Positionning of Vacuum Clamps using Genetic Algorithms (using EvA)

A problem of geometric optimization known in woodworking industries is solved by genetic algorithms. This project is supported and sponsored by Homag Maschinenbau AG, Schopfloch. Homag is the world leading machine-building corporation for woodworking machines.

The special task is to find an optimal positioning of vacuum clamps to fix a flat woodpiece in limited time. These wood-boards are fixed on a working table by vacuum clamps in order to treat them automated (shaping, sawing, snorring, boring, glueing ...). The goal is the optimal positioning of these clamps to avoid interference with robots or other clamps and to avoid fluttering in the work process. Grades of quality like "clamps should be near working traces" assess found solutions by a simple fitness function which needs to be optimized.

We apply genetic algorithms because of the high-dimensional parameter space and the existence of many local optima; especially EvA is used, a toolkit developed at this department. In a first step geometric algorithms prepare given data (polygons). For optimization we use the strategies Evolution Strategies (VeES), Genetic Algorithms (VeGA) and Co-Evolution (VeES) which are all implemented as parts of EvA. The application is written in C++, ported and tested on Linux, Solaris and Windows NT. There is a graphical user interface for visualization of input and output data (Tcl/Tk).



left: woodworking automata, rigth: found optimal positions of vacuum clamps

The results satisfy both by the quality of the output and by the simple formulation of the fitness function. Furthermore, they show that Evolutionary Algorithms are not only a real alternative to classical algorithms but that they are often even superior.