|
  Our
research on computational theory of haptics is focused on developing a
theoretical framework for studying the information processing and control
strategies common to both humans and robots performing haptic tasks. For
example, although the "hardware" of the tactile apparatus in humans and
robots are different, they have the common feature of mechanosensors embedded
in a deformable medium. Therefore the mechanistic analyses needed to solve
the computational problem of coding (predicting sensor response for a
given mechanical stimulus at the surface) and decoding (inferring the
mechanical stimulus at the surface by suitably processing the sensor response)
are sufficiently similar for human and robot tactile sensing systems that
a theory common to both systems can be developed.
We first developed such a "computational theory" using a simplified 2D
half-space model of the human or robot finger subjected to arbitrary pressure
or displacement loading conditions normal to the surface, and gave explicit
formulae for the coding and decoding problems (Srinivasan, 1988). We have
now expanded these results to a more general 3D half-space model where
the load direction can be completely arbitrary (Karason, Srinivasan, and
Annaswamy, 1994). Explicit solutions for the coding problem are given
and enable the selection of a useful set of relevant stimuli as well as
the choice of sensors appropriate for maximizing the information about
the stimulus on the skin surface. The solution of the decoding problem
is also given, both for the idealized noise-free case and for the realistic
case with measurement noise. For the latter, the solutions are shown to
be numerically stable and optimal.
In our work during the previous years, we were successful in answering
basic identification and control issues that arise during manipulation
of compliant objects using compliant fingerpads (Annaswamy and Srinivasan,
1994; 1996). In order to understand the fundamental aspects of these tasks,
we have analyzed the problem of identification of compliant objects with
a single finger contact, as well as under a two-finger grasp. Using lumped
parameter models, we have carried out the identification of human and
object parameters, using either force or displacement inputs to the rigid
backing of the end-effector. Based on identified parameters, control strategies
are developed to achieve a desired manipulation of the object in the workspace.
We have also modelled the dynamic interactions that occur between compliant
end-effectors and deformable objects by a class of nonlinear systems (Annaswamy
and Seto, 1993). It was shown that standard geometric techniques for exact
feedback linearization techniques were inadequate. New algorithms were
developed by using adaptive feedback techniques which judiciously employed
the stability characteristics of the underlying nonlinear dynamics. In
both theoretical and simulation studies, it was shown that these adaptive
control algorithms led to successful manipulation. The theoretical results
can be used to generate testable hypotheses for experiments on human or
robot haptics.
|
