Mechatronics is a current trend in the development of technical systems. By integrating cognitive components, mechatronic systems can be equipped with inherent partial intelligence making autonomous and goal-directed action feasible. The result is advanced information processing, which allows the systems to independently and proactively plan the required behavior of their function modules (behavior planning) to fulfill a single order. Because planning is a complex problem, and the given real-time condition of a mechatronic system allows only a short time to carry out a planning process a short-term planning horizon is given. In addition this is so far limited by the objective of the current single order. Information about possible follow-up orders, which is assigned to the system beyond the limited planning horizon, is missing. Thus the system cannot autonomously adapt the planning of individual orders in the course of a further execution or the scope of required resources in advance. This thesis contains a concept to solve this problem based on the schema of hierarchical planning. The challenge is the integration of a closed-loop circuit to regulate behavior and in the definition of the necessary interdependencies. During the rolling behavior planning executed (partial), plans are recorded and a predictive model for the behavior anticipation is built up. The behavioral and result-oriented anticipation, as well as the classification of the collected execution characteristics (especially of states) are of particular interest.