Endocrine Inspired Modulation of Artificial Neural Networks for Mobile Robotics

Abstract

The desire to operate highly autonomous robots in harsh conditions which may threaten their survival has demonstrated the need for artificial systems which can adapt to their environment. Traditionally many have attempted to control robots with artificial neural networks (ANNs). These provide reasonably successful instantaneous reactive behaviours in response to stimuli (e.g. correcting course deviations). However they lack the ability to respond in a longer term fashion to more gradually changing conditions. The same problem is true in biology, especially with regards to species without any higher brain functions that are able to consider longer term factors. In mammals two other systems play a key role in longer term changes to the neural system, the endocrine and immune systems. The endocrine system is able to modulate the behaviour of a variety of cells (including neural cells) with a time frame lasting between a few seconds and several months. Whereas the immune system provides longer term responses lasting between minutes and years. However, the immune system is far more complex than the endocrine system and its capabilitiesbetween species, therefore it will not be dealt with here.