Newsgroups: comp.robotics
Path: brunix!uunet!pipex!uknet!comlab.ox.ac.uk!rwd
From: rwd@robots.ox.ac.uk (Ron Daniel)
Subject: intention discrimination
Message-ID: <1993May10.145449.3010@oberon.robots.ox.ac.uk>
Originator: rwd@oberon.robots
Organization: Robotics Research Group, Engineering Science Dept, Oxford, UK.
Date: Mon, 10 May 1993 14:54:49 GMT
Lines: 29

We are looking at methods of controlling robot aides for disabled people.
At present we are focussing on people with Athetosis, which means that they
are subject to tremor and writhing movements often associated with the
need to carry out a task.

It appears that experienced carers can filter out non-intentional
movement in order to communicate.  We postulate that it may be
possible for a filter to be used in conjunction with a joystick to
control the motion of a robot to carry out a similar task.  The filter
would not be just a straight-forward linear filter, but would have to
discriminate intention from the behaviour of the operator.  Is anyone
working on similar problems on the net?

A possible approach would be to identify a set of simple primitive
actions required of the system and to construct most likely
trajectories through its associated state space.  Robot control would
then achieved by switching between control modes, the mode being used
depending on the current task primitive being carried out.  The
question we hope to answer is whether a joystick provides more
information than a straight switch control for a person with
Athetosis, and if so, how the degrees of freedom inherent in the joystick
can be mapped into the most important degrees of freedom of the robot.
For example, during a grasp, the joystick should control the opening
and closing of the gripper plus speed along the approach vector.  The
next stage would be to control motion along a trajectory independent
of the state of the gripper.


Ron Daniel rwd@uk.ac.oxford.robots
