PC-based Design Of Visual-guided DELTA Manipulator High-speed Sorting System

A real-time sub-system taking advantage of KRTS (Kithara Real-time Suite) was realized in order to integrate motion control with machine vision conveniently, simplify motion control system as well as cost less. Real time tasks ran in real time sub-system and other tasks ran in Windows in parallel. KRTS was a modular real-time extension for the Windows.In this way, standard network cards instead of motion control cards and image capture cards were used to transfer data of motion control and images. The solution which integrated motion control with machine vision was proposed and achieved standardization, commonality and network as a motion control platform.Based on that platform, a visual-guided high-speed DELTA sorting system was developed. Both vision process system and robot motion control system were built in one Visual Studio project and it could be a reference for deeper research on machine vision integrated motion control system. Meanwhile, mathematic model of the whole sorting system was presented and for easier analysis, it was divided into three subsystems, kinematic model of DELTA manipulator, vision model as well as objects-tracking model. After that, mathematic algorithms which were extracted out of subsystems were verified.Vibrations and shocks resulted from DELTA manipulator’s reciprocal motion in a high speed were detrimental to DELTA mechanical structure and grabbing effects. To relieve those vibrations and shocks, a novel acceleration and deceleration algorithm was proposed according to the way which S acceleration and deceleration curve was proposed. Much more smoothness was obtained by this new type of curve in comparison with S-Curve because the jerk was continuous in the whole curve. What’s more, the jerk curve was made up of several sine and cosine curves and through integral, the acceleration curve, velocity curve as well as displacement curve could be obtained respectively. Finally, this new curve was tested and analyzed through Matlab.To overcome the repeat shooting to workpieces by vision system in sorting process, an image deduplication algorithm based on system time and workpieces’ positions was proposed. The running time of the real-time sorting system and the current location of workpieces on running conveyors was combined to uniquely identify a part. Then, the non-linear mathematical model was established for workpieces tracking, which was solved by Newton-Raphson iteration. Finally, in prototype test the maximum sorting speed reaches 110 times per minute, grab rate is 99.97%, which proves that the algorithm can meet the real-time, the accuracy and the stability requirements.