The field of cellular networks is moving towards a more nuanced understanding of coverage quality, with a focus on quantifying usability and stability. Researchers are developing new metrics and frameworks to better capture the complexities of wireless propagation and network behavior. Notably, the use of artificial intelligence and machine learning is becoming increasingly prominent in the design of resilient radio access networks. Innovative approaches, such as reinforcement learning, are being applied to optimize network performance and adapt to changing conditions.
Some noteworthy papers in this area include: Quality of Coverage (QoC) introduces a novel set of key performance indicators to characterize network behavior, offering a more fine-grained representation of cellular coverage. Universal Maximum Likelihood Decoding via Fast Vector-Matrix Multiplication presents a simple framework that reduces the worst-case complexity of maximum-likelihood decoding for arbitrary block codes. Resilient Radio Access Networks: AI and the Unknown Unknowns examines the challenges of designing AIs for resilient radio access networks, highlighting the limitations of current statistical learning methods. End-to-end Learning of Probabilistic and Geometric Constellation Shaping with Iterative Receivers presents an end-to-end learning method for constellation shaping, demonstrating improved bit error rate performance. QoSGMAA: A Robust Multi-Order Graph Attention and Adversarial Framework for Sparse QoS Prediction proposes a novel architecture for predicting Quality of Service in complex network environments. Deep Reinforcement Learning Approach to QoSAware Load Balancing in 5G Cellular Networks under User Mobility and Observation Uncertainty presents a deep reinforcement learning framework for autonomous load balancing. Adaptive Design of mmWave Initial Access Codebooks using Reinforcement Learning proposes a hybrid reinforcement learning framework for adaptive initial access codebook design. PolarZero: A Reinforcement Learning Approach for Low-Complexity Polarization Kernel Design investigates kernel construction under recursive maximum likelihood decoding using a reinforcement learning framework.
