The Design and Control of High Performance Robotic Systems for the Nuclear Industry

KEPCO and EDF Research Project




Principal Investigator

 Professor S. Dubowsky

Group Members

Dr. Steven Dubowsky

Dr. Byung-Hak Cho, Visiting Scientist

Karl Iagnemma, PhD

Peter Jaffe, MEng

Marco Meggiolaro, PhD

Problem Statement

This page describes the Korean Electricity and Power Corporation (KEPCO) and Electricite de France (EDF) sponsored research being conducted in the Field and Space Robotics Laboratory at MIT.

The motivation of this research comes from the annual process of installing nozzle dams to enable nuclear reactor refueling. In order to replace the uranium fuel cells, the reactor core must first be filled with water. However, the water can not be allowed to flow down the hot and cold access pipes of the reactor. Currently, workers (referred to as "jumpers") to enter a small chamber in the steam generator to secure a 70Kg, 1 meter in diameter aluminum disc over the pipe opening. The problems associated with this process are as follows:

If a robotic arm could accomplish the task efficiently, this could result in incredible time savings as well as a significant human safety increase.


There are many difficulties inherent in designing an automated placement system for a nozzle dam. First of all, the payload is heavy and therefore requires a strong manipulator. Strong manipulators tend to have very high joint friction, and, as a result, it is very difficult to achieve accurate small motions.

The tolerances of the nozzle dam placement process are on the order of 1mm. This makes a teleoperated placement technique very difficult and would require at least the same amount of time as the worker process. However, this would at least reduce worker radiation exposure.

In order to decrease the procedure time, the manipulator could perform self calibration and then place the nozzle dam directly into position. Current control methods have shown that strong electric and hydraulic manipulators can successfully achieve 5mm tolerances, but the 1mm accuracy demanded by the nozzle dam problem seems to be out of the range of standard control methods.

The way to achieve such a level of precision is to compensate for the friction through a joint torque feedback path. This can be implemented by either developing extremely accurate (and complex) models of the nonlinear joint friction, or else by retrofitting the manipulator with expensive joint torque sensors. Using the calculated or measured joint torques a frictionless manipulator can be emulated. Our lab has pursued a third approach which has achieved submillimeter precision at a much lower cost.

The technique employed by the Field and Space Robotics Lab is to use a single base force/torque sensor to measure the wrench at the base of the manipulator, and, for small motions, this can be used to estimate the torques at all the joints of the robot. This method does not require a complex model of the robot or expensive joint torque sensors.

Our research is being conducted on a 7 degree of freedom hydraulic manipulator from Schilling Corporation. This robot is capable of lifting 1200lbs and exerting 600ftlbs of force. With the addition of a base force/torque sensor and a test bed of variable tolerances, our hope is to achieve successful placement of a heavy object into a tight fitting receptacle.


1. Achieving Fine Absolute Positioning Accuracy in Large Powerful Manipulators

2. A Force-Updated Kinematic Virtual Viewing System with Application to Nuclear Power Plant Maintenance


Inserted Nozzle Dam

Schilling Holding the Nozzle Dam



Copyright 1995-2001 by MIT.

All rights reserved.