Seungjoon Choi

Image-based Visual Servoing using Sliding mode control

Visual servoing consists of controlling the displacements of a robot using, in closed loop, the data provided by the camera. In this paper, we propose to design the nonlinear controller for image-based visual servoing which is more robust than classical controller on the parameter uncertainties. It can be guaranteed not only stability of the system but the tracking errors boundness which considers the gain. Additionally using the modified uncertainty bound matrix, the range of admissible uncertainties is increased

Development of a new variable remote center compliance for assembly robots

In this paper, we develop a new elastomer shear pads (ESPs) variable remote center compliance for peg-in-hole assembly. It consists of one top plate and one bottom plate, which are linked by four ESPs as in a conventional RCC. However, it has a structure to enable the four ESPs to rotate about the vertical axis of the bottom plate so that the position of the compliance center can be adjusted freely by changing the rotation angle. As a result, the new VRCC can be applied to pegin-hole assembly with various peg lengths. Furthermore, it has a very simple mechanical structure so that it can be manufactured easily.

The coil recognition system for an unmanned crane using stereo vision

In steel works, there is a need to know where is a coil in a yard for an unmanned crane operation. Image based vision processing can be effective in detecting important geometrical features of an object and lead to accurate estimation of the objects position/orientation required for pickup operations such as a crane. In this paper, we investigate techniques for recognizing the coils position/orientation with established a 3D map by stereo vision system which is implemented on FPGA chips and generates disparity values from stereo camera images at high speed. In order to recognize the coils position/orientation, a surface segmentation technique is used. Our system is tested at coil yard in steel works for an unmanned crane.

Development of a new variable remote center compliance using stiffness adjusters

A new variable remote center compliance (VRCC) for peg-in-hole assembly is presented. The proposed VRCC consists of elastomer shear pads (ESPs) and stiffness adjusters. The stiffness adjuster (SA) is clamped to a part of the ESPs in order that the ESP parameters can be changed freely. As a result, the position of the compliance center can be adjusted to the various peg lengths. Furthermore, the proposed VRCC has a very simple mechanical structure and is easy to manufacture. Experimental results verify the effectiveness of the VRCC.

Non-linear observer-based trajectory tracking control for wheeled mobile robot

In this paper, non-linear observer-based trajectory tracking control for a wheeled mobile robot is proposed. The non-linear observer is designed for not only an unmeasurable orientation error but current position error. This full-order observer shows robustness to measurement noise. Besides, a controller which exponentially stabilizes the tracking error is used. The non-linear observer with the controller is asymptotically stable by means of persistently exciting (PE) condition and cascaded system theory. A proof of stability and simulation results are given.

Set-up Model Design for Silicon-Steel Cold Rolling Mill

Abstract Silicon steel is used as soft magnetic materials in electric applications and devices, especially for transformers and motors because of their enhanced electrical characteristics. Since silicon steel has different characteristics from general steel, an adequate setup model for rolling process is needed to guarantee quality and efficiency. Current setup model is a lookup table based on the experience of field operators. It is hard to meet the various customer’s demands. Therefore, there is a need to build a suitable setup model. In this paper, we attempt to design a setup model for silicon steel based on accurate roll force prediction. An corrective neural network is applied to improve prediction accuracy of roll force.

A 3-D Object Recognition System Based on a Rapid 3-D Vision System 1

Abstract In this paper, we develop a 3-D object recognition system based on fast 3-D stereo vision. This system can locate position and orientation of several coil-shaped objects. Representing a target object, the supertoroids is used and the position/orientation of the object can be found in a depth image through optimization techniques. The vision system which is implemented on FPGA chips generates the depth image from stereo camera images at high speeds. This systemcan be applied to coil yards for crane automation.