reinforcement learning (RL) and artificial intelligence (AI) to optimize network performance. Recent research has focused on developing innovative solutions to address the challenges of dynamic networks, unknown environments, and heterogeneous deployments. One notable trend is the use of offline RL to improve bandwidth estimation and resource allocation in real-time communication systems. Additionally, researchers are exploring the application of laser chaos-based RL to achieve stable acoustic relay assignment and high throughput in underwater networks. Experience-centric resource management schemes are also being proposed to enhance user quality of experience in integrated sensing and communication networks. Furthermore, there is a growing interest in developing generalizable RL frameworks for radio access networks (RAN) that can operate effectively in diverse environments. Noteworthy papers include:
- Robust Bandwidth Estimation for Real-Time Communication with Offline Reinforcement Learning, which proposes a robust bandwidth estimation framework based on offline RL.
- Generalization in Reinforcement Learning for Radio Access Networks, which presents a generalization-centered RL framework for RAN control that improves average throughput and spectral efficiency by ~10%.