Wolfgang Stolzmann: Learning Robots using Anticipatory Classifier Systems
Anticipatory Classifier Systems (ACS) are learning classifier systems that learn by using the cognitive mechanism of anticipatory behavioral control which was introduced in cognitive psychology by Hoffmann (1993). A stepwise introduction to ACS is given. First, the basic learning algorithm, the anticipatory learning process (ALP) is introduced. It enables an ACS to learn an internal world model. The ALP can be combined with a genetic algorithm (GA) so that the ACS is able to learn an internal world model that consists of maximally general rules. Second, an application in robotics is presented that consists in a simulation of an experiment about latent learning in rats. Latent learning is defined as learning in the absence of environmental reward. Therefore latent learning cannot be simulated with usual reinforcement learning techniques.
generated: 10 May 2000; last update: 10 May 2000 / WST