Probabilistic Robotics Applied to Self-Localization Inside Oil Wells of an Autonomous System

Hugo Francisco Lisboa Santos¹; Marco Antonio Meggiolaro²

¹ PUC-RIO/PETROBRAS; ² PUC-RIO

doi:10.20906/CPS/COB-2015-1158

Resumo

The use of robots in the petroleum industry is still in the beginning phase. There are several studies for using robots on topside installations, but few for inside wells. This paper describes part of a research project that aims to develop a robot capable of performing maintenance operations at producer and injector wells. The goal is to design an autonomous robot for well intervention. In this development, one of the most challenging problems is to define a reliable self-localization system. Normally, positioning inside wells is performed with cables or with pipe strings. Since the proposed robot is not connected to a cable, an alternative is needed. A possible solution is explored here, based on sensor fusion of a tachometer and a magnetic sensor. For this fusion, several probabilistic and non-probabilistic robotic techniques are considered. Kalman filters are implemented, with adaptations for the current problem. Particle and Histogram Filters are also evaluated, with good results despite their high memory and computational requirements. All techniques are evaluated and compared using field data.

Palavras-chave: Positioning; Autonomous; Robot; Well; Offshore; Probabilistic Robotics; Kalman; Sensor Fusion; Particle Filter; Histogram Filter; Localization