UBC Theses and Dissertations
Vision assisted system for industrial robot training Sang, Tao
In an industrial environment, robotic manipulators are trained to perform repetitive operations at high speed. For safety reasons, the power to the robotic manipulator must be shut off whenever the operator enters the robotic work-cell. However, during training, human operators are required to enter the work-cell frequently to verify the manipulator positions. This cycle of "power-off-check-power-on-train" is time consuming and expensive. This work presents a self-calibrated vision system for human-guided robot training. The objective of the system is to simplify the training process thus reducing the training time and cost. The system assists robot operators in training industrial robots quickly, safely, and accurately. The system can be used in various industrial environments and requires no predetermined model of the working space. First, the system self-calibrates to compute the initial camera locations and orientations. Next, the user demonstrates the tool path that is to be followed by the robot arm using a 6-degree of freedom training medium. Finally, the system optimizes the camera locations so that, at those locations, the cameras can provide sufficient images for the operator to observe and perform further remote training of the robot as required without entering the workcell. The result shows that with the help of the vision system, the number of times that a human is required to enter the robotic work-cell is reduced to no more than 3 times for robot task training. At the same time, the accuracy of the training process is not compromised with the use of the vision system.
Item Citations and Data