2025 Asian Conference on Communication and Networks (ASIANComNet 2025)

The advent of sixth-generation (6G) mobile networks, marked by an exponential proliferation of wireless devices and novel applications, has dramatically intensified the need for intelligent and adaptive spectrum management. Cognitive radio (CR), initially conceived for dynamic spectrum access, is now being fundamentally reshaped for the 6G era through its convergence with artificial intelligence (AI), edge computing, and reconfigurable intelligent environments. This survey presents a comprehensive overview of this transformation, structured around three foundational pillars: the evolution of spectrum sensing, the rise of AI-native resource allocation, and the challenges of multi-domain coexistence. We begin by charting the progression of spectrum sensing from classical techniques like energy detection to advanced, data-driven approaches that harness deep learning, federated intelligence, and generative AI, tailored for the high-frequency regimes of THz and mmWave bands. We then explore the shift in resource allocation from conventional optimization frameworks to end-to-end AI-native paradigms, encompassing deep reinforcement learning, multiagent collaboration, and the use of digital twins for predictive modeling. The survey further investigates the intricate coexistence of terrestrial, non-terrestrial (NTN), and RIS-assisted networks, tackling emergent issues such as cross-technology interference, security vulnerabilities, and dynamic regulatory landscapes. Concluding with a forward-looking perspective, we pinpoint critical open challenges from semantic-aware spectrum sharing and privacy-preserving learning to quantum-resistant security and ”green sensing” that are pivotal for building sustainable 6G ecosystems. By integrating concepts from diverse fields, this work serves as both a definitive reference and a strategic roadmap for researchers and engineers pioneering the next generation of autonomous wireless systems.

Cognitive radio, 6G, spectrum sensing, AIdriven, resource allocation, coexistence, federated learning