An Organic Architecture for Traffic Light Controllers Fabian Rochner 1 , Holger Prothmann 2 , J¨ urgen Branke 2 , Christian M ¨ uller-Schloer 1 , Hartmut Schmeck 2 1 Institute of Systems Engineering Universit¨ at Hannover Appelstr. 4 30167 Hannover, Germany {rochner,cms}@sra.uni-hannover.de 2 Institute of Applied Informatics and Formal Description Methods Universit¨ at Karlsruhe (TH) 76128 Karlsruhe, Germany {hpr,jbr,hsch}@aifb.uka.de Abstract: Efficient control of traffic networks is a complex but important task. A successful network management vitally depends on the abilities of the traffic light controllers to adapt to changing traffic situations. In this paper a control architecture for traffic nodes is presented that is inspired by the principles of Organic Comput- ing. It allows a node to quickly adapt to changing traffic situations and enables it to autonomously learn new control strategies if necessary. 1 Introduction Control of road traffic networks in urban areas is a challenging task. This is mainly due to the great dynamics of traffic. Therefore, a mere generation of an optimal controller for a certain traffic situation is not sufficient. It is necessary to provide the capability of adjusting quickly to changes in traffic situations and, in particular, to react reasonably in situations that had not been anticipated by the designer of the traffic controller. In this paper we show how different principles of Organic Computing [Sch05] like self- organization, self-adaptation and the Observer-Controller paradigm can contribute to im- prove traffic controllers, making them at the same time more flexible and easier to set up, to maintain, and give better results. First ideas that have been incorporated into this architecture have been presented in [RMS04]. From the extensive literature on traffic control, the recent works of Bull et al. [BS + 04] and Helbing et al. [HL + 05] should be mentioned in the context of this paper. Similar to the ideas presented here, Bull uses a Learning Classifier System (LCS) to control a simplistic traffic node, while Helbing proposes a decentralized control strategy for traffic flows (which not yet considers legal regulations like minimum or maximum green times). 2 Motivation A central requirement for any control architecture for a traffic node is to guarantee the safe functionality of the node at any time, ensuring that conflicting traffic streams are 120