Many industrial processes are difficult to optimize due to the lack of performance index definition, unavailability of sensors (and, indeed, measurements), and difficulties in setting up objective functions. In such scenarios, expert operators' knowledge drives the tuneup phase of the industrial processes/applications. Indeed, a programming-free approach to transfer such human knowledge to the production plant can be implemented to allow any operator to naturally/intuitively transfer his/her expertise to the target machine/robot.
The HCPbO project exploits preference-based optimization algorithms to address such needs. By adopting such an approach, it is possible to train an algorithm by means of experiments performed by an expert operator, guiding the optimization process. The optimization algorithm can then elaborate a machine configuration depending on different objective functions. The system provides suggestions to the human operator, assisting him/her in the optimization activities. In addition, an enhanced version of this algorithm (including both qualitative and quantitative optimization capabilities) will be developed to maximize the flexibility of the optimization toolbox. The developed algorithms will be tested in two relevant use cases.