Schnieder, E.; Hübner, M.; Lück, T.:
Engineering of Car2Car Interactions by Means of Colored Petri Net Road Models.
WCTR - 12th World Conference on Transport Research, Lisboa, Juli 2010.
This contribution presents a new approach to traffic flow optimization on highways by means of vehicle sided rule-bases. Regarding present traffic, global traffic dynamics, expressed by state variables as density and flow as well as phenomena as traffic jams, is a direct consequence of the microscopic behavior of the vehicles. The latter depends on the driver‟s behavior that additionally diverges from vehicle to vehicle. A homogenization shall be a remedy to unintentional behavior of traffic according to Helbing, D. (1997). This may be realized by means of common vehicle sided rule-bases. For this purpose driver‟s behavior is initially not considered and autonomously driving vehicles are assumed. Every vehicle possesses an exact positioning system, robust longitudinal and lateral control, as well as an ad-hoc network adapter designated to car2car communication.
The major objective in this article is presenting a fundamental concept for the engineering of a common vehicle-sided rule-base. Based on the formal concept of vehicle classes and clusters, which are roughly speaking groups of spatially allocated vehicles, rule bases are developed by implementing the standard consensus algorithm for the coordination of the microscopic variables velocity and longitudinal distance (see Hübner et al. (2009a)). Decision-making is realized by spatial discretization of the highway to permissible positions, which are formally represented by means of Petri nets. Its places refer to permissible positions, whereas its transitions denote possible interactions (see Hübner, M., Lück, T. and Schnieder, E. (2009b)). Due to the introduction of different vehicle classes, it is reasonable to model the vehicle-vehicle-interactions by means of colored Petri Nets with a common places‟ capacity that equals one.
A desired global behavior of clusters may generally be achieved by choosing the timing and corresponding subset of transitions that shall fire. One approach is to choose the latter manually out of an engineering perspective. From this set of firing transitions it is possible to formally derive a vehicle-sided rule-base. The developed method presented in this paper is based on the reachability analysis of Petri Nets. In addition, by the automatic generation of a truth table a direct coding of the required vehicle-sided rule-base is possible. Several algorithms, which generate different global behaviors for encountering of clusters and managing passing maneuvers are demonstrated. Those algorithms are compared with respect to a formal definition of traffic safety. The latter depends on the basic states of the present vehicle formation and mean velocity, as well as on the number of interactions per coordination.
The Petrinet based formulation of the coordination problem is a quite natural method for the derivation an optimal vehicle-sided rule-base.
Keywords: multi agent systems, automatic control of traffic, cooperative control