Implementation Methods to model High Traffic Loads

摘要

The continuous increasing traffic volume within the recent years leads to novel appearances of traffic characteristics. Clearly, a proper way of modeling these conditions is prerequisite for an adequate design of simulation and prediction tools. Although the permanent rising use of vehicles often implies a very high degree of bursty traffic behavior, great efforts are still taken to adapt the implemented ARMA and Markovian prediction tools to the current situation. However, theses models possess the disadvantage of smoothing out turbulent traffic due to its finite memory. In contrary, for high traffic loads long-range dependent (LRD) processes have proved to be the superior, adequate alternative in similar other fields like electronic network traffic (e.g. TCP/IP and ethernet traffic), video-on-demand streaming, stock market prediction etc.[4][9]. Thus, this paper first gives a precise, unique understanding into the mathematical background of LRD processes. Based upon this knowledge, appropriate, economic implementation techniques for real traffic control systems will be developed. Particular attention is laid on two major points: first, economic constraints require cost-efficient solutions for already existing systems. So the mainly used ARMA/Markovian tools will be upgraded to LRD systems by the expansion to long-memory FARIMA processes. Second, importance has to be laid to obtain a clear relation between abstract, mathematical parameters and existing boundary conditions. Here, the superposition of heavy-tailed distributions, each of them representing the local traffic condition of individual roads, gateway entries etc.leads to the desired Fractional Brownian Motion. Therefore, turbulent conditions, e.g. on highway corridors, can be depicted by the result of its incoming traffic.