Science of Sensor System Software

Every sensor system—from a smart thermostat to a flood warning network—carries hidden uncertainty that can silently corrupt its readings and decisions, yet engineers lack the fundamental principles to manage it. This matters because sensors are increasingly embedded in critical infrastructure: they monitor water treatment plants, guide autonomous vehicles, control energy grids, and inform medical diagnostics. But sensors drift out of calibration, get knocked out of position, or are deployed in environments that degrade their accuracy. Worse, these small errors cascade through software layers, turning a slightly off reading into a bad decision. Current engineering practice cannot handle this multi-level uncertainty because there is no underlying science to guide it. This project aims to create that missing science—a systematic framework for designing sensor software that remains reliable despite pervasive uncertainty. If successful, it would allow engineers to verify that a sensor system will behave as intended across a wide range of real-world conditions, not just in a lab. The impact would be felt in any system that depends on sensor data: safer autonomous vehicles, more resilient power grids, and water treatment plants that detect problems before they become failures. The research is fundamental, but it targets a concrete gap that currently limits how confidently we can deploy sensor-driven systems at scale.