SAB 2006 | START Conference Manager |
In place of the continuous time recurrent neural network used by Beer we used a one-dimensional ring of cells within which the concentration of two coupled chemicals changed according to two differential equations describing intra-cell reactions and inter-cell diffusion. Output from whisker-like proximity sensors was fed to the cells in the RD-ring via weighted links, perturbing the concentration of the two chemicals. Weighted links in turn allowed the concentration of particular chemicals in designated cells to specify motor activation, completing a sensor-motor loop. In some of the controllers described the RD-ring also received input from the motors. Links were made symmetrically about the animat's longitudinal axis. Parameters specifying the weighted links between cells, motors and sensors were evolved as were the values of a dimensionless feed rate and rate constant for the RD-system. A simple fitness-function encouraged animats to fixate circles and avoid diamonds.
Animats capable of discriminating between similarly sized circles and diamonds were regularly evolved. By then allowing evolution to continue while randomly deleting links we were able to reduce the size of the resulting networks to around 30-40 symmetrical links. Full analysis of the controllers will take a while yet but some progress has been made in this regard.
START Conference Manager (V2.53.9) |
Maintainer: rrgerber@softconf.com |