Mobile autonomous service robots operate in a highly dynamic environment, requiring a capable high-level planning system to achieve their tasks. In this paper the characteristics of task planning systems for autonomous service robots are discussed. RGOAP, a robotic planning system based on the concept of goal oriented action planning, is developed. It is implemented as a Python library with three packages. A third-party-independent core package, an adapter package for the robotic software framework ROS and an adapter package for the state machine library SMACH. Finally, RGOAP is used successfully to add an autonomous behaviour to the mobile service robot Scitos.