Balancing Subjectivity and Objectivity in Network Selection: A Decision-Making Framework Towards Digital Twins

TL;DR

This work tackles the network selection problem in heterogeneous wireless networks (HWNs) by addressing two key limitations of traditional MADM-NS methods: rank reversal and insufficient alignment with user/service requirements. It introduces a novel BWM-GWO weighting framework that blends subjective weights from the Best-Worst Method with objective weights from Grey Wolf Optimization through a convex combination, enabling a balanced MADM-NS approach and paving the way toward a digital twin (DT) of HWNs. The method is integrated with TOPSIS-NS and SAW-NS, tested in MATLAB on an HWN model with WiFi, WiMAX, LTE, and 5G, and shown to reduce rank reversal by substantial margins (up to 71.3% in some scenarios) while improving QoS for time-sensitive services. The results suggest that the BWM-GWO framework effectively personalizes network selection to service class requirements and system dynamics, offering a practical route to DT-based, AI-assisted optimization in next-generation networks.

Abstract

Selecting the optimal radio access technology (RAT) during vertical handovers (VHO) in heterogeneous wireless networks (HWNs) is critical. Multi-attribute decision-making (MADM) is the most common approach used for network selection (NS) in HWNs. However, existing MADM-NS methods face two major challenges: the rank reversal problem (RRP), where the relative ranking of alternatives changes unexpectedly, and inefficient handling of user and/or service requirements. These limitations result in suboptimal RAT selection and diminished quality of service, which becomes particularly critical for time-sensitive applications. To address these issues, we introduce in this work a novel weighting assignment technique called BWM-GWO, which integrates the Best-Worst Method (BWM) with the Grey Wolf Optimization (GWO) algorithm through a convex linear combination. The proposed framework achieves a balanced decision-making process by using BWM to compute subjective weights that capture user/service preferences, while employing GWO to derive objective weights aimed at minimizing RRP. The development and validation of this framework establish a digital model for NS in HWNs, marking the initial step toward realizing a digital twin (DT). Experimental results show that integrating the proposed BWM-GWO technique with MADM-NS reduces RRP occurrence by up to 71.3% while significantly improving user and service satisfaction compared to benchmark approaches.

Figures (4)

• Figure 1: HWNs environment.
• Figure 2: General architecture of using BWM-GWO in MADM-NS.
• Figure 3: Average rank reversal ratio.
• Figure 4: Optimized weight calculation for RAT attributes.