Using Higher Order Synapses and Nodes to Improve the Sensing Capabilities of Mobile Robots R. J. Duro, J. Santos, J. A. Becerra, F. Bellas, J. L. Crespo 1 Grupo de Sistemas Autónomos Universidade da Coruña, Spain gsa@cdf.udc.es Abstract In this paper we present three types of higher order artificial neural networks that may be included in heterogeneous ANN architectures to improve the perceptual performance of mobile robots. Two of the networks are based on synaptic processing, with the advantage that this type of processing works with the raw data and not an average, as is the case of nodes. The first one of the structures is designed for handling temporal relations using synaptic delays. The second one, through gaussian functions in the synapses, endows the networks with the capacity of recognizing particular objects in images independently of the background. By integrating these gaussian synapse networks in a global visual architecture, this detection becomes independent of position, orientation and scale. Finally, the third network presented is based on the use of persistence by means of the implementation of habituation neurons as input nodes of networks. 1. Introduction Mobile Robots are very good examples of systems that can be autonomous. They interact with their environment performing actions in order to achieve objectives as a function of perceptions and previous actions. Obviously, as Van de Velde [12] indicates, for a system to be autonomous it must organize its own internal structure in order to behave adequately with respect to its goals and the world, that is, it must learn. Learning involves several aspects that affect the cognitive and physical structure of a robot. It must be carried out all the way from the organization of perceptual information to the synchronization of the actuation of the robot. Several approaches have been employed for the implementation of learning structures in robot cognitive systems, but the most successful paradigm in this area has been that of artificial neural networks. Many systems, specially in the realm of behavior based robotics, implement artificial neural networks that can be trained for perceptual tasks, control tasks, actuation tasks or, more often, for a combination of these in the form of some type of complete behavior that directly links perception and actuation. In general, the ANN systems governing the behaviors have been 1 This work was funded by Xunta de Galicia under project PGIDT99PXI10503A. D-Facto public., ISBN 2-930307-00-5, pp. 81-88 B orks , ES tw r 0 A Ne u 0 N l g 0 N ra e 2 '2 Neu s l 000 l i icia ( r p tif B p ro Ar e A ce on l edi m g 28 ngs iu i - - pos u 6 E ym m 2 uro S ) pean ,