Ute C. Sigmund and Thomas Stutzle
To survive in a reed world environment, a reasoning agent has to be equipped with severed capabilities in order to deed with incomplete information about its target domain. One aspect of the incomplete information is the uncertainty about the effects of actions and events. Although the concrete outcome of an action or event cannot be predicted, often their relative likellhood can be quantified. In this paper we will discuss planning under uncertainty exploiting the available probabi] istic information in dynamic worlds, especially considering delayed effects of actions modeled as causally invoked events.