Zusammenfassung:
In the development of multi-robot systems, simulation plays a crucial role. On the one hand, simulation can save money, because physical robots are very valuable and damage to them results into high maintenance costs. On the other hand, experiments with physical robots are very time consuming, since usually only one experiment at a time is conducted and there is no possibility to fast forward when there is no interesting robot behavior at the moment. Here, damage to the robots results into a delay of the experiment. Furthermore, it usually isn’t easy to create dynamic environments and controlled failures such as faulty sensors or actuators within the robots. Even if a trivial solution to these problems is found, it’s very laborious to conduct a number of experiments with the exact same changes in the environment and/or failures. It is often desired to repeat an experiment with identical conditions, for example to test di erent compositions of robot teams and compare their performance under those conditions. This thesis presents a simulator that makes it possible to conduct simulation runs with (possibly automated) changes in the environment or failures within the robots. As a basis, the Swarmulator, a simulator for swarm robotics, which was developed by Martin Burger within the scope of his diploma thesis [Bur12] is used and contributed the main part of the internal logic of the simulator. However, it needed to be extended to design experiments with the aforementioned environmental changes or robotic failures. The ALLIANCE architecture, as introduced by Parker in [Par98] is implemented as the mechanism for the coordination of a team of robots. Additionally, this thesis also discusses the experimental results of a case study, which has been adopted from Parker [Par98].
Download:
thesis_gutschale.pdf