Analysis and comparison of optimization techniques supporting the engineering planning of reconfigurable robot cells

Abstract

The present paper studies the automated optimization of reconfigurable robot cells to support engineering planning activities. The specific type of reconfigurable robot cells under investigation allows an arbitrary arrangement of functional machine modules on a base plane accessible to the robotic manipulators. The research is motivated by the fast-growing configuration space of reconfigurable robot cells which results from the configurability of the cell layout and the performed workflows. Searching for optimized configurations within this configuration space is an iterative procedure which currently requires lots of manual efforts and expert assessment skills for each iteration. Performing many optimization iterations requires much time, which typically results in only a few performable iterations and thus in suboptimal configurations. To compensate for this, automated support for the underlying planning problem is required. The specific planning problem under investigation concerns finding motion-time optimized layout configurations for fixed workflows. This planning problem was chosen because of its strong association with the productivity of robot cells and the fact that it is the basis for further considerations on the influences of performed workflows. The main contribution of the research is a guideline concerning the decomposition of the mentioned optimization problem. This decomposition concerns the specific formulation of the optimization problem as well as its solving principle. The approach of the paper is to compare combinations of single- and multi-objective problem formulations as well as single- and multi-step solving procedures. Respective variants of evolutionary and conventional algorithms are tested for solving the optimization problems. The experiments are conducted using a specifically developed simulation and optimization framework which supports layout reconfiguration, robot path planning, and time parametrization based on state-of-the-art methods.