FL-DBA: Fuzzy Logic Based Dynamic Bandwidth Allocation Algorithm for Next Generation Passive Optical Network
Résumé
The 6G system’s stringent latency and throughput requirements have imposed challenging constraints on Passive Optical Networks (PONs) and stretched classical Dynamic Bandwidth Allocation (DBA) algorithms to their limits. Moreover, the exponential increase in the number of connected users and the surge in generated traffic have pushed current scheduling and network management algorithms to their limits. Therefore, it has become essential to rely on artificial intelligence techniques with human-like reasoning to implement high-performance 6G compliant online DBA schemes. In this paper, we propose a Fuzzy-Logic based DBA (FL-DBA) algorithm to significantly enhance the performance of the PON and fully meet the 6G requirements. The FL-DBA performs fast and simultaneous, online time-wavelength scheduling with Mode Division Multiplexing, following an expert-like reasoning. Initially, a search algorithm is proposed to perform optimal scheduling for data generation and benchmarking. Subsequently, the resulting simulation data is evaluated and used as a reference to fine-tune a Mamdani Fuzzy Inference System (FIS) with respect to an experienced operators reasoning. The FIS is utilized to achieve fast and efficient time-wavelength scheduling by estimating the variation in the number of channels at the output, using the latency and traffic cost as inputs. Simulation results demonstrate the effectiveness of the proposed FL-DBA technique, achieving near-optimal performance as compared to the optimal search algorithm and state-of-the-art. The proposed FL- DBA algorithm fully meets the 6G KPI requirements, accommodating a network throughput exceeding 1.6 Tbps with latency and jitter levels below 100 μs and 10 μs, respectively, while achieving high channel utilization efficiency.