تولید مدل اتوماتیک و پیاده سازی کد PLC برای سیاست های به هم پیوسته در سلول های ربات های صنعتی

In industrial production lines, for example in the automotive industry, cells with multiple industrial robots are common. In such cells, each robot has to avoid running into static obstacles and when the robots work together in a shared space they must also avoid colliding with each other. Typically, the latter is enforced by manually implementing interlocks in programmable logic controllers (PLCs). This is a tedious, error-prone task that is a bottleneck in the development of production lines. The PLC-code being man-made also greatly complicates the maintenance and reconfiguration of such production lines. However, in industry today, a lot of development of robot cells is made offline in 3D simulation environments which enables the use of computers also for deciding and implementing the necessary coordination. This paper presents a method that makes use of information in a robot simulation environment in order to automatically extract finite state models. These models can be used to generate supervisors for ensuring that the deadlock situations that may arise as a consequence of the introduced interlocks are avoided. It is also possible to optimize the work cycle time for the cell. Finally, PLC-code to supervise the production cell can be automatically generated from the deadlock-free and possibly optimized system model. This approach results in a high flexibility in that the coordination function can be quickly reimplemented whenever necessary. A prototype implementation has been developed making use of a commercial 3D robot simulation tool, and a software tool for supervisor synthesis and code generation. The approach is general and should be possible to implement in most offline robot simulation tools.

Today's trend of mass customization and the ambition to shorten the time-to-market result in a demand for more flexible manufacturing systems, faster development of new production lines and faster adaptation of old production lines for producing new products. The automotive industry is a very illustrative example with mass production of long series of car models being replaced by customized products in short series. Production in the body-in-white shops of the automotive industry is typically centered around an assembly line, where the car body moves from cell to cell. Within a cell multiple robots weld parts such as wheel-houses and the roof, to the stationary car body. As the robots sometimes work inside each others’ workspaces there is a risk that the robots collide. Naturally, this must always be avoided. To prevent robot collisions, the access to shared spatial volumes, here called zones, must be coordinated. Only one robot may occupy a zone at a time, to guarantee freedom of collision. However, this coordination may instead introduce blocking problems, where a number of robots get into a circular wait, indefinitely waiting for each other to release a currently occupied zone. Naturally, this must also always be avoided, hence the coordination problem is by no means trivial.

A method for automatic generation of nonblocking and work cycle optimized supervisors for coordination of industrial robot cells has been described. The method extracts models for the robot cell from a 3D simulation environment. The models are generated via a generic XML-format which allows for the use of different robot simulation software while keeping the same algorithms. The contribution of this paper lies not in the novelty of the individual steps in this process, but rather in assembling these analysis and synthesis steps together into an integrated whole. Most importantly, the presented method makes the use of formal methods for analysis of industrial robot cells possible. This is an important step since formal methods can bring greater flexibility and the possibility to do optimization analytically. The authors believe that this can significantly speed up the offline development of robot cells. The introduction of via-points leads to improved performance compared to the method proposed in Flordal et al. (2004) since it allows waiting at the border of a zone, potentially saving some time.