Collective Self-Tuning for Complex Product Design

A complex product is generally a system composed of numerous interdependent components, each one representing specific disciplines and developed using associated expertise. When analysing the problem from another point of view, we can see that for each design domain, a generally huge set of real already designed elements exists. Thus, when constructing a new element, it is interesting to use this already known and acquired knowledge. This knowledge does not only contain the discipline's information but also the engineers' experience. Considering this point of view, the design of complex products defines a new generic class of complex problems. In this paper, we address this class of problems using the Self-Adaptive Population Based Reasoning (SAPBR) generic approach. It is based on the Adaptive Multi-Agent System (AMAS) theory that takes advantage from cooperation to design robust and open multi-agent systems. In SAPBR, agents use cooperative self-tuning principles in order to estimate and discover new characteristic values for the design of new elements. The obtained system is compared to the Self-Organising Map (SOM) and the Multilayer Perceptron (MP) algorithms that address similar problems.