Attention map-driven compressive sensing for stable and high-accuracy distributed data storage in mobile crowdsensing systems

In Mobile Crowdsensing (MCS) systems, distributed data storage (DDS) faces challenges of unstable reconstruction accuracy due to uneven environmental data distribution and reliance on randomized measurement collection. Existing approaches often fail to adapt to regional data heterogeneity, leading to suboptimal performance. To address this, we propose an intelligent compressive sensing strategy that integrates an attention map-guided measurement selection mechanism. Initially, a small subset of measurements is utilized to generate an attention map, which dynamically identifies data-dense regions. The central server then prioritizes additional measurements based on this map, enhancing both reconstruction accuracy and stability. Theoretical analysis demonstrates that the proposed method achieves a significant reduction in reconstruction error compared to conventional random sampling strategies. Extensive simulations under real-world urban sensing scenarios further validate its superiority, showing marked improvements in both reconstruction quality and stability. This approach establishes a robust framework for adaptive data collection in large-scale MCS applications, effectively addressing the challenges of uneven data distribution while maintaining resource efficiency.