The use of recursively generated iterated structures to constrained neural network architectures

Summary form only given, as follows. Previous work applied the principles of constrained embryologies to reduce the complexity of the neural networks controlling structures with attributes of symmetry and segmentation. The authors have extended the treatment to include iterated structures generated by recursive procedures. Expressions were derived to show the reductions in search space achieved by this strategy. A four-legged walking platform was used to illustrate the order of complexity of the problems found in developing practical robots on realistic time-scales