ANFIS-DBA: ANFIS Based Dynamic Bandwidth Allocation Scheme for Latency Driven Cost Effective Next Generation PON

Motivated by the six generation (6G) system’s stringent latency and throughput requirements, we propose a high performance dynamic bandwidth allocation (DBA) scheme for the next generation passive optical network (PON). The proposed algorithm performs simultaneous time and wavelengths scheduling, which make the optimization problem extremely complex. For this reason, we propose a novel adaptive neuro-fuzzy inference system based DBA (ANFIS-DBA) algorithm, which simplifies our complex scheduling algorithm. The proposed ANFIS-DBA has the latency and traffic cost as inputs and the allocated channels and bandwidth as output. Its prediction model is based on input–output experimental data generated from the original optimization problem which is then used to create a Sugeno type fuzzy inference system (FIS) for estimating the number of channels that will be scheduled for an optical network unit (ONU). Predicted data resulting from the proposed ANFISDBA are compared with the optimal values, indicating that both models perform nearly the same with a root mean square error (RMSE) of less than one. In addition, a sensitivity analysis is conducted showing that the proposed ANFIS-DBA is more sensitive to the traffic cost input. The numerical achievements of the paper show that the proposed ANFIS-DBA is capable of accommodating a network throughput higher than 1.5 Tbps while securing a latency and jitter below 100 μs and 10 μs, respectively, which fully meets the 6G network performance requirements.