Sensorless Tactile Exploration:

An application to oil well mapping

Principal Investigator

Prof. Steven Dubowsky

Prof. Richard Wiesman

Group Members

Francesco Mazzini, Phd Student


         Rising oil demand and advanced oil recovery techniques made it economically attractive to rehabilitate previously abandoned oil wells. These wells often have a number of  junctions where divergent branches leave the main well at unrecorded depths.  To rehabilitate a branch, the location and shape of its junction need to be determined in the least possible time.

         Cameras and lasers cannot be used in oil wells. Sonic sensors have low accuracy and problems of reflection.

         Similar situations can be found in gas pipes, sewers, mines.

Autonomous Tactile Exploration

A manipulator, mounted at the end of a wireline tool, is brought close to the junction. The manipulator base is fixed,and then  the autonomous tactile exploration starts:

         The robot tip probes a few points on the surface

         From these point, the surface geometry is reconstructed

Oil well branching structure, with cutaway detail of robot exploring the junction

Robot control

         The robot is moved under impedance control, which allows interaction with the environment and detection of contact without any force or tactile sensor. Only joint encoders are needed. This makes the robot robust, inexpensive and reliable, even in challenging environments.

         To increase robot precision, a specific compensation method for backlash is being developed when the robot is in contact with the surface

Robot design

To maximize the robot workspace, a design study selected a 4 degrees of freedom manipulator: a translational link along the wireline tool and an elbow robot. A laboratory prototype was designed and built, reproducing the last three links. The robot is mounted inside a tank modeling the junction. Robot and tank are designed to be operated while submerged in a fluid.

Experimental robot arm mounted inside the oil-well junction model, during exploration

 Search Strategy

         Tactile exploration is inherently slow, and keeping an oil well inactive is very expensive. Hence, the time needed to map the surface is a crucial factor.

         The surface model is constructed using very few points, representing its shape as combination of simple geometric primitives.

         Key factor is the selection of the best location to probe the surface, to obtain more information. Heuristic strategies have been implemented to guide the robot movements. A strategy using
Information Theory concepts is being developed.

         The strategy needs to deal with the (initially unknown) limitations on robot kinematics by the surrounding environment, and the detection of the contact location without any sensor.

         The strategy is kept general enough to be applied to several tactile exploration problems.


The control algorithm and the search strategies have been tested and validated experimentally. The robot behavior has been tested while mapping several shapes and when submerged in fluids with different viscosities.


Reconstructed junction using two different search strategies, showing the number of data point needed in each 





Copyright 1995-2010 by MIT.

All rights reserved.