A Multi-Agent System uses Artificial Neural Networks to Model the Biological Regulation of the Lower Urinary Tract
Joe Garcia, Francisco Maciá Pérez, Antonio Soriano, Francisco Flórez‐Revuelta · 2002
Abstract:- The robustness that shows the biological regulation of the human lower urinary tract provides a suggestive paradigm for the artificial control. The biological regulator consists of a heterogeneous group of nervous centres that act cooperatively, in a distributed way. That regulation conforms a behaviour of several types (autonomous work or conscious one) and it reduces the consequences in situations of bad operation. Related to that system, we propose a model of the paradigm of heterogeneous and distributed control that can be found in biological systems. The objective is to artificially reproduce the benefits of robustness in order to use it in the control of natural systems and artificial devices. The distributed aspects have been obtained using multiple intelligent agents, each one of which represents one of the biological centres. The interaction pattern among agents provides a heuristic based on the OAM neural network (Orthogonal Associative Memory). The knowledge has been added to the system by training, using correct patterns of behaviour of the urinary tract and wrong behaviour patterns due to the inoperability in up to two of the agents (representing deficiencies in up to two nervous centres at the same time). The experiments show that the model is robust and it satisfies the expectations of providing a model of the regulator system that allows to break into fragments the problem, in simple modules with own entity each.