This paper proposes a graph design for the action planning of autonomous mobile robots in cities. When moving in city environments, not only movements but also various other actions such as opening a door or crossing the street need to be considered. Autonomous robots are required to plan their routes while balancing the cost of both movements and required actions. To enable this, we developed the "action node graph, "a graphical representation of a robot's mobility environment. Autonomous robots can use this action node graph to obtain the optimal route to the destination and determine the required actions according to their specifications. An action node graph can be easily constructed from geospatial information and is automatically converted into a behavior tree to be used for the autonomous navigation of a variety of mobile robots. We used a wheeled robot to demonstrate autonomous navigation around crossings and buildings.