IIC Journal of Innovation 12th Edition | Page 41

Digital Twin Development for Serial Manipulators: Data Driven Optimized Planning and Sequencing of Tasks Initially, the part recognition system of shaft assembly parts is constructed with the trained deep learning model using computer vision (CV) techniques. The CV system at the station requires a 3-dimensional (3D) camera in this application as the point cloud data provides more accuracy for cobot motion and material handling processes. Afterward, the new cobot motion planning and simulation for specific shaft assemblies are performed by a robotic engineer as shown in Figure 8. Once the new cobot motion path is achieved, the engineer can review and validate it with the virtual cobot in VML. When the confirmation is complete, the new cobot path is released to the robot controller, and the cobot setup is ready to execute the customized shaft sub-assembly operations. Figure 8: The cobot simulation model running in the robot motion planning software or a real physical cobot on the shop floor, based on the requirements. In VML, the cobot DT model is running and links with the actual cobot at the shaft assembly station. The detail cobot DT design and implementation steps are published in a Novelty & innovation of the DT model In order to minimize the effect on the on- going full gearbox assembly line process, applying the cobot digital twin model to validate and verify a newly developed cobot motion plan is cost-effective and efficient. This is because the planning, simulation, analysis and verification of the new cobot motion plan can be completed with the virtual cobot (cobot DT model), which can then be connected to the simulation model - 36 - November 2019