A Human-in-the-Loop Sensing System for Industrial Flow Field and Pollution Monitoring

Experimental measurements are essential for advancing applications such as profiling industrial systems, but extensive setup times often limit their practicality. We propose a method that eliminates setup time, enabling focused, high-quality measurements. The operator simply outlines the domain of interest, and our algorithms direct efficient sampling of the flow field.

In addition to velocity, our system samples gas concentrations, allowing mass flow calculations and integrated environmental and pollution assessments within the same workflow.

While sensing and learning algorithms are common in robotics, their application in fluid dynamics remains limited due to challenges in real-time data access, interactive control, and system infrastructure. Traversing systems also face constraints in reach, articulation, and usability. Our approach uses human-guided probe movement, with an augmented reality interface directing measurements to algorithmically defined locations.

We present the current implementation, assess its strengths and limitations, and outline directions for future development.