Dynamic Neuro-modelling Using Bacterial Foraging Optimisation with Fuzzy Adaptation

This paper presents current work on fuzzy adaptation of chemotactic step size of bacterial foraging algorithm and its application to optimisation of parameters of a neural network, i.e. weights, biases and slope parameters of activation function, in modelling of a single-link flexible manipulator. Experimental input-output data pairs gathered from a laboratory-scale single-link flexible manipulator rig are used both in the modelling and validating phases. Moreover, a set of correlation tests is used to validate the resulted model. The objective of the work is to assess the performances of the improved bacterial foraging algorithms in comparison to standard one based on the cost function value achieved, convergence speed, and time-domain responses.