Expert Systems and Mathematical Optimization Approaches on Physical Layout Optimization Problems

This work presents a new approach to the problem of component placement on printed circuit boards. It describes a system for automatic placement (SAP) on printed circuit boards (PCB) that uses both artificial intelligence techniques (expert systems) and classical optimizing algorithms. The previous approaches model the placement problem as a classical optimizing problem and they do not take into account the intrinsic circuit features. This work researches the use of empirical knowledge acquired from PCB designers and classical algorithmic techniques to improve the placement algorithm performance and final art result. Starting from component and net lists, SAP identifies and classifies groups of these objects, which are important to the problem domain. Afterwards, it uses a rule base to find the relative placement between components. Finally, the relative placement is optimized to minimize the total wire length and to equalize the distribution of wires on the board.