COATI: a chaotic optimization framework for joint eMBB-URLLC scheduling in 5G/6G NR networks

Abstract The proliferation of heterogeneous traffic with divergent Quality of Service (QoS) requirements in 5G/6G networks necessitates fundamentally new resource allocation strategies. The existing deep reinforcement learning methods suffer from high training complexity, while heuristic approaches lack adaptability. We propose COATI (Chaotic Optimization Algorithm with Traffic Intelligence), a metaheuristic framework that synergizes the Coati Optimization Algorithm (COA) with chaotic dynamics and traffic-aware operators. Formulated as a mixed integer nonlinear program (MINLP), COATI introduces: (i) logistic map-based chaotic initialization for diversity preservation, (ii) adaptive Lévy flights guided by buffer status reports for dynamic exploration–exploitation trade-offs, and (iii) a QoS-aware fitness function incorporating proportional fairness constraints. Simulations based on 3GPP scenarios demonstrate that COATI reduces URLLC latency violations by 38%, improves eMBB throughput by 9.2%, and converges 27% faster than state-of-the-art methods. The algorithm’s bio-inspired yet mathematically rigorous approach establishes a new paradigm for real-time resource management in dense cellular networks.