funded research

Funded Research

A main problem of spot welding in automotive assembly lines is failures of welding robots. Such a failure can cause an immediate stoppage of the line and a great loss of throughput until the robot is fixed. A possible solution in such a circumstance is the reallocation of the welding spots from the failed robot(s) to a backup robot(s) during the repair period. To that end, an optimal recovery plan in terms of the throughput loss has to be determined under various operational constraints. Despite its importance, this particular problem and its practical implications have not been addressed in the literature. Part of the reasons, might be its computational intractability features. In this report, we will present an optimal recovery plan which minimizes the throughput loss in case of robots' failures under certain scenarios. The problem description and characteristics based on the real data provided by the GM R&D and a literature review are presented. Following, five Mixed Integer Linear Programming (MILP) models for different variations of the single model problem are presented: i) a single-robot backup strategy; ii) a multi-robot backup strategy; iii) a multirobot backup strategy with a limitation on the number of backup robots; iv) a multi-robot backup strategy with a linear cycle time estimation; and v) a mixed-model multi-robot backup strategy. Examples for both small-scale and large-scale problems are given for the multi-robot backup strategy. Next, a tutorial of our software application GM Welds Reallocation (v0.1) is presented. The new tool is illustrated via a test case based on the Pontiac regular cab model 803, ex ning the computational performance of the application and the sensitivity analysis of the solution. Last, conclusions and potential research directions are presented, such as stochastic simulation for the mixed-model environment and a 3D graphic simulation for the avoidance of robot interference.

Remote Manipulation and Control of Robotic Cells
Goldstain O., Lipman R., Bukchin Y., Ben-Gal I.

This research and teaching project, which is funded by VATAT, proposes a new approach for remote learning, based on Internet and simulation tools. The proposed program is designed to enable students and developers to remotely set up and manipulate a robotic work cell or a production line, without the operator's necessity of being at the actual work site.
Preliminary research has shown that remote robotic control interfaces are already being used in several laboratories throughout the world. However, the main use of these interfaces was found to be for demonstration or for the activation of a small set of predetermined actions. In this work, a general approach is being developed to comply with more complicated tasks in flexible work cells; this new approach will enable the end user to manipulate different objects within the work cell with enough degrees of freedom to achieve various goals.
In the early stages, three different approaches were examined for the solution architecture:

1. Manipulating a work cell remotely by controlling the activating machine (computer) of the cell via remote access programs i.e. Windows Remote-Desktop-Control (RDC) or Virtual-Network-Connection (VNC).
2. Simulating the work cell at the remote computer (home or office), enabling redesign and optimization of the process. Once the process planning is completed, connecting to the work cell as mentioned above, for the actual performance of the tasks.
3. Creating a web server on the work cell's computer, enabling the end user to connect to the workstation through the Internet. This approach provides us with a much robust solution and reduces the bandwidth problems that were detected in the first two solutions.