Learning Force Control for Robotized Sewing

Learning Force Control for Robotized Sewing

The robotics group of the department of the
Mechanical Engineering and Aeronautics has been
working the last years on the robotic handling of fabrics.
The robotic sewing is one of the tasks that consist the
group’s research. In this framework, a feedforward neural
network (FNN) controller [5], able to guide a wide range
of fabric types, was implemented. The target of the
controller was to apply a desired constant tensional force
to the fabric during the whole sewing process. In order to
investigate further this area, the implementation of a
FMRL controller was decided. Also, the fact that no
work using fuzzy adaptive control has been found in the
robotics handling of fabrics area consisted an additional
motive to use this method.VS Enterprises

In this paper, a FMRL control scheme is
adopted. The adaptation mechanism observes the signals
from the control system and adapts the parameters of the
fuzzy controller to maintain the performance even if there
are changes in the plant. The desired performance is
characterized with a reference model and the controller
seeks to make the closed-loop system behave as the
reference model would. The FMRLC, besides of tuning,
remembers to some extend the values tuned in the past, in
contrast to the conventional approaches which simply
continue to tune the controller parameters.
The proposed FMRL control scheme is shown in
Fig. 2. The fabric and the sewing machine’s velocity are
unknown. The goal of this system is considered to be
achieved when its controller maintains a desired constant
force applied to the fabric.
The presented FMRLC includes: the fuzzy
controller, the robot-fabric system to be controlled, the
learning mechanism, the reference model and a filter [8].

The robotic handling of non-rigid objects, such
as fabrics, is a very complicated problem since it is very
difficult to model and predict the behavior of the fabric.
The non-linearity, the large deformations and the very low
bending resistance of the fabrics increase the complexity
and difficulty of the robotic handling. In this paper, the
robotized sewing is examined, where the fabric must be
held taut and unwrinkled. Actually, a constant tensional
force, which must be applied to the fabric throughout the
feeding to the sewing machine, affects the seam’s quality
to a great extent [1]. Gershon [2] clearly justified the need
for force feedback control, in order to obtain a fabric’s
constant tensional force in the sewing task.
While there is a big variety of force control
methods on handling of rigid objects, the robotic research
literature is not so rich concerning the handling of limp
materials.
In the FIGARO system [3], a PI controller is
used, where the gains were chosen by trial and error.
These gains should be modified when a new type of fabric
should be handled. Gershon [2] strongly suggested that
the conventional control methods are inadequate to handle
the fabric tensional force. An adaptive control approach
was adopted by R. Patton et.al [4] for controlling the
tensional forces applied to a fabric, where the fabric’s
stiffness was the unknown variable. They mentioned that
non-adaptive control schemes are unsuitable for fabric
handling due to high variations of fabrics’ stiffness.
Koustoumpardis et al [5] introduced an intelligent force
control scheme based on a feedforward neural network
(FNN) controller, using a force sensor mounted on the
robots wrist

Learning Force Control for Robotized Sewing

H. F. Ho et.al. [6], used direct fuzzy adaptive
control for a nonlinear helicopter system. The control
objective was to maintain the elevation and azimuth
angles to maintain the desired trajectories. Rehman [10]
used a fuzzy model reference learning controller in order
to regulate the speed of an induction motor. Tarokh [7],
proposed an adaptive fuzzy control scheme for explicit
force control of a robot manipulator in contact with an
environment whose parameters are unknown and vary
considerably but slowly. The adaptation mechanism
modifies the fuzzy force controller according to the
difference between the actual and the desired force
responses.
The robotics group of the department of the
Mechanical Engineering and Aeronautics has been
working the last years on the robotic handling of fabrics.
The robotic sewing is one of the tasks that consist the
group’s research. In this framework, a feedforward neural
network (FNN) controller [5], able to guide a wide range
of fabric types, was implemented. The target of the
controller was to apply a desired constant tensional force
to the fabric during the whole sewing process. In order to
investigate further this area, the implementation of a
FMRL controller was decided. Also, the fact that no
work using fuzzy adaptive control has been found in the
robotics handling of fabrics area consisted an additional
motive to use this method.https://www.vssewingmachine.in/

https://g.page/vssewing?share

Leave a Reply

Your email address will not be published. Required fields are marked *