Due to recent advancements in technology, computer simulation has become a necessary tool for transportation professionals to model, analyze and optimize traffic. Unfortunately, they are usually overwhelmed by the size and detail of the simulation results, which prevent them from making high level conclusions, such as identifying bottlenecks and how routes change during heavy traffic.
Additionally, a typical transportation study usual deploys multiple simulation runs because various parameters and alternatives need to be tested until desired traffic solution(s) emerge. Without an effective visualization approach, it may be difficult to identify the optimal solution(s)/alternative(s).
Using a simple visualization of the raw data, plotting individual vehicles travelling through the road network, might make it possible to recognize such occurrences in very conspicuous cases. However, in general, this na?ve approach is not very reliable or informative due to its lack of quantitative measurements.
The objective of this work is to provide visualization methods that highlight the differences in simulated traffic data. Examples in this work will focus on analyzing properties of traffic trajectory data relevant to route planning and route selection, and then graphically display these results for the user. For example, the developed tools allow users to study and visualize the impact of a road or bridge closure by highlighting the changes in optimal routes through the network, as well as changes in the actual routes taken by drivers. In addition, as the time scale in these simulation can range from minutes to decades, using the proposed visualization tool, traffic experts can see the changes in traffic flow decades apart after a bridge is built, see the divergence of highway traffic into the regional network minutes after an accident, and see the effect of a short-term road closure. Although we will use only simulated data in this paper, our technique is also suitable for historical data.
Our technical contributions include (1) mixed trajectory clustering using graph-based Hausdorff distance, which provides significant improvement over the traditional geodesic distance measurement, and (2) in-flux and out-flux difference analysis of traffic flow, which highlights the divergence and convergence of traffic flow in two simulations.