Integration of agents and planning systems

Agents and planning systems represent two modern intelligent techniques. Although these two techniques can be very successfully combined, not much work has been done on their integration. One of the main reasons for this is the fact that these two kinds of systems evolved in an independent manner. Planning systems started their evolution in the early 1960’s and 70’s when systems GPS (General Problem Solver) and STRIPS (STanford Research Institute Problem Solver) [Fikes, Nilsson 1971] were created. In 1989, ADL (Action Description Language) [Pednault 1989] was defined, and this language was the base for PDDL (Planning Domain Definition language) [McDermott et al. 1998], [Fox, Long 2003] the most commonly used planning language today. Another planning standard that is still in use is HTN (Hierarchical Task Network) [Nau et al. 2003], which was created in 1994. The agent paradigm was created in the early 1990’s and quickly became one of the most important fields of investigation in artificial intelligence. Very often, it is combined with other techniques to successfully solve a given problem. This paper will present how this paradigm can be efficiently combined with planning systems to solve complex tasks.