Computing with activities II: ruling robots by the activity patterns of hierarchical sums
Matthias Reuter, B. Rosendo · World Automation Congress · 2004
In this paper a first example is given, in which way robots can be steered by global activity patterns of self-organizing maps (SOMs). In the context of the Computing with Activities a hierarchical neural network structure was trained which codes the diverse situations of (simulated) soccer playing robots not by a winner neuron but by the whole activity structure of the SOMs.