摘要: Learning Classifier Systems use reinforcement learning, evolutionary computing and/or heuristics to develop adaptive systems. This paper extends the ZCS System improve its internal modelling capabilities. Initially, results are presented which show performance in a traditional learning task incorporating lookahead within rule structure. Then mechanism for effective without external reward is examined enables simple system build full map of task. That is, shown learn under latent scenario using scheme. Its ability form maps tasks then considered.