This paper demonstrates a proactively reconfigured edge network that can adaptively configure IP topology for dynamic service demand. Simulation result show the average utilization has improved from 32.50% to 68.98% compared traditional static network.

Smart home technologies are growing actively all around the world. As a result, great pressures are imposed on internet of things networks by dynamic traffic and plenty of devices. The passive optical network is considered one of the most promising fronthaul technologies. In particular, the time and wavelength division multiplexing passive optical network has shown the advantage of high capacity and received attention recently. In support of internet of things networks, the energy and transmission efficiency has emerged as an important issue on the time and wavelength division multiplexing passive optical network enabled fronthaul networks. In this paper, we try to enhance the energy and transmission efficiency of the time and wavelength division multiplexing passive optical network enabled reconfigurable fronthaul. Fronthaul links’ load balancing is also taken into consideration. An integer non-linear programming model is employed to formulate the joint optimization problem. We also provide an adaptive genetic algorithm-based approach with fast convergence. The simulation results show that the active units of fronthaul can be dynamically switched on/off with the traffic variation and a significant energy saving is achieved. In addition, the maximum transmission efficiency increases by 87% with integer non-linear programming method in off-peak periods.

We propose a multi-agent federated reinforcement learning algorithm for privacy- enhanced service provision in multi-domain optical network. Experimental results show blocking probability is reduced from 11% to 1% and the deviation ratio is within 3%.

No abstract

No abstract