Agent-based approach to model-based dynamically reconfigurable control algorithm

There is a class of control algorithms, which are based on models of the controlled processes. The common feature of these algorithms is the fact that at the design stage the algorithm can be modified to take into account multiple sensors available in the controlled plant. Each additional independent measurement increases the accuracy of the simulation carried on by the model embedded in the algorithm. In this work the authors propose to improve this established static approach and implement the control system which dynamically derives the control law during its operation, depending on the availability of individual sensors. Availability may be reduced in the event of sensor failure, or increased after new sensors are connected. It is proposed to employ the agent-based programming paradigm as it has the dynamic reconfiguration capabilities which are crucial for the proposed technique. The presented idea is still a work on progress, so no experimental results are available, but this paper focuses on the details of the proposed technique in order to collect the opinions and conclusions of the research and engineering community.