Chih-Sheng Chou

Performance Measures for Adaptive Signal Control: Case Study of System-in-the-Loop Simulation

The simulation of local signal controllers has become increasingly sophisticated in recent years and has been paralleled by improvements in the integration of adaptive systems into simulation. This paper describes and demonstrates an emerging methodology for the evaluation of adaptive signal control that is termed “system-in-the-loop simulation.” This methodology extends existing software-in-the-loop simulation by linking virtualized traffic controllers with real-world adaptive-control systems. In addition, the authors propose an analysis methodology that fuses data on simulated probe vehicles with data on high-resolution controller events. Through this data fusion, traditional measures of simulation performance such as delay can be enhanced with operational measures of performance that characterize quality of progression and capacity utilization. In addition, adaptive-control performance can be characterized in relation to overall impact on traveler delay and also described in terms that are meaningful for improvement of control schemes. An example case study is presented: the ACS-Lite adaptive system was tested on a 19-intersection system in Morgantown, West Virginia, under a special-event scenario. Free, fully actuated control was compared with traditional time-of-day and traffic-responsive control both with and without the use of the adaptive-control system ACS-Lite. Overall delay results are presented and contrasted with more detailed analysis of event-based performance measures at a single intersection and on a networkwide basis.

Recommended Procedure to Adjust Inaccurate Weigh-in-Motion Data

Weigh-in-motion (WIM) stations are used to monitor traffic volumes and loadings on roads and bridges. When a station is installed, a vehicle of known axle weights is typically used to calibrate the sensors. Because of environmental changes, pavement changes, sensor fatigue, and repeated loadings, the initial calibration doesn’t always maintain its validity. Therefore, the WIM station output must be monitored so that potential problems can be identified and corrected. It is common for analysts to discover that WIM data are inaccurate when they are attempting to use the data for design or planning. An analyst must then decide whether to conduct the analysis with inaccurate data or to not conduct the analysis at all. There is a need to develop a procedure that allows inaccurate WIM data to be adjusted and remain usable. Adjustment of WIM data after they have been collected may not be an acceptable practice to all agencies or for certain applications. A new accuracy metric was developed that relates Class 9 front axle weight with gross vehicle weight with data from the Long-Term Pavement Performance (LTPP) program. This new metric and other existing accuracy metrics in the literature will serve as baselines for adjusting the data. The adjustment procedure involves time series analysis, which eliminates temperature-induced variations in weight measurements that are known to occur with certain WIM sensors. The adjustment procedure is applied to an LTPP WIM station to illustrate the data before and after adjustment.

Characterizing Emergency Vehicle Preemption Operation with High-Resolution Traffic Signal Event Data

This paper proposes the use of a signal phase spectrum plot to analyze high-resolution traffic signal event data. Specifically, the controller performance related to emergency vehicle preemption operation is characterized in an effort to identify performance measures that will allow a traffic engineer to understand better the impact that various configurations have on intersection operations. Performance measures for individual intersections in coordinated systems include preemption duration, transition duration, and total interruption time. Performance measures for networks are based on an emergency vehicle reidentification process for deriving an emergency vehicle’s trajectory through a network, and the results can further be used to estimate travel time, travel speed, and origin–destination. These performance measures are illustrated by simulated signal system in Morgantown, West Virginia. Transition modes are varied in the simulation network to determine the relative performance measures. Case studies are presented for using high-resolution data to troubleshoot field preemption operation using signal systems in Morgantown and Huntington, West Virginia.

Methodology to Map Routes from Truck Permit Database, Using a Linear Reference System and Network Analysis

Trucking companies can obtain permits from state agencies allowing them to haul oversize and overweight (OSOW) loads. Most permit systems assign a specific route to be followed that accounts for various network travel restrictions. The routes assigned in these permits can provide a wealth of information for different types of transportation analyses. Unfortunately, the permit data are not always generated by a geographic information system (GIS) or archived in a manner that facilitates analysis in a GIS platform. This paper presents a methodology that was developed to process archived permit records from the West Virginia Department of Transportation (DOT) so that the records could be imported into a GIS and plotted by using the existing West Virginia DOT linear referencing system (LRS). Some agencies do not have a GIS-based OSOW permitting system, and those that use a GIS permitting system are still faced with the challenge of integrating archive data with existing GIS–LRS platforms. The methodology presented here should be widely applicable for those facing such challenges. The automated procedure was able to assign an LRS code and map 91.4% of the permits that contained route data for the month of July 2011.

Evaluation of triangabout as alternative for intersection with nonthrough arterial movement

At the majority of signalized intersections along an arterial, the mainline movement passes straight through the intersection. At four-leg intersections where the major arterial movement is not a through movement (i.e., an L-shaped corridor vertex), the intersection can sometimes experience more congestion than adjacent intersections along the corridor because the two minor approaches, which are adjacent conflicting movements, must be served in a split-phase fashion. A new, unconventional intersection design, referred to as a “triangabout,” is proposed to improve mobility and enhance safety for such intersections. The two approaches that are the mainline arterial form two legs of a triangle and a new diagonal section of roadway connecting those two legs is the hypotenuse. Flow within the triangabout is one-way counterclockwise, similar to a roundabout. The three intersections at the vertices of the triangle are signalized to provide a dedicated right-of-way. A case study is performed for the intersection of Van Voorhis Road and Chestnut Ridge Road in Morgantown, West Virginia, along the WV-705 corridor. Simulation results from VISSIM show that the triangabout design can reduce travel delay by 50% compared with the existing configuration. Also, the number of conflict points compared with a conventional intersection is reduced by 34%. There are a number of qualitative benefits of this design compared with conventional intersections at an L-shaped corridor vertex.

Reidentification of Trucks on Basis of Axle-Spacing Measurements to Facilitate Analysis of Weigh-in-Motion Accuracy

This study examined weigh-in-motion (WIM) data from two states to evaluate the performance of an improved reidentification methodology that had been used to match vehicles between WIM stations. The improvement allowed the reidentification model to be trained without the use of ground truth data (i.e., true vehicle matches). The training data set was instead developed by following a three-step manual investigation of the characteristics of assumed vehicle matches between two WIM stations. The trained reidentification methodology was validated with data from Oregon, where the model was able to match identical vehicles with 70% to 90% accuracy. The reidentification was then applied to data from two WIM stations in West Virginia, where the downstream station had 6,178 vehicles with 10,427 possible matches at the upstream station. At a high confidence threshold of .98 (of 1.0), the algorithm identified 526 likely matches. Furthermore, this study examined the differential calibration of the weight sensors at each WIM location by comparing the axle weight measurements between the matched vehicles. The data from the two WIM stations in Oregon illustrated good correlation in weight measurements. However, the two WIM stations in West Virginia did not illustrate consistent relationships across all axle measurements.

Empirical Analysis of Controller Event Data to Select Vehicle Detector Fault Triggers

The proper detection of both vehicles and pedestrians is critical for operations of signalized intersections, especially intersections with adaptive traffic signal control. Faulty detectors must be identified as quickly as possible so that they can be repaired and their erroneous output corrected (e.g., preventing an adaptive algorithm from continuing to allocate green time to a detector that is stuck on). This study establishes a procedure to determine static detector fault thresholds (e.g., maximum detector on time and maximum detector off time) that can be programmed into a traffic signal controller. This procedure analyzes detector on and off durations over multiple days to define distributions of those parameters. The study analyzes data for 154 detectors in Morgantown, West Virginia, to generate on and off distributions and calculate the resulting 90th, 95th, and 99th percentile values. The detectors are divided into three groups on the basis of their physical locations relative to the intersection (e.g., stop bar or advance) and the approach function (e.g., mainline or minor). Default thresholds for the three groups are recommended for the signal system through the application of a 95% confidence limit to the 99th percentile values and local knowledge of the system. The procedure defined here can be applied to other signal systems to determine these thresholds.