| To be effective, intelligent transportation systems (ITS) need timely and accurate information for decision making. This dissertation work addresses the information available---and the information needed---from proven, public sector sensors such as being used in the major upgrade of New York City DOT's ITS infrastructure for traffic control.;That ITS infrastructure will include several thousand intersection controllers designed to NYC specification (the ASTC controller), a wireless communications backbone, a network of video and microwave detectors, and a new software environment and joint traffic management center (JTMC) for implementing the system. This researcher has been deeply involved in these efforts.;There are three foundation questions of critical importance in making cost effective use of this ITS environment: (1) How precise do the measurements have to be, in order to exercise effective control, and what metrics have to be measured?, (2) How sparse can the detector network be, and still be effective?, (3) When---if at all---is responsive control necessary to realize significant improvements?;The work and recommendations presented in this dissertation work address the first two questions, and lay the basis for addressing the third.;The dissertation work suggests that flow and occupancy observed at key locations in a network can be used to deduce demand sufficiently accurately for an effective feedback control system. There are imbedded issues of the use of data smoothing, and the risks of over-smoothing, that are addressed in the dissertation work. Information from 4 data sets at test locations within NYC was used.;The dissertation work can make a contribution to the literature not only because it lays the foundation for effective control, but also because it will in fact be implemented in a real, large, ITS environment.;Based upon the analysis shown in this dissertation, it can be said that: (1) occupancy/queue is the key to action, but the precise values of these indicators are not needed for action; (2) because flow observations will generally not allow demand to be observed due to the metering effects of the signal plan, demand has to be estimated by {flow + Delta}, where Delta is influenced by observed occupancy/density/queue; (3) accuracy is best thought of as "sufficiency for action" and not for precisely replicating the ground truth data, at least for the purpose of control; (4) because of the dynamics of traffic where timely response is needed, the data measurement period should be 180 seconds or less, while at the same time it should be 30 seconds or more due to the communications burden within a wireless system.;This dissertation expands upon these points, particularly with regard to (a) the implications for how control can be achieved within these findings/assertions, (b) a synchronous framework for collecting data, with regard to sensor information, (c) a matrix of recommended actions using the above points. The dissertation work concludes with a framework for control actions based upon demand and occupancy (or queue), at a pace suited to practical levels of detectors in a large ITS environment, distinguishing between areas that have highly regular patterns and those that need more responsive response.;The unique contribution of the research lies in (1) Using non-intrusive devices that do not require high maintenance, once calibrated; (2) Not using high number of detectors, and focusing on the critical intersections in dynamic environments where advanced traffic control can be useful, or on a "by exception" basis; (3) Taking advantage of both occupancy and flow rate, which in combination both reasonably accurate for the needs of the control and in thus providing the foundation for appropriate policies, concurrent with providing the operational environment. |
