In this paper we propose a trajectory planning Approach for autonomous vehicles on structured road maps. Therefore we are using the well-known A* optimal path planning algorithm. We generate a safe optimal trajectory through a three-dimensional graph, considering the two-dimensional position and time. The graph is generated dynamically with fixed time differences and flexible distances between nodes, based on the vehicle's velocity, using a structured road map. Furthermore the position of dynamic obstacles is predicted over time along the road lanes. The proposed Flexible Unit A* (FU-A*) algorithm was tested for real-time applications with execution times of less than 50 MS on the car's mail Computer. The feasibility and reliability of FU-A* is validated by implementing on simulated autonomous car of Freie Universität`s "MadeInGermany" using the roadmap of Tempelhof, Berlin.