Abstract :
[en] In this paper, we address the problem of contact
state recognition for compliant motion robotic systems. The
wrench (Cartesian forces and torques) and pose (position and
orientation) of the manipulated object in different Contact
Formations (CFs) are firstly captured during a certain task
execution. Then for each CF, we develop an efficient Takagi-
Sugeno (T-S) fuzzy inference system that can model that specific
CF using the available input (wrench and pose) - output (the
desired model output for each CF) data. The antecedent part
parameters are computed using the Gravitational Search- based
Fuzzy Clustering Algorithm (GS- FCA) and the consequent parts
parameters are tuned by the Least Mean Square (LMS).
Excellent mapping and hence recognition capabilities can be
expected from the suggested scheme. In order to validate the
approach; experimental test stand is built which is composed of a
KUKA Light Weight Robot (LWR) manipulating a cube rigid
object that interacts with an environment composed of three
orthogonal planes. The manipulated object is rigidly attached to
the robot arm. The robot is programmed, by a human operator,
to move in different CFs and for each CF, the wrench and pose
readings are captured via the Fast Research Interface (FRI)
available at the KUKA LWR. Using the suggested approach,
excellent modeling is obtained for different CFs during the robot
task execution. A comparison with the available CF recognition
approaches is also performed and the superiority of the
suggested scheme is shown.
Scopus citations®
without self-citations
7