No abstract

The ability to accurately measure and cost-effectively collect traffic data at road intersections is needed to improve their safety and operations. This study investigates the feasibility of using laser ranging technology (LiDAR) for this purpose. The proposed technology does not experience some of the problems of the current video-based technology but less expensive low-end sensors have limited density of points where measurements are collected that may bring new challenges. A novel LiDAR-based portable traffic scanner (TScan) is introduced in this report to detect and track various types of road users (e.g., trucks, cars, pedestrians, and bicycles). The scope of this study included the development of a signal processing algorithm and a user interface, their implementation on a TScan research unit, and evaluation of the unit performance to confirm its practicality for safety and traffic engineering applications. The TScan research unit was developed by integrating a Velodyne HDL-64E laser scanner within the existing Purdue University Mobile Traffic Laboratory which has a telescoping mast, video cameras, a computer, and an internal communications network. The low-end LiDAR sensor's limited resolution of data points was further reduced by the distance, the light beam absorption on dark objects, and the reflection away from the sensor on oblique surfaces. The motion of the LiDAR sensor located at the top of the mast caused by wind and passing vehicles was accounted for with the readings from an inertial sensor atop the LiDAR. These challenges increased the need for an effective signal processing method to extract the maximum useful information. The developed TScan method identifies and extracts the background with a method applied in both the spherical and orthogonal coordinates. The moving objects are detected by clustering them; then the data points are tracked, first as clusters and then as rectangles fit to these clusters. After tracking, the individual moving objects are classified in categories, such as heavy and non-heavy vehicles, bicycles, and pedestrians. The resulting trajectories of the moving objects are stored for future processing with engineering applications. The developed signal-processing algorithm is supplemented with a convenient user interface for setting and running and inspecting the results during and after the data collection. In addition, one engineering application was developed in this study for counting moving objects at intersections. Another existing application, the Surrogate Safety Analysis Model (SSAM), was interfaced with the TScan method to allow extracting traffic conflicts and collisions from the TScan results. A user manual also was developed to explain the operation of the system and the application of the two engineering applications. Experimentation with the computational load and execution speed of the algorithm implemented on the MATLAB platform indicated that the use of a standard GPU for processing would permit real-time running of the algorithms during data collection...

No abstract

Pedestrians are regarded as vulnerable road users as they experience the most severe consequences in collisions with motor vehicles. Pedestrians were involved in fewer than 4% of all traffic accidents but represented more than 12% of traffic fatalities in Medellin, Colombia, from 2009 to 2016. A better understanding of the factors affecting the likelihood of such accidents is an area of interest across local and governmental agencies in Colombia. The paper provides a spatial lag model estimated using spatial two–stage least squares (S2SLS) method to analyze the impact of land use, socioeconomic factors, and transportation modes on vehicle–pedestrian collisions (“crashes”). Consideration is given to spatial spillovers, avoiding inconsistent and inefficient estimates. Commercial land use has a positive correlation with the number of crashes while residential land use correlates with fewer crashes. Analysis of transportation modes demonstrated a higher risk of crashes associated with public transit. In Medellin, buses stop at the passenger’s request instead of at regular bus stops, which increases the likelihood of pedestrians being involved in accidents. The effect of taxi as transportation mode, however, analysed using the number of taxi stands as proxy, is that the number of crashes is reduced. In this transportation mode, the entire trip is covered from point of origin to destination, reducing exposure to the risks of walking. These results might provide a better understanding of the dynamics involved in pedestrian crashes while validating the importance of including spatial econometrics for safety analysis.

No abstract

To improve traffic performance and safety, the ability to measure traffic accurately and effectively, including motorists and other vulnerable road users, at road intersections is needed. A past study conducted by the Center for Road Safety has demonstrated that it is feasible to detect and track various types of road users using a LiDAR-based system called TScan. This project aimed to progress towards a real-world implementation of TScan by building two trailer-based prototypes with full end-user documentation. The previously developed detection and tracking algorithms have been modified and converted from the research code to its implementational version written in the C++ programming language. Two trailer-based TScan units have been built. The design of the prototype was iterated multiple times to account for component placement, ease of maintenance, etc. The expansion of the TScan system from a one single-sensor unit to multiple units with multiple LiDAR sensors necessitated transforming all the measurements into a common spatial and temporal reference frame. Engineering applications for performing traffic counts, analyzing speeds at intersections, and visualizing pedestrian presence data were developed. The limitations of the existing SSAM for traffic conflicts analysis with computer simulation prompted the research team to develop and implement their own traffic conflicts detection and analysis technique that is applicable to real-world data. Efficient use of the development system requires proper training of its end users. An INDOT-CRS collaborative process was developed and its execution planned to gradually transfer the two TScan prototypes to INDOT’s full control. This period will be also an opportunity for collecting feedback from the end user and making limited modifications to the system and documentation as needed.

Autonomous vehicles must balance a complex set of objectives. There is no consensus on how they should do so, nor on a model for specifying a desired driving behavior. We created a dataset to help address some of these questions in a limited operating domain. The data consists of 92 traffic scenarios, with multiple ways of traversing each scenario. Multiple annotators expressed their preference between pairs of scenario traversals. We used the data to compare an instance of a rulebook [1], carefully hand-crafted independently of the dataset, with several interpretable machine learning models such as Bayesian networks, decision trees, and logistic regression trained on the dataset. To compare driving behavior, these models use scores indicating by how much different scenario traversals violate each of 14 driving rules. The rules are interpretable and designed by subject-matter experts. First, we found that these rules were enough for these models to achieve a high classification accuracy on the dataset. Second, we found that the rulebook provides high interpretability without excessively sacrificing performance. Third, the data pointed to possible improvements in the rulebook and the rules, and to potential new rules. Fourth, we explored the interpretability vs performance trade-off by also training non-interpretable models such as a random forest. Finally, we make the dataset publicly available to encourage a discussion from the wider community on behavior specification for AVs. Please find it at github.com/bassam-motional/Reasonable-Crowd.

RoadHAT is a tool developed by the Center for Road Safety and implemented for the INDOT safety management practice to help identify both safety needs and relevant road improvements. This study has modified the tool to facilitate a quick and convenient comparison of various design alternatives in the preliminary design stage for scoping small and medium safety-improvement projects. The modified RoadHAT 4D incorporates a statistical estimation of the Crash Reduction Factors based on a before-and-after analysis of multiple treated and control sites with EB correction for the regression-to-mean effect. The new version also includes the updated Safety Performance Functions, revised average costs of crashes, and the comprehensive table of Crash Modification Factors—all updated to reflect current Indiana conditions. The documentation includes updated Guidelines for Roadway Safety Improvements. The improved tool will be implemented at a sequence of workshops for the final end users and preceded with a beta-testing phase involving a small group of INDOT engineers.