Introduction to Ph.D. Thesis

Over the last decades, the world economy has constantly grown resulting in a continuously increasing demand for transportation. This is true for both goods and people. Mobility, as a sign of independence, is commonly regarded as one of the most important factors of the quality of life. In 1995, there were 40.4 million vehicles registered in Germany which corresponds to one vehicle for every other citizen. It is not surprising that traffic has become an important factor in our daily lives. Lately, however, we have been experiencing more and more of the downside of vehicular traffic. As street networks have not been extended at such a pace as the demand for transportation, the occurrence of traffic jams has become a major problem in densely populated areas. The resulting delay has direct and indirect consequences:

Time spent ``on the road'' can generally not be used for productive work. People who have to travel professionally lose part of their income through traffic jams. Also, there is an indirect economic effect: being confined to a car in a traffic jam is a stressful situation which reduces the ability to work efficiently once the work place has been reached.
Vehicles stalled in traffic jams use additional fuel. This is not only another cost factor to drivers, but also causes damage to the environment through pollution.
Recurrent congestion largely restricts the independence of drivers. Instead of adjusting trips to activities, drivers have started to plan their activities carefully in order to avoid congestion.

Since the construction of new streets is either financially not affordable or politically unsupportable (This is especially true for European countries). it has become more and more important to use the existing network more efficiently. There are three different time-scales at which one can influence the efficiency:

Long term considerations (over years) refer to the construction of new infrastructure. Here, it is important to decide where to add the infrastructure to obtain the best overall benefit.
On smaller time-scales like days or weeks, one can try to analyze the impact of temporary measures such as road constructions or road blocks. Moreover, sudden, non-recurrent peaks of the demand structure can be investigated: additional traffic caused by simultaneous cultural events (i.e. concerts or sports events).
The lower end of the time-scale is dominated by decisions made just before or during a trip. So-called Advanced Traffic Management/Information Systems (ATMS/ATIS) analyze the current traffic situation and give advice to drivers on how to avoid traffic jams.

In all cases, traffic simulation has proven to be an important tool to model and analyze the traffic state. In this thesis, we will concentrate on different aspects of traffic simulation on modern computer architectures.

In Chapter 2 we will describe models handling traffic on a single multi-lane street segment. The Cellular Automaton (CA) model for traffic flow on a link developed by Nagel and Schreckenberg will serve as a starting point. The CA rule-set will be extended to allow for multi-lane traffic to capture lane-changes. We will present results obtained from simulations of traffic on loops with periodic boundary conditions.

In Chapter 3 the CA multi-lane traffic segment will be used as one of the building blocks of the micro-simulation PAMINA for traffic in street networks. We will also present the structure of other components, such as highway junctions and city intersections. As a first test we will use a route-set consisting of individual routes generated for the TRANSIMS case-study to drive the simulation. Traffic lights and speed-limits will serve as switches to influence the ``fidelity'' of the simulation. We will investigate the phenomenon of grid-locks and how they are influenced by the choice of fidelity.

PAMINA will be used in Chapter 4 to iteratively adapt route-sets from a given initial route-set based upon an origin-destination matrix. The link travel-times generated by the micro-simulation will serve as feedback to improve the dynamic performance estimate for each individual link. We will see that the speed of the relaxation process can be monitored effectively by inspecting the overall travel time of all vehicles in the simulation area (called study-area). We will identify parameters that accelerate the relaxation process without having an impact on the final state. Two artifacts which occurred during some iterations and their remedies will be pointed out. The chapter will be concluded by a comparison of three micro-simulations processing route-sets for the same study-area.

In Chapter 5 the previously generated route-sets will be used to conduct an experiment of online routing. During a simulation run a certain fraction of all drivers will be allowed to access online information. If a driver forecasts an increased trip time due to an impending congestion, a shortest path algorithm computes a sufficiently good alternative route. The main issues of this chapter will be the questions, (a) how much the quality of the re-routing process will depend on the market saturation of the online service, and (b) how much impact (both positive and negative) the system will have on drivers who do not have access to the re-routing service.

The computational aspects of traffic simulation will be the main issue of Chapter 6. Based upon simple parameters which can be obtained from the street network and the underlying hardware, an upper bound for the parallel efficiency will be derived. Also, this chapter we will present benchmark results of the PAMINA micro-simulation both for dynamic load-balancing and static load-balancing with external load feedback.

In Chapter 7 we will summarize the work presented in this thesis and discuss the results obtained from the simulation runs. Whenever applicable, we will give suggestions as to how the micro-simulation can be improved in future experiments.

There are two appendices. The first appendix deals with the technical implementation of the PAMINA micro-simulation. The second one contains details related to traffic simulation and the iterative process used in Chapter 4.

Aktualisiert am 26.11.2006 19:26 Uhr