Virtual Sensing
Overview
Virtual sensing infers hidden states or missing measurements from available signals, improving monitoring and control when physical sensors are sparse, costly, or impractical. Common approaches include soft sensors and latent-variable models, physics-informed neural networks and digital twins, graph neural operators for spatial systems, and Kalman-style filters for online estimation. Applications span structures, energy networks, batteries, traffic, fluid flows, and robot interaction. Benchmarks show no single model works best everywhere, while robustness to sensor failures and changing conditions, reliable uncertainty estimates, and the trade-off between accuracy and computational energy remain open challenges.
Latest papers
27 papers4 papers in the last four weeks, down 20% on the four weeks before. 0.0% of all new papers.