Toward a Self-adaptive Multi-Agent System to Control Dynamic Processes

Bioprocesses are especially difficult to model due to their complexity and the lack of knowledge available to fully describe a microorganism and its behavior. Furthermore, controlling such complex systems means to deal with their non-linearity and their time-varying aspects. A generic approach,relying on the use of an Adaptive Multi-Agent System (AMAS) is proposed to overcome these difficulties, and control the bioprocess. This gives it genericity and adaptability, allowing its application to a wide range of problems and a fast answer to dynamic modifications of the real system. The global control problem will be turned into a sum of local problems. Interactions between local agents, which solve their own inverse problem and act in a cooperative way thanks to an estimation of their own criticality, will enable the emergence of an adequate global function for solving the global problem while fulfilling the user's request. This approach is then instantiated to an equation solving problem, and the related results are presented and discussed.