Federal regulations (23 CFR 630 Subpart J, 23 CFR 630 Subpart K) place emphasis on smart work zone technologies within and around work zones to improve safety and mobility. Given the increasing number of smart work zone deployments, cross-studies have been performed to summarize the lessons learned, and work zone implementation guidelines were recently published by the Federal Highway Administration (FHWA) to assist departments of transportation (DOTs) in determining the feasibility and design of smart work zones for a given application. Two critical components for the success of a smart work zone deployment are the quality of the traffic data collected by sensor networks and the algorithms used for data processing, which, when combined, provide real-time traffic information in the work zone. Accurate and reliable traffic estimation is the basis for many smart work zone systems regardless of the specific application. For example, the effectiveness of a portable changeable message sign (PCMS) is reduced if the message does not accurately correspond to current traffic conditions. Therefore, the accuracy of the traffic estimation can be regarded as a critical metric for the potential effectiveness of smart work zones. Using the estimation accuracy as a metric of the potential effectiveness of work zones, this study focuses on quantitatively evaluating a large variety of sensor network configurations and traffic estimation algorithms in microsimulation to obtain insights on best practices for designing smart work zone systems. Two work zones located on I-80 in Will County and I-57 in Jefferson County were modeled and calibrated with field data in the microsimulation environment. Dedicated sensor error models were developed to generate realistic measurements corresponding to Doppler radar sensors (radar), remote traffic microwave sensors (RTMS), and low-energy radar (LER). To assess the importance of algorithms for the estimation accuracy of the velocity, queue length, and travel time, three algorithms with different levels of sophistication were implemented: (1) spatial interpolation used in practice, (2) spatio-temporal filtering, which integrated a smoothing component in the temporal horizon, (3) and a traffic flow model—based nonlinear Kalman filter. To identify the critical factors on the sensor network design in a smart work zone, 242 different configurations of sensor networks were quantitatively assessed, each with three algorithms that varied in the number and spacing of sensors, the type of sensors, and the accuracy of individual sensors. In summary, this study assessed 726 combinations of sensor network configurations and traffic estimation algorithms. The main findings are as follows: The spacing of sensors is an important factor for improving the accuracy of traffic estimation, especially at a large sensor spacing (e.g., 1 mile). When the sensor spacing is smaller than 0.5 mile, the benefit of additional sensors or the choice of algorithm is marginal (i.e., less than 5% improvement per sensor). The nonlinear Kalman filter generally provides significantly more-accurate estimation of the velocity, the queue length, and the travel time compared with other algorithms when the spacing of sensors exceeds 1 mile. It has the potential to reduce the cost of the existing sensors by approximately 50% while achieving the same level of traffic estimation accuracy. However, the performance of the nonlinear Kalman filter relies on the appropriate selection of algorithmic parameters, which requires field data collection and expertise to apply the technique. The RTMS provides more-accurate flow measurements than radar and LER because of its less prominent occlusion issue. The accurate flow measurement can significantly improve the estimation accuracy of the nonlinear Kalman filtering algorithm. The spatial interpolation and the spatio-temporal filtering algorithms use velocity measurements only; hence, they have less accuracy variation across three types of sensors. In the cost accuracy analysis, the radar sensors are the most cost effective for estimating the velocity and queue length. At the same system cost, the additional number of radar sensors (a lower unit price allows more to be installed) provides higher improvement of the estimation accuracy than using more-accurate but fewer RTMS. It should be noted that the cost accuracy analysis was conducted based on limited cost data. It is recommended that the cost accuracy be re-assessed, given the updated cost data for each specific deployment. Existing sensor technologies are sufficient for good performance across all algorithms considered, and little additional benefit can be expected from improvements of the quality of individual sensors because measurement error is dominated by the quantization error and errors related to occlusion (for radar and LER). This finding is based on the assumption that all sensors are properly calibrated to achieve the error magnitudes as specified by the sensor manufacturer specification and operate reliably. More benefit can be achieved by improving the reliability of sensors instead of increasing individual sensor accuracy. This conclusion is made based on the rate of significant missing data during congestion in the field dataset in a work zone where no traffic estimation algorithm can produce accurate traffic estimates. All classes of implemented algorithms perform relatively poorly on travel time estimation owing to the use of the instantaneous travel time estimation scheme. The use of Bluetooth sensors will not improve travel time estimation accuracy when the travel times are quickly changing. This is because the sensors only record the travel time of the vehicle that just completed the trip, which may no longer be a good estimate of the travel time of the vehicle just entering the stretch of roadway. Analytics with the capability of travel time prediction for smart work zone monitoring systems are recommended to obtain a better travel time estimation. The findings in this study are intended to Help DOTs in their decision-making process regarding the acquisition of smart work zone systems. Assist vendors working with state DOTs on the development of improved systems for smart work zones. All source code developed in this study can be found at https://github.com/Lab-Work/IDOTSmartWorkzone. (Author/publisher)
Samenvatting