Abstract:Sensor placement is an important and ubiquitous problem across the engineering and physical sciences for tasks such as reconstruction, forecasting and control. Surprisingly, there are few principled mathematical techniques developed to date for optimizing sensor locations, with the leading sensor placement algorithms often based upon the discovery of linear, low-rank sub-spaces and the QR algorithm. QR is a computationally efficient greedy search algorithm which selects sensor locations from candidate positions with maximal variance exhibited in a training data set. More recently, neural networks, specifically shallow decoder networks (SDNs), have been shown to be very successful in mapping sensor measurements to the original high-dimensional state space. SDNs outperform linear subspace representations in reconstruction accuracy, noise tolerance, and robustness to sensor locations. However, SDNs lack principled mathematical techniques for determining sensor placement. In this work, we develop two algorithms for optimizing sensor locations for use with SDNs: one which is a linear selection algorithm based upon QR (Q-SDN), and one which is a nonlinear selection algorithm based upon neural network pruning (P-SDN). Such sensor placement algorithms promise to enhance the already impressive reconstruction capabilities of SDNs. We demonstrate our sensor selection algorithms on two example data sets from fluid dynamics. Moreover, we provide a detailed comparison between our linear (Q-SDN) and nonlinear (P-SDN) algorithms with traditional linear embedding techniques (proper orthogonal decomposition) and QR greedy selection. We show that QR selection with SDNs enhances performance. QR even out-performs our nonlinear selection method that uses magnitude-based pruning. Thus, the combination of a greedy linear selection (QR) with nonlinear encoding (SDN) provides a synergistic combination.

Reconstruction of Fields from Sparse Sensing: Differentiable Sensor Placement Enhances Generalization

Intelligent Reconstruction for Spatially Irregular Seismic Data by Combining Compressed Sensing with Deep Learning

Data-Driven Sparse Sensor Placement for Reconstruction

Global field reconstruction from sparse sensors with Voronoi tessellation-assisted deep learning

Deep Learning Improvements for Sparse Spatial Field Reconstruction

Sparse Sensor Placement for Interpolated Data Reconstruction Based on Iterative Four Subregions in Sensor Networks

Continuous Field Reconstruction from Sparse Observations with Implicit Neural Networks

Efficient Sensor Placement for Signal Reconstruction Based on Recursive Methods

Spatially-Aware Diffusion Models with Cross-Attention for Global Field Reconstruction with Sparse Observations

Learn to Sense: a Meta-learning Based Distributed Sensing Algorithm in Wireless Sensor Networks

A Multi-resolution Adaptive Encoding Method for 3-D Reconstruction of DPO-SDF

Learn to Sense: A Meta-Learning-Based Sensing and Fusion Framework for Wireless Sensor Networks

Data Reconstruction Using Smart Sensor Placement

Efficient Sensor Placement from Regression with Sparse Gaussian Processes in Continuous and Discrete Spaces

Data-Induced Interactions of Sparse Sensors

Optimization of Wireless Sensor Network Deployment for Spatiotemporal Reconstruction and Prediction

Data-driven sensor placement with shallow decoder networks

A physics-driven sensor placement optimization methodology for temperature field reconstruction

Enhancing Estimation Performance in Distributed Sensing Through Autoencoder-based Sensor Array Feature Extraction

Spatial Data Augmentation: Improving the Generalization of Neural Networks for Pansharpening.

ProxNF: Neural Field Proximal Training for High-Resolution 4D Dynamic Image Reconstruction