Optimizing Inter-Cell Resource Partitioning in Network Slicing: A Game-Theoretic Approach

Network slicing is pivotal for next-generation mobile networks, enabling efficient resource sharing among multiple tenants with distinct service needs. While traditional methods offer structured solutions, they often fall short in capturing real-world complexities. This paper introduces a hybrid optimization approach that synergistically combines promising game theoretic concepts such as Convex Optimization (CO) and Reinforcement Learning (RL), called Convex Optimization based Resource Allocation (CORA). It employs CO for global optimization under stable conditions and RL for real-time adaptability in fluctuating environments. The algorithm is both computationally efficient and scalable, capable of delivering globally optimal solutions across a variety of network conditions. It demonstrates a spectral efficiency of 0.91 for 3 slices with a small latency. CORA is ideal for smart cities where multiple services like emergency response and autonomous vehicles share network resources.