Super Resolution Deduction: Inferring Fine-Grained Capacity for Urban Signal Station Deployment

With the boosting of mobile devices, wireless sensor networks, and the internet of things, abundant multi-modal data, such as GPS signal, sensor data, are produced intentionally or unintentionally, which can represent the people's active patterns, vehicle's routes, and city's flows to develop a smart city. These multi-modal data are usually transmitted and received by signal stations deployed in the city. However, reasonably choosing the signal stations' locations is still an open issue for enhancing people's life quality in the smart city. To this end, we propose the Super Resolution Deduction (SRD) model for solving the signal station selection problem. SRD first initializes the city map as a coarse-grained heat map representing the capacity of the signal stations. Then an image-based super-resolution deduction model is proposed to obtain a fine-grained signal station capacity for deploying. To be specific, we employ Dense Block to capture the spatio-temporal correlations, C-Attention to selectively enhance useful feature maps, and S-Distribution to impose structural constraints. By sharing the GPS data load with the new deployment of signal stations, we ensure the smart city's efficiency and effectiveness. Extensive experimental results on real-world dataset Changchun City demonstrate that our proposed model achieves the superior performance among the state-of-the-art baselines.

Automated summary

In the era of smart cities, the challenge of choosing signal station locations for efficient transmission and reception of multi-modal data persists. To address this, we propose the SRD model, which optimizes signal station deployment through image-based super resolution deduction.