TWO-STEP UNCERTAINTY NETWORK FOR TASKDRIVEN SENSOR PLACEMENT

Abstract

Optimal sensor placement achieves the minimal cost of sensors while obtaining the prespecified objectives. In this work, we propose a framework for sensor placement to maximize the information gain called Two-step Uncertainty Network(TUN). TUN encodes an arbitrary number of measurements, models the conditional distribution of high dimensional data, and estimates the task-specific information gain at un-observed locations. Experiments on the synthetic data show that TUN outperforms the random sampling strategy and Gaussian Process-based strategy consistently.

Tasks

Reproductions