Industrial Robots

Robotic soldering of flexible PCBs

We present the first result on the dynamic contacting and manipulation of deformable objects and propose a new vision-based control scheme which smoothly integrates both operations of contacting and manipulation into a single controller. With the proposed controller, the robot is able to automatically contact the flexible PCB then actively manipulate it into the desired configuration. In addition, the dynamic stability of the closed-loop system is rigorously proved with Lyapunov methods such that the proposed control scheme is theoretically grounded for high-speed applications.

The following shows the performance of a flexible-PCB-soldering robot, consisting of a vacuum pad which actively deforms the flexible PCB, and a soldering end effector which is activated only after the deformation is stabilized.

Robotic manipulation of USB wires

We consider the problem of robotic grasping and manipulation of USB wires. In particular, a vision-based controller is proposed by organizing the control objective in a two-level structure.

At Level-I, the robot grasps then manipulates a specific wire;

At Level-II, the robot aligns the wires by referring to the USB color code.

The two-level structure is embedded into the controller, and the transition among multiple operations (i.e. Grasping, Manipulation, Alignment) is smooth and simultaneous, which is realized by assessing the status of deformation. Such formulation guarantees the feasibility and the autonomous capability of robotic manipulation of USB wires.

The above demonstrates vision-based robotic grasping and manipulation of USB wires. In the first experiment, all wires were not fixed at the grooves initially, and the robotic gripper was controlled to grasp then manipulate the wires, and carry out the alignment according to the color code sequentially. In the second experiment, the green wire was manually pulled away after it was grasped, to simulate the scenario that the wire falls off during the operation of grasping. In the third experiment, the red wire was manually pulled away after it was manipulated at the desired position, to simulate the scenario that the wire escapes from the groove during the operation of alignment.

Robotic renovation in indoor environment

Compared with the advancing construction technology, the improvement of renovation techniques is relatively low, and current works are highly dependent on the manual operation, which has two problems: lack of skilled labour and high manpower cost. Introducing robotics into renovation works can alleviate the problems. However, challenges for the development of renovation robots include: limited workspace; localization in unknown environment; real-time and high-accuracy detection; large displacement of robot; external disturbance and temporary loss of features. The collaboration project aims to develop an autonomous multi-functional renovation robot, and systematically overcome the limitations. In particular, CUHK will construct the vision system and develop the vision-based control approach, where the novelties are summarized as: (i) a 3D vision system with structured laser light is developed to fulfil the requirement of high accuracy (error<0.5mm) and real-time detection; (ii) a new visionbased robot controller will be developed to guarantee the positioning accuracy in the presence of uncalibrated camera, limited field of view, and singular configuration; (iii) Control techniques will be proposed to attenuate external disturbances due to movement on uneven grounds or existing of assembly gaps. The outputs of the proposed project will lead to commercial robotic systems, which will significantly automate the renovation works.

Robotic sanding of wooden box

As a subsidiary company of Bright Ford International Limited, Artdeco specializes in the production of wooden boxes for European luxury products. Currently, the production of these boxes heavily relies on labor, where workers stand next to a spinning sanding belt to remove a thin layer of polyethylene coating from the boxes. The aim of this research project is to develop a robotic system to automate the wood sanding process, without losing the quality and the efficiency of the manual process, and hence keeps human from unhealthy working environment and also lowers the lung disease incidents for sanding workers. In order to accomplish this objective, we propose to develop a four-DOF robot arm and a controllable sanding belt mechanism. A new impedance controller is developed to regulate the interaction force between the wooden box and the sandpaper, and thus guarantee the sanding quality. Neural-network techniques are also employed to estimate and compensate the disturbance existing in the system and thus improve the accuracy and the stability. Compared to the traditional manual sanding operation, the proposed system will help Artdeco to standardize production processes.