ASSESSING ATMS/ATIS PERFORMANCE USING STOCHASTIC DTA MODELING OR TMC FIELD DATA

This paper addresses aspects of the Advanced Traffic Management Systems and Advanced Traveler Information Systems (ATMS/ATIS) strategies and gives examples of recommended treatments. The test method used in this work is simulation based, using stochastic assignment of destinations and multiple replications that could be applied equally well to mathematical programming formulations. The modeled demands and network are idealized and used as illustrations of valid statistical treatments that would also be applicable to Traffic Management Center (TMC) evaluation of field data. To address the problem of unrepresentative peak period performance, multiple simulation replications are used. The use of 20 multiple replications in simulation is analogous to a month of AM or PM peak period weekdays. Using field observations, a TMC's statistical sampling of performance measures by vehicle probes or roadway sensors might be used in place of the exhaustive sampling possible in a simulation model. When Monte Carlo simulation is applied to modeling of ATMS/ATIS strategies or formulations, appropriate statistical treatments must be applied in determining if alternative strategies actually are significantly different. These statistical treatments include single-factor analysis of variance for normally distributed results and nonparametric methods, such as the Kruskal-Wallis test, for maximum values that would not be normally distributed. Performance can be treated as having demand and supply components. In this paper the driver, or demand-side, is treated as having a trip-based orientation. The average, variance, and maximum values of trip time as performance measures are examined for their statistical treatment of significant differences. Also, an estimation of two standard deviations of mean trip time is used to augment maximum trip time comparisons among strategies. The demand-side trip oriented significance tests in simulations might be extended to the case of TMC field evaluations by using limited sampling of probe vehicles. The network, or supply-side, is examined primarily with respect to density on the critical arcs of the network. Network performance is designed to capture the critical dynamic states of the network's supply utilization over time. The maximum density conveys the worst conditions on any arc. The number of arc-minutes that density is above a critical free flow density conveys the duration of congested conditions. The number of density-arc-minutes that density is above a critical free flow density conveys a measure of total congestion. This density data can be made available for a TMC, as well as in simulation. An example is given of the use of the Kruskal-Wallis test statistic in finding significant differences in maximum densities for four routing strategies.