Performance evaluation and analysis of SENMP in robotics experiments

The idea of stochastic evolutionary neuron migration process (SENMP) is to use artificial evolution process to arrange spatially interacting computational entities, i.e. artificial neurons, into a pattern in 2-space so that desired behavior or dynamics emerges within the pattern. In this paper, we analyze the role of space in regard to SENMP performance using the well known double pole balancing problem as a test case. We also study the effect of environmental change to the adaptation process during a robot navigation experiment. This analysis suggests that synaptic scaling like dynamics, resembling inverted Hebbian rule, can emerge in the stochastic pattern formation process between the laterally interacting computational entities.