Compliance-Aware AI Framework for Anti-Interference and Adaptive Routing in Telecommunication Networks

Abstract

The rapid evolution of heterogeneous telecommunication networks has increased the difficulty of maintaining communication performance while operating within spectrum and transmission constraints. Existing network optimization approaches primarily focus on throughput, routing, and interference mitigation, whereas regulatory compliance is commonly treated as a separate post-processing activity. This separation limits the ability of network-control mechanisms to respond proactively to compliance risks. This study develops a Compliance-Aware AI Framework (CA-AIF) that integrates Coalition Formation Games (CFG), variance-aware DUCBQ adaptive routing, and Proximal Policy Optimization (PPO) with a policy-constrained compliance monitoring layer. The framework represents compliance risk through telemetry-derived violation indicators associated with spectrum occupancy, transmission power, and interference conditions and incorporates these indicators into network decision-making. A controlled simulation environment is developed using synthetically generated mobility, interference, channel, and spectrum-occupancy data. The proposed framework is evaluated against Q-learning, Dyna-Q, UCBQ, and DUCBQ using throughput, packet acceptance, routing stability, convergence, compliance violation rate, and computational cost. Across repeated simulations, the proposed framework achieves higher throughput and packet acceptance while reducing compliance violations relative to the evaluated baselines. The results demonstrate that incorporating compliance risk into network optimization can improve the stability of communication decisions without treating compliance as an independent post-processing task. The study contributes a reproducible simulation-based framework for compliance-aware intelligent networking and provides a basis for future validation using operational spectrum-monitoring and regulatory datasets.