BOLT: A Bayesian Online Learning Framework for Time Sensitive Networks in Metaverse
Source of Publication
2023 International Conference on Intelligent Metaverse Technologies & Applications (iMETA)
Efficient load balancing in wireless networks is a critical challenge that directly impacts resource utilization and user experience for time-sensitive networks in the Metaverse. This paper addresses this challenge by proposing a Bayesian online learning approach, enhanced by Software-Defined Networking (SDN) principles and the OpenFlow API. The objective is to dynamically optimize traffic distribution across multiple access points (APs) in wireless networks using real-time observations. The proposed solution integrates SDN concepts and leverages the OpenFlow API to enable centralized control and dynamic configuration of network elements. By employing Bayesian online learning, the load-balancing policy is continuously updated based on real-time observations and historical data. This approach minimizes total load imbalance across APs, accounting for uncertainties inherent in network dynamics. The Bayesian online learning framework incorporates proba-bilistic models to capture load fluctuations and estimation errors. By utilizing historical data and observed network behavior, the model refines its load-balancing decisions over time. MATLAB is utilized for simulations, demonstrating the effectiveness of the proposed approach in achieving load-balancing objectives for time-sensitive networks in the Metaverse. The results underscore the adaptability and robustness of the system in managing load variations while leveraging the power of SDN and Bayesian learning.
Adaptation models, Metaverse, Wireless networks, Bandwidth, Fasteners, Load management, Real-time systems
Balasubramanian, Venkatraman; Bouachir, Ouns; and Al-Habashnat, Ala'a, "BOLT: A Bayesian Online Learning Framework for Time Sensitive Networks in Metaverse" (2023). All Works. 6206.
Indexed in Scopus