Decomposition of a Traffic Flow Model for a Parallel Simulation

With advent of Intelligent Transportation Systems (IST), it is required that the observation and control of traffic demand and flow be performed in real-time with more accuracy and reliability than before. In order to accelerate the traffic flow simulation the paper presents an algorithm for the decomposition of a mesoscopic traffic flow model for a parallel implementation on multiprocessor hardwareplatforms with distributed memory. The purpose of the this algorithm is to ensure a nearly equal distributed workload on all physical processors on the one hand and to provide possibly large coherent regions with minimal interconnections on the other hand. Therefore different optimization criteria have been considered. The used traffic flow model DYNEMO estimates spatial traffic variables, such as mean speed and density as well as speed and acceleration of individual vehicles. The model considers the flow on different lanes and the influence of traffic light control and is well suited for large city networks. Finally, first simulation results for the network of the Berlin-Brandenburg region are given. 1 THE TRAFFIC FLOW MODEL DYNEMO The mesoscopic traffic flow model DYNEMO (Schwerdtfeger 1984), as shown below, combines the advantages of microscopic and macroscopic models. For each stretch in the network the model needs as input the relationship between traffic density ρ and mean speed mean v . This relationship (Schmidt and Schäfer 1997) is called fundamental relation of the road. The relationship reflects the fact that the behaviour of a driver strongly depends on how many vehicles he/she notices on the road and particularly on the distance to the vehicle immediately in front. The maximum value of traffic density depends upon the vehicle length, and on the distance bumper to bumper (spacing when they come to a stop). The minimum value characterizes the free flow state where the mean speed is nearly independent of the traffic volume. A basic assumption of the model is that each individual vehicle-speed at a given traffic condition characterized by its density varies within the interval [ ] ) ( ), ( ρ ρ v v . The actual speed v of an individual car depends on its desired speed wu v and is restricted by the given interval. The equation of the actual speed is