Common use of RESEARCH BACKGROUND Clause in Contracts

RESEARCH BACKGROUND. Aside from their negative impacts on traffic safety, freeway incidents have been identified as one of the major causes for congestion. According to a Federal Highway Administration (FHWA) research report (Cambridge Systematics, Inc., 2005), approximately 50 percent of congestion on freeways is non-recurrent congestion caused by incidents (25 percent), work zones (10 percent), and bad weather (15 percent). The situation is even more severe in urban areas. Approximately one-half to two-thirds of the total travel delay in large metropolitan areas is incident-related (Center for Urban Transportation Research, 2005). Since congestion mitigation and safety enhancement are among the main goals of most transportation agencies, a number of state (and local) departments of transportation have invested in incident response (IR) programs in a variety of forms. In Washington state, the Washington State Department of Transportation (WSDOT) established its Incident Response Program in collaboration with the Washington State Patrol (WSP) and the Washington State Association of Fire Chiefs (WSAFC), a group called the Washington Traffic Incident Management Coalition (WaTIMCo). Besides prioritizing responder and motorist safety, one of WaTIMCo’s goals also involves congestion mitigation when incidents occur. Estimation of IID is highly desirable for the following reasons: • Measurement of IID is important in assessing the effectiveness of congestion countermeasures. • IID estimates help engineers understand the impacts of various types of incidents under various traffic and roadway conditions. • Accurate IID estimates can help in identifying appropriate decisions regarding IR so that limited monetary and labor resources can be allocated to maximize its benefit-to- cost ratio (BCR). • IID estimates are key components of incident cost calculations and are essential for the development of active traffic management and integrated corridor management strategies. However, it is not an easy task to quantify IID because existing traffic sensors cannot directly measure IID, and algorithms are needed to estimate IID by using available traffic sensor measurements. Therefore, IID estimation has become a hot research field. Most research efforts have been based on either Deterministic Queuing Theory (DQT) or shock wave analysis. The former calculates IID by using a queuing diagram formed by cumulative vehicle arrival and departure curves. The area enclosed by the two curves represents the total delay. The latter involves several attempts to apply the methods of kinematic waves to explain the characteristics of traffic flow, which lead to the development and application of shock wave analysis for estimating IID. In Washington state, ▇▇▇▇▇▇▇▇▇▇ et al. (2003) developed a loop-occupancy-based algorithm to identify incident occurrence and to estimate the impacts of incidents on freeways. Although it was easy to apply, this algorithm suffered from false alarms in terms of incident detection because loop-measured lane occupancy is not always a good indicator of actual traffic conditions. Also, IID estimated with this algorithm could be subject to another source of error because of the use of point-sensor measured speeds for sectional travel time calculations. In 2008, ▇▇▇▇ et al. (2008) applied a DQT-based approach for estimating IID. This approach improved the accuracy of IID estimates by using traffic data (e.g., vehicle count and loop occupancy) and a Dynamic Volume-Based Background Traffic Profile (BTP) instead of a fixed occupancy background profile to quantify IID. Use of this approach significantly improved the incident detection rate in comparison to results from the loop-occupancy-based method. The accuracy of the DQT-based approach was evaluated by using microscopic traffic simulation models. However, the performance of this DQT-based approach was not stable. Further investigation found that IID estimated from this approach was sensitive to the travel time estimation from the upstream loop station to the incident location. To improve IID results, the estimation accuracy of the space-mean speed or travel time from the upstream loop station to the incident location had to be enhanced. This would be fairly difficult, given the sporadic deployment of detectors and sensor measurement constraints. Hence, the first challenge to improving IID estimation using DQT methods was that accurate speed or travel time data would be difficult to obtain. Not many traffic sensors deployed on the existing freeway system are capable of measuring traffic speed or travel time. For example, most existing traffic detectors on Washington freeways are single loops. Although Athol’s method (Athol, 1965) can be used to estimate traffic speed from single-loop output, the accuracy may not be great when a significant number of long vehicles are present (▇▇▇▇ and ▇▇▇▇▇, 2000). Furthermore, capturing each vehicle’s speed requires loop event data (high resolution loop status data), which are available only in traffic controllers unless special data collection devices, such as the Advanced Loop Event Data Analyzer (▇▇▇▇▇) (Cheevarunothai et al., 2006), are used to record the data. The second challenge to using the DQT-based approaches for quantifying IID was to separate recurrent traffic delay from the total delay under incident scenarios. ▇▇▇▇ et al. (2008) applied a BTP matching approach to find a traffic volume series in an incident-free scenario to match the current volume series under incident impact. Once an acceptable match had been found, its volume series at the downstream loop was applied to calculate recurrent delay. The difference between total delay and recurrent delay was IID. However, not all traffic arrival series have an incident-free matching series in the historical database. Also, the matching process is computationally expensive. Therefore, the BTP matching approach may not always be capable of quantifying the delay introduced by an incident. A more robust IID estimation approach was desired. As a continuation of an earlier study (referred to as the Phase I study in this report) by ▇▇▇▇ et al. (2008), this study (or Phase II) sought to develop a new IID estimation approach to overcome the two challenges described above.