BIJO: Bilevel Interactive Joint Optimization for Resource Allocation and Routing Generation

In the age of 5G and the advent of 6G technology, meeting various applications' diverse and stringent QoS requirements has gained greater urgency. Most existing works only address this problem by optimizing network resource allocation or routing generation, ignoring their mutual influence. In this paper, we jointly optimize resource allocation and routing generation while comprehensively considering the mutual influence and demonstrating that joint optimization is a bilevel optimization problem. We propose BIJO, consisting of two-level agents, to explore using DRL to address the bilevel optimization problem. The upper agent allocates resources for different types of services and the lower agents choose the forwarding path for each type of service. The two-level agents are optimized interactively and iteratively. Extensive experiments on Mininet confirm that BIJO provides a win-win outcome, where various requirements are met as much as possible and network resources are utilized evenly.