Dynamic Bandwidth allocation algorithm based on traffic classification with the aid of LSTM and GRU for industrial passive optical networks

With the rapid development of the Industrial Internet of Things (IIoT), the Industrial Passive Optical Network (IPON) has become an important technology to meet the future needs of access networks due to its deterministic and low latency. However, due to the traffic characteristics of IoT devices, the traditional dynamic bandwidth allocation method cannot make full use of network bandwidth, resulting in low resource utilization. We propose an algorithm to jointly predict classified IoT/non-IoT traffic using LSTM and GRU neural networks, and dynamically allocate bandwidth to SDN-based IPON networks according to the classification results. Results show that it reduces latency by 15.5% and packet losses by 47% under a load of 0.6, compared to existing typical dynamic methods.