Title:   NSF/Tokyo Report:  Force Control and Contour Following of Constraine
            Robotic Systems
Date:   January 14, 1998
Replaces: None


The National Science Foundation's offices in Tokyo and in Paris periodically
report on developments abroad that are related to the Foundation's mission.
These documents present facts for the use of NSF program managers and policy
makers; they are not statements of NSF policy.

Special Scientific Report #98-02 (January 09, 1998)


Force Control and Contour Following of Constrained Robotic Systems


Dr. Hemanshu (Roger) M. Lakhani prepared the following report. In June of
1997, Dr. Lakhani began a 24-month JSPS fellowship at the Tokyo Institute of
Technology (TIT). His host is Professor Toshiaki Ohkami of the Department of
Engineering Sciences at TIT. Dr. Lakhani can be reached via email at
roger@mass.mess.titech.ac.jp.

This research is being conducted in the Mechanical Engineering Sciences
department of the Tokyo Institute of Technology under the supervision of Dr.
Yoshiaki Ohkami. The goal of the project is to demonstrate novel force control
and contour following approaches for robotic systems. Much of the theory
behind the approach was developed during my doctoral studies in the same lab.
This phase of the research involves the implementation of the developed ideas
at an actual factory site where force control is required in an automated
manufacturing process.

Since robotic control began to be heavily studied, many approaches to force
control and contour following have been proposed. Of these, the most
significant are hybrid position/force control and impedance control. Whereas
these approaches have been shown to work in both laboratory and field
applications (mostly the former), there remain certain field applications for
which it is not clear that they can be easily realized or are the best
approach. The current research begins from the premise that automated
factories are already equipped with a certain amount of robotic hardware and
that the control hardware to drive these robots is more or less determined and
thus can not be easily modified. To implement sophisticated control approaches
such as hybrid and impedance control, either the control hardware must be
modified so that the robot actuators can be controlled according to the new
control law, or the factory must purchase and install a dedicated robot for
the force control t! ask. The first option is often met with resistance since
the factory is hesitant to tamper with a system that already performs position
control tasks quite well. The second option is expensive, and further reduces
cost effectiveness because two robots are used for a process that should
require only one. The best solution would be to use the factories existing
facilities as much as possible while ensuring that the industrial robot,
through minor modifications, can in addition perform the force control task.

Such a robotic structure would then consist of a high-performance
kinematically controlled manipulator (the factory robot) to which is mounted a
smaller system for the precise endpoint force control. The control of the two
systems is separate, yet in general the dynamic behavior of each system
affects the motion of the other. To further simplify the scenario, we assume
that the larger kinematically controlled manipulator is position controlled
using very stiff position servos (as is the case with the overwhelming
majority of industrial robots), so that motions of the smaller end-effector do
not affect the motion of the larger manipulator. We thus concern ourselves
only with the effect of the motions of the larger manipulator on the smaller
end effector. If the end effector is used for force control applications, then
these affects are primarily the inertial accelerations of the larger
manipulator, which translate directly into force variations at the tip of the
end-effector.

In modeling such a system, we must consider the tip motion of the larger
manipulator as a kinematic constraint applied to the base of the smaller
manipulator. This leads to a very different force control model of the smaller
manipulator than is obtained when assuming that the entire structure is force
controlled, from base to tip. Implementation is also challenging, since some
type of acceleration measurement of the tip of the kinematically controlled
subsystem is generally needed.

The current research builds upon force control algorithms designed around this
structure during my doctoral work in the laboratory. Laboratory experiments
were also carried out during that time to demonstrate the effectiveness of the
approaches. In the current phase of the research, these ideas will be
implemented and tested in an actual factory application.

Factory implementation and testing

The above force control and contour following approaches are being implemented
at a factory located in the town of Tsuyama in Okayama prefecture. The factory
produces fiber-reinforced plastic molds used in the construction of large
tanks and other storage equipment. Currently most of the manufacturing
process, including the spraying of plastic fibers onto the melted plastic base
and the subsequent rolling of these fibers into the plastic, is carried out by
factory employees under extreme laborious and harsh conditions. The goal of
the current development at the factory is to automate most of the coarse
spraying and rolling tasks and to use employees only for fine post-processing
of the plastics. The spraying task has been automated relatively easily since
it is a purely position control task. The laboratory entered the project in
the subsequent phase to consider approaches for automating the rolling task.
This task is particularly difficult to automate not only because it requ! ires
contact, or force control, but also because of the large variety of mold
shapes that need to be considered.

Since beginning the JSPS fellowship in June of 1997, we have worked
extensively on-site at the factory to install a complete automated
manufacturing system. The entire line consists of four major subsystems: 1) a
conveyor system to transport the plastic sheets to the rolling booth (this
system also handles the spray robot); 2) a robot used for the rolling task; 3)
a slave computer to control a customized force control hand which is attached
to the end of the contour following robot; and 4) a master computer to
supervise all the other subsystems. Of the various tasks involved, I have been
directly responsible for the integration of the force control hand, as well as
developing all the software necessary to run the system. This includes not
only the control software for the force control hand, but also the software on
the main supervisory computer. In addition, I am in charge of integrating all
the various subsystems, a task involving direct organizational contact with
all the s! ubcontractors and other related parties (factory workers and
managers, robot manufacturer, conveyor system manufacturer, computer
representatives, and electrical and cabling technicians.)

To date, the system integration has been completed and the manufacturing line
is in working order. We are now in a position to directly test force control
algorithms on the customized hand. As a preliminary step, a force measurement
system developed by me during my doctoral tenure was used to measure actual
force profile outputs during rolling tasks performed by factory employees. We
are now in the process of analyzing these data to see how the robot can be
controlled to yield similar performance. The first quarter of 1998 should see
extensive experimental work done on site to test the various force control
approaches.