The subject “kinematics“ concerns the systematic description of movements and movement apparatuses on the basis of their geometric connections when it comes to chronological sequence and is thereby a fundamental subject of robotics and automation technology. In the context of eRobotics, MMI is additionally working on developing techniques for modelling, simulation, analysis and controlling for this. Using these techniques enables the desired kinematics – be it 3D objects, robots typical to industry or anthromorphic structures – to be modelled uniformly and move along predefined trajectories in the simulation or in real-life. Building on this, extensions or optimizations are developed in order, for example, to offer application-specific forms of presentation for trajectories, generate collision-free robot paths automatically or to create humanoid-like movements. The results of these techniques are used, among others, to control and drive robots directly or in order to generate ideal nominal values for further algorithms.
“Multi agent systems” (MAS) allow for emulating the hierarchical and heterarchical structures of complex kinematics, but can be expanded to cover automated systems in general. At MMI, we develop highly modular control system architectures based on MAS, in which agents are self-contained controller entities that implement unified interfaces to virtual and real subsystems at once. Each agent represents an element from the real system. Sensor agents have the task to trigger sensor devices, e.g. camera systems or laser scanners. Actuator agents control robots and grippers as well as other actuated hardware, such as fixtures or switches. Product agents maintain virtual representations of the products and offer interfaces for the other agents, e.g. to query product positions and dimensions.