Devices, including robotic devices, operating in viscous fluid flow can use passive sensor data collected to represent fluid parameters at an instant in time to derive information about the flow, the motion and position of the device, and parameters of the physical system constraining the flow. Using quasi-static analysis techniques, and appropriate feature selection for machine learning, very accurate determinations can be made, generally in real time, with very modest computational requirements. These determinations can then be used to map systems, navigate devices through a system, or otherwise control the actions of, e.g., robotic devices for clean-up, leak detection, or other functions.
