Newsgroups: comp.robotics.research
Path: cantaloupe.srv.cs.cmu.edu!europa.chnt.gtegsc.com!gatech!hookup!nstn.ns.ca!newsflash.concordia.ca!sunqbc.risq.net!hobbit.ireq.hydro.qc.ca!NetNews.IREQ.Hydro.QC.CA!mboyer
From: "Victor B. Naumov" <nau@iias.spb.su>
Subject: Russian Robotics 2/8: Force Control for Elastic Robots
Message-ID: <MBOYER.95Jul5164234@pellan.ireq-robot.hydro.qc.ca>
Lines: 72
Sender: news@ireq.hydro.qc.ca (Netnews Admin)
Organization: St.-Petersburg Institute for Informatics and Automation, RAN
Date: Wed, 5 Jul 1995 20:42:34 GMT
Approved: mboyer@ireq-robot.hydro.qc.ca, crr@ireq-robot.hydro.qc.ca

                 ACADEMY OF SCIENCES OF RUSSIA
   ST.-PETERSBURG INSTITUTE FOR INFORMATICS AND AUTOMATION
                         ROBOTICS LAB.
                     of prof.F.M.Kulakov



         Generalization of the Force Robot Control Methods
         -------------------------------------------------
                       for Elastic Robots
                       ------------------

Fields of Application
---------------------

Design of control systems for assembly robots and robots for execution
of grinding, polishing and other manipulations with objects which have
geometrical and/or velocity constraints on the displacement of the
objects.

Description of idea of research
-------------------------------

If robots carry out manipulations with grasped objects, having
constraints, contact forces and torque arise (reaction forces).  The
interactions deform elastic elements of the robot.  Larger reaction
forces result in greater displacement the desired trajectory of the
gripper of robot from it's desired path.  Usually it is impossible to
follow the desired trajectory precisely.

The execution of these manipulation is possible if the robot has a new
property, which provides the ability to control the reaction forces.
This method of robot control is called the force control method.  The
most universal methods of robot force control is the active continuous
method of robot force control.

Despite the successes in the research of robot force control,
there are still some problems which need to solved.

For instance
- for known methods of the active continuous force control, the
  stability of the perturbed motion was found to be small.  This is
  especially true for manipulators which display elastic deformation 
  of the transmission from drive to joints, of the wrist, and of the
  manipulator's links;
- this method was only considered for cases of special constraints,
  which has object grasped and moved by the manipulator etc.

A model of a real elastic manipulator, grasping an object with
constraints, has led to a newly proposed control law which attempts to
address some of these problems.  This new control law uses feedback to
change the object position increment vector (delta p) based on
knowledge of the elastic part of the manipulator instead of using
feedback from the reaction forces/torques.  

It has been shown that the newly proposed control laws are a
generalization of the known control laws.  They coincide with the last
in some particular cases and degenerate into the known control laws, 
if the manipulator is absolutely rigid.

Tests of the new force control methods have been conducted using a
real anthropomorphic robot, PUMA.  The settling times were less then
that using the old, force control methods.  This control law appears
to have great potential for the practical control of robots in such
applications as assembly.  However, more detailed theoretical analysis
and tests need to be conducted.


--
 *********************** (moderated) ***************************
 Submissions:                 Meta-discussions/information:
 crr@ireq-robot.hydro.qc.ca   crr-request@ireq-robot.hydro.qc.ca
