Artificial intelligence driven spatiotemporal query framework for mobile wireless sensor networks

Mobile Wireless Sensor Networks (MWSNs) with dual mobility enable dynamic data collection in smart agriculture, but lack efficient spatiotemporal query support. We propose AI-SQF, a lightweight framework integrating LSTM-based predictive clustering, Q-learning-guided multi-sink mobility, NSGA-II optimization, and grid-based indexing for real-time SQL-like queries (e.g., “soil moisture in Zone A, past 10 min”). Cluster heads maintain a 50 × 50 m metadata index; sinks resolve queries locally, reducing transmissions. Evaluated in a 2000 × 2000 m field using NASA SMAP/MODIS data, AI-SQF achieves 190 ± 10 ms latency, 13.5 ± 0.6 pkt/s throughput, 95.5% coverage, > 91% accuracy, and 2.3 ± 0.25 s freshness — outperforming static, KNN, and DRL baselines by 22–50% (p < 0.01). Fault tolerance sustains 91 ± 2.5% delivery after 12% node failure.