Vanasse Hangen Brustlin, Inc.

Many existing approaches for estimating cycle-by-cycle traffic volume from vehicle trajectories require either a high market penetration rate of sampled trajectories or auxiliary intersection data (e.g., detector counts or calibrated parameters), largely because they rely only on queued vehicles. In contrast, the motion of nonqueued vehicles-particularly slowed vehicles approaching a dissipating queue-contains informative spatiotemporal cues about queue dynamics. This paper proposes a deep learning-based approach that uses slowed connected vehicles (CVs) to estimate cycle-by-cycle traffic volume under low market penetration. Using a prior-trajectory data set with signal timing information only, a multilayer perceptron (MLP) learns (1) a spatiotemporal capacity-state speed surface, (2) the speed of the last queued vehicle, and (3) the queued-vehicle position (queue length). In operation, the trained models use one slowed probe vehicle per cycle to infer queue length and compute cycle volume without detector counts or externally calibrated shockwave parameters. In simulation (30 representative cycles; 10% market penetration), the method achieved 88.2% accuracy [mean absolute percentage error (MAPE) = 11.8%]. On real-world Federal Highway Administration (FHWA) Next Generation Simulation (NGSIM) trajectories, the method achieved 76.43% accuracy (MAPE = 23.6%). Overall, the results indicate that acceptable volume estimates can be obtained at market penetration rates at which existing approaches typically require higher penetration or additional data sources.

The Interstate System is a fully access controlled surface transportation network connecting diverse geographical areas for the movement of people and goods. It is of national interest to preserve and enhance this system. Service interchanges are critical facilities linking State and Local roads to the Interstate System. When the need arises to insert a new interchange or modify an existing interchange, the request is presented in the Interchange Justification Report (IJR). Although many IJRs are submitted annually, the types of interchanges proposed are limited. Improvements in traffic operation due to proposed interchanges are usually well-described based on traffic analysis results. However, the safety performance of proposed interchanges is often not well substantiated due to a lack of safety modeling tools. Historic data indicates that high-traffic facilities like interchanges exhibit stable traffic crash patterns over time, meaning they should be predictable. Although many factors are known to induce traffic crashes, the safety community hasn't fully grasped how to properly quantify those factors. Decomposing an interchange into simple components and adding up the components' annual crash predictions generally doesn't yield results that match field data. The Federal Highway Administration (FHWA) developed the Interchange Safety Analysis Tool (ISAT) and supported the development of its enhanced version, ISATe. Both of these tools decompose interchanges into simple components for analysis. The FHWA recently completed a study treating the service interchange as a whole and using input data typically available in the planning phase to predict the safety performance of planned service interchanges. It includes the safety predictions of eight interchange configurations representing 78 percent of all interchanges considered in IJRs reviewed by FHWA. This paper presents the functionalities of this tool and explains how the tool may be used. This tool gives the IJR reviewers a consistent methodology for assessing the safety performance of proposed service interchange projects.

Vehicle miles traveled (VMT) is an essential input for many aspects of transportation engineering, and an accurate estimation of VMT is critical for practicing engineers. Linear regression models are a popular method to estimate VMT as they provide insight into the relationships between VMT and other external factors. In linear regression models the prediction of the response variable has a non-zero probability of resulting in a negative value. For this reason, the natural logarithm of VMT is often used as the response variable to force a positive outcome. However, these log-linear regression (LLR) models provide median VMT estimate instead of the mean estimate. To overcome this limitation of LLR models, this study proposes using heteroskedastic LLR and count data methods to estimate VMT. These methods are found to have better performance than LLR models in terms of data fit and prediction accuracy.

Dual left turn lanes are typically operated with protected left turn signals, which means extra delay during non-peak hours compared to permissive operation. To reduce this delay, the authors designed, installed, and tested a "dynamic left turn intersection" (DLTi), which is a new way to operate dual left turn lanes. With DLTi, the both lanes are operated with protected phasing during the peak hours when higher capacity is needed, and only the leftmost left turn lane remains open for protected-permissive operation during off-peak hours. The test showed substantial delay savings, the crash experience has been minimal, and the public comments received have been predominantly negative but not overwhelming. Around 85 percent of left-turning motorists complied when the rightmost left turn lane was closed. The team believes that the DLTi test has been a success, and that agencies should begin searching for other suitable locations for DLTi installation.