Begin: Tue, 02. of Jun 2015 ( 6:00 PM)
Location: Warburger Str. 100, Lecture Hall O2
On June 2, 2015, Dr. Sebastian von Mammen from the University of Augsburg will give a talk about "Ready for Action: Challenges of Swarm-Based M&S for Real-World Applications" in the context of the SFB 901 colloquium.
Abstract:
Computationally modelling systems based on their individual, often reactive components bears many promises: Consider, for instance, the possibility to predict novel emergent behaviours of the overall system, to allow for open systems in terms of numbers, heterogeneity and configurability of the components, or to tracing and understanding the arising dynamics based on local interactions. Yet, such swarm-based modelling & simulation approaches bear several challenges as well. In this talk, I present my research efforts towards overcoming some of these challenges, including the accessibility and scalability of building and simulating swarm-based models.
We have been working on a multi-level behavioural representation that allows for simple model abstraction and visual programming right in the simulation space. Since recently, we have begun to translate these efforts into the actual object spaces, where, for instance, swarms of quadcopters survey spaces or work on architectural facades. The communication with an actual swarm, to visualise its current state and to allow the user to re-configure it on the fly, can be realised by means of augmented reality technologies. Beneath the surface, the swarm dynamics have to be computed efficiently to ensure smooth realtime interactions with the user. Yet, the multitude of interaction possibilities, and ever-changing topologies among large numbers of swarm individuals let the computational costs soar. Therefore, we have been working on algorithms which detect patterns in the simulated processes and which simplify the model accordingly while the simulation is running. Repeated abstraction can lead to level of detail (LOD) hierarchies that dynamically adjust in accordance with the changing situation complexity and the required simulation accuracy.
