In-Tank Sensor Error Prediction and Modeling

Internal inspection of the floors of large storage tanks presents significant localization challenges. Standard localization techniques such as GPS are not viable due to the steel walls. Navigation using internal beacons is feasible, but requires an initial survey and significant effort to set up. Consequently, an autonomous stand-alone navigation approach is desirable. In previous work, we studied the localization of an in-tank inspection robot using angular and range sensors and performed a Monte Carlo study to determine optimal sensor orientation. Improved in-tank localization can be achieved using optimal estimators which require error models for individual sensors. In this work, we numerically model the range sensor error magnitude as a function of sensor angular orientation to permit online relative error estimation.