Panagiotis N. Koustoumpardis

Fuzzy Logic Decision Mechanism Combined with a Neuro-Controller for Fabric Tension in Robotized Sewing Process

A new approach for flexible automated handling of fabrics in the sewing process is described, which focuses to control the cloth tension applied by a robot. The proposed hierarchical robot control system includes a Fuzzy decision mechanism combined with a Neuro-controller. The experts actions during the sewing process are investigated and this human behavior is interpreted in order to design the controller. The Fuzzy Logic decision mechanism utilizes only qualitative knowledge concerning the properties of the fabrics, in order to determine the desired tensional force and the location of the robot hand on the fabric. A Neural Network controller regulates the fabric tension to achieve the desired value by determining the robot end effector velocity. The simulation results demonstrate the efficiency of the system as well as the robustness of the controller performance since the effects of the noise are negligible. The system capabilities are more evident when the controller uses its previously acquired “experience”.

Admittance neuro-control of a lifting device to reduce human effort

In this paper, two admittance-based control schemes for a power-assisted lifting device are presented. This device can be used to hoist a heavy object interactively for reducing the operator’s burden. The proposed system integrates an admittance controller with an inner control loop that regulates the velocity of the object. The admittance is the outer loop that establishes the desired relation between the applied force to the object and its velocity. For the adaptation to a variety of loads, an online learning controller is implemented based on a neural network (NN) with backpropagation training. The overfitting of the NN is resolved with weight decay to decrease the oscillations around the equilibrium point. Alternatively, a gain scheduling PID controller is designed for the inner loop, which measures the object weight and tunes the gains with predefined rules. The performance of these two adaptation methods is demonstrated on an experimental setup and the results illustrate that better generalization can be achieved with the NN.

Robotized sewing of fabrics based on a force neural network controller

The robotized sewing of a wide range of fabrics is presented. A conventional industrial sewing machine in cooperation with a scara robot manipulator is used for producing straight line seams in single pieces of rectangular fabrics. A force controller based on neural networks is implemented in order to control the robot end-effector motion so as to ensure that a desired tensional force is applied to the fabric throughout the process of the sewing. The force controller scheme is designed and closes the loop outside the standard robot control law. All the necessary algorithms are implemented and executed inside the robots controller board. The 12-specimens of fabrics, which have been used for the robotized sewing experiment, represent a broad range of fabric types and compositions. The seams produced by the robot are compared with seams that are produced by a human operator under the same sewing machine conditions. The results reveal that the seams produced by the robot are comparable with that produced by the human operator and in few cases better.

Neural network force control for robotized handling of fabrics

A neural network force controller for regulating the fabrics tensional forces applied by a robot in fabric handling tasks is presented. The proposed controller enables the robot to apply a desired constant force to the fabric in a dynamically changing environment (fabrics variable and unknown velocity and length). A force sensor mounted on the wrist of the robot manipulator measures the actual force applied to the fabric, and the formulated force error is used in the backpropagation algorithm which trains the controller. Two real fabric handling examples are presented for testing the effectiveness of the controller, where the robot demonstrates a satisfactory real-time response. The feedforward neural network (FNN) controller is tested and the results are discussed and compared to those obtained using a PID controller under the same conditions. The performance of both controllers is investigated and a computational comparison is carried out.

Human Robot Collaboration for Folding Fabrics Based on Force/RGB-D Feedback

In this paper, the human-robot collaboration for executing complicated handling tasks for folding non-rigid objects is investigated. A hierarchical control system is developed for the co-manipulation task of folding sheets like fabrics/cloths. The system is based on force and RGB-D feedback in both higher and lower control levels of the process. In the higher level, the perception of the human’s intention is used for deciding the robot’s action; in the lower level the robot reacts to the force/RGB-D feedback to follow human guidance. The proposed approach is tested in folding a rectangular piece of fabric. Experiments showed that the developed robotic system is able to track the human’s movement in order to help her/him to accomplish the folding co-manipulation task.

Underactuated 3-finger robotic gripper for grasping fabrics

In apparel industries, the handling of fabrics still remains a manual work and its automation is a real challenge. In this paper, a three finger gripper is developed for grasping (pinching and clamping) fabrics under different ways. It is inspired by the human fingers movements in order to grasp a piece of fabric that is laid on a table. The conceptualization, the design and the prototype of the gripper are presented, along with the kinematic and static analysis of its mechanism. The proposed versatile gripper is based on a simple mechanism, where the two of the three fingers (pointer and middle) are underactuated by a tendon. Also, the plan for the sequence and the synchronization of the movements of the fingers is defined according to the grasping task. A prototype of the gripper is produced using 3D printing technique, which keeps the total cost very low. The prototype has been tested experimentally under several grasping tasks, where its agility is demonstrated.

Intelligent Robotic Handling of Fabrics Towards Sewing

Handling of flexible materials is one of the most challenging problems that have arisen in the field of robot manipulators. Besides the difficulties (e.g. geometrical uncertainty, obstacle avoidance, etc.) that emerge when handling rigid materials using robots, flexible materials pose additional problems due to their unpredictable, non-linear and complex mechanical behaviour in conjunction with their high flexibility and high bending deformations. The fact that sewing fabrics is a “sensitive” operation, since fabrics distort and change their shape even under small-imposed forces, poses barriers in the development of automated sewing systems. On the other hand, the need for great flexibility in garment assembly system is really imperative, since cloth manufacturing should cope with the fast fashion changes and new materials for fabrics and responding to the consumer demands. Our research efforts are focused on the development of intelligent control systems with multi-sensor feedback, enabling robots to perform skillful tasks in realistic environments towards higher flexibility and automation. In this work, the robot control approaches based on artificial intelligence for handling fabrics towards feeding in the sewing machine are described in detail.

On the Vibration Control of a Flexible Metallic Beam Handled by an Industrial Robot Within an ARX-Based Synthetic Environment

This study addresses the problem of vibration control of a flexible metallic beam which is transferred by an industrial robot. The control is designed in a special Matlab/Simulink synthetic environment that is founded on AutoRegressive with eXogenous (ARX) stochastic modelling of the robot-beam system through exclusively experimental data. Based on this, a simple closed-loop control system consisting of a feedforward typical Proportional-Integral (PI) controller and a feedback that enables the minimization of the induced force at the wrist of the robot is developed for beam vibration control. The data-based modelling of the robot-beam system allows for the precise design of the control system in offline mode, without interrupting normal production conditions, and achieves excellent performance in real time application.

ROBOT-HUMAN COOPERATION HOLDING AND HANDLING A PIECE OF FABRIC

A Neural Network force control of a robot manipulator collaborating with a human to handle a fabric is presented. A robotic gripper holds the one end of a piece of fabric while a human hand holds the opposite end. As the human moves the fabric arbitrarily, the robot tries to manipulate the fabric according to the required handling task. The task could be defined in a higher-level decision making process which is based on the artificial constrains of the handling task and influenced by the human needs. A force sensor mounted on the wrist of the robot manipulator measures the 3-components (Fx, Fy, Fz) of the actual force applied by the human hand to the fabric, and the formulated force errors are used in the backpropagation algorithm, which trains the Neural Network force controller. The controller is tested in a simple case of a desired handling task and the results are discussed and compared with the reversed case, i.e. the robot moves the fabric arbitrarily and the human tries to manipulate the fabric according to the handling task. The response of the controller show that the robot is capable to handle the fabric according to the desired constrains.

Manipulator Collision Detection and Collided Link Identification Based on Neural Networks

In this paper, a multilayer neural network based approach is proposed for the human-robot collisions detection during the motions of a 2-DoF robot. One neural network is designed and trained by Levenberg-Marquardt algorithm to the coupled dynamics of the manipulator joints with and without external contacts to detect unwanted collisions of the human operator with the robot and the link that collided using only the proprietary joint position and joint torque sensors of the manipulator. The proposed method is evaluated experimentally with the KUKA LWR manipulator using two joints in planar horizontal motion and the results illustrate that the developed system is efficient and very fast in detecting the collisions as well as the collided link.