Andreas Poxrucker, Gernot Bahle, Paul Lukowicz


Collaborative learning in collective adaptive systems is an active, open research area. In the Allow Ensembles project, we investigate this problem by a component called Evolutionary Knowledge. One problem arising in this context is that concepts of collaborative learning can hardly be studied without an actual real-world system. In this paper, we present our concept of a simulation tool of a real-world urban traffic system used as a framework to investigate collaborative learning. In contrast to existing ready-to-use traffic simulators, its purpose is not the accurate simulation of microscopic or macroscopic traffic flow models. Instead, it is used to generate data to train a knowledge model learning context parameters and their interrelations, which cannot be deduced from an analytical description of the system, but arise as emergent properties from the complexity of the system. Using the simulation we want to investigate the effects of different collaborative learning strategies on emergence in a complex urban mobility system applying different knowledge exchange patterns among entities. We describe the need for and the area of application of our simulator, show the differences to existing traffic simulation tools, and present an outline of its conceptual architecture.   [Download]


@inproceedings {Poxrucker:Towards:2014:7833,
	number = {}, 
	month = {}, 
	year = {2014}, 
	title = {Towards a Real-World Simulator for Collaborative Distributed Learning in the Scenario of Urban Mobility}, 
	journal = {}, 
	volume = {}, 
	pages = {44-48}, 
	publisher = {IEEE}, 
	author = {Andreas Poxrucker, Gernot Bahle, Paul Lukowicz}, 
	keywords = {learning (artificial intelligence);road traffic;traffic engineering computing;Allow Ensembles project;collaborative distributed learning;collective adaptive systems;evolutionary knowledge component;knowledge exchange pattern;learning context parameters;ma}