A model-free self-organizing fuzzy logic control system using a dynamic performance index table

The ‘raison d’être’ of self-organizing fuzzy logic control (SOFLC) algorithms is the performance index table, which normally issues the adequate corrections to the low-level control given certain performance criteria. In the standard SOFLC architecture, the performance index table is generic, fixed a priori and is of a ‘grid-partition’ structure making the whole scheme inefficient in terms of computational complexity and performance. In this paper, we propose a new model-free SOFLC architecture whereby the performance index table is ‘dynamic’, of free structure and starting from an empty table. The new proposed architecture includes the required mechanisms to ensure self-organization such as those associated with the individual evaluation via an online genetic algorithm (GA) and the fine-tuning of individual regeneration range for micro-GAs, to optimize the rules of the performance index table and enhance the system’s performance. Results of experiments on a non-linear muscle relaxation process show that the proposed control scheme is superior to the standard SOFLC algorithm in terms of performance and robustness against the system input/output scaling factors selection and parameter variations.