Subsurface Flow Path Modeling From Inertial Measurement Unit Sensor Data Using Infinite Hidden Markov Models

Subsurface flows are challenging to map due to the inaccessibility and extreme nature. This coupled with global positioning system (GPS) denied, sensorially deprived, complex, and locally self-similar environments result in most of the traditional mapping methods failing. This results in extensive subsurface environments that remain unknown and unexplored. In this paper, we show that it is possible to detect features using infinite hidden Markov model (iHMM) from inertial measurement unit data (IMU) collected along surface and subsurface flow paths. By assuming zero velocity at the beginning of each feature, we present a data driven model to reconstruct unknown 2D subsurface flow paths using 9 degrees of freedom (dof) IMU sensor data and two known location points along the path. In addition, we show the advantages of using iHMM compared to simpler feature extraction methods. The model is validated on multiple controlled examples, as well as experimental real world datasets.