Cost-efficient routing, modulation, wavelength and port assignment using reinforcement learning in optical transport networks

Abstract With the rapid growth of global network traffic, more and more high-capacity network services are required. Traditional wavelength division multiplexing (WDM) networks are inadequate in terms of service scheduling and network management, so optical transport networks (OTN) have been proposed. OTN offers electric-layer switching ability to support finer granularity and higher spectrum efficiency. In OTN, we need to realize routing, modulation, wavelength, and port assignment (RMWPA) for supporting optical-electric (O/E) conversion (i.e., O/E port). Efficient RMWPA will significantly reduce the cost of operators. In this paper, we explore multi-modal information from OTN, and image the topology and routes to capture the information of the OTN and services. We design a cost-efficient RMWPA (CE-RMWPA) algorithm based on reinforcement learning to realize RMWPA for reducing cost in OTN. The proposed algorithm can interact with the OTN environment and learn how to make improved decisions based on the environment’s feedback. Simulative results demonstrate that the CE-RMWPA algorithm can achieve the optimization of RMWPA for reducing cost.