Multilevel Testing of Control Software for Teams of Autonomous Mobile Robots Sebastian Petters, Dirk Thomas, Martin Friedmann, and Oskar von Stryk Technische Universit¨ at Darmstadt, Department of Computer Science Hochschulstr. 10, D-64289 Darmstadt, Germany { petters, dthomas, friedmann, stryk }@sim.tu-darmstadt.de, WWW home page: http://www.sim.tu-darmstadt.de Abstract. Developing control software for teams of autonomous mobile robots is a challenging task, which can be facilitated using frameworks with ready to use components. But testing and debugging the resulting system as teached in modern software engineering to be free of errors and tolerant to sensor noise in a real world scenario is to a large extend beyond the scope of current approaches. In this paper multilevel test- ing strategies using the developed frameworks RoboFrame and MuRoSimF are presented. Testing incorporating automated tests, online and offline analysis and software-in-the-loop (SIL) tests in combination with real robot hardware or an adequate simulation are highly facilitated by the two frameworks. Thus the efficiency of validation of complex real world applications is improved. In this way potential errors can be identified early in the development process and error situations in real world op- erations can be reduced significantly. 1 Introduction Development of control software for teams of autonomous robots imposes many challenges on the developer. The software is usually highly complex, containing modules for very different tasks (like motion generation, sensor data fusion or behavior control). To ensure operation of such systems, each module of the con- trol software must (1) be free of errors and (2) tolerate noise and errors from other sources. A special class of robots targeted at in this paper are ”lightweight” robot systems characterized by inertially stabilized high motion dynamics and limited onboard sensing and computing capabilities due to payload restrictions like small humanoid robots, small unmanned aerial or marine vehicles. As autonomous mobile robots are operated in environments with large uncer- taines, the software must be tolerant to noise and disturbances. To examine the abilities of an autonomous robot all individual modules of the control software as well as the complete system have to be tested extensively. Testing the software for autonomous mobile robots is a complicated challenge, which can only be met if the developer is equipped with appropriate tools. One major problem when testing such software is the fact, that the source of an error is often not obvious. An error can usually be caused by one of the modules involved, or it can be Preprint of a paper which appeared in the Proceedings of the 2008 International Conference on Simulation, Modelling and Programming for Autonomous Robots (SIMPAR) Venice, Italy, Nov. 03 - 07 2008