Multimodal Optimal Base Station Selection Network for Intelligent Communications.

Area of Science:

• Wireless Communication Systems
• Artificial Intelligence in Networking

Background:

• Next-generation wireless systems face challenges in base station selection due to high base station density and dynamic channel conditions.
• Traditional single-modal approaches using only channel or location data are insufficient for complex wireless scenarios, leading to suboptimal decisions.

Purpose of the Study:

• To develop a novel network, MOBS-Net, that integrates multimodal spatial and temporal information for optimal base station judgment and proactive prediction.
• To overcome the limitations of single-modal methods in accurately selecting base stations.

Main Methods:

• The judgment module uses Convolutional Neural Networks (CNNs) for image semantics and a Transformer-based fusion mechanism for combining image, location, and channel features.
• The prediction module employs multimodal sequential data and a large-scale multimodal model to extract temporal semantics for proactive switching.

Main Results:

• MOBS-Net achieved 92.20% accuracy in optimal base station judgment.
• MOBS-Net demonstrated 91.5% accuracy in prediction tasks, significantly outperforming single-modal baselines.

Conclusions:

• MOBS-Net effectively integrates multimodal information for intelligent base station decision-making in dynamic wireless environments.
• The proposed network offers a reliable and effective solution for optimizing base station selection and prediction.