Organizational Unit:

School of Civil and Environmental Engineering

Permanent Link

https://hdl.handle.net/1853/70760

Parent Organization

Organizational Unit

College of Engineering

Includes Organization(s)

Organizational Unit

Georgia Water Resources Institute

Organizational Unit

Georgia Transportation Institute

Organizational Unit

Environmental Microbial Genomics Laboratory

ArchiveSpace Name Record

https://finding-aids.library.gatech.edu/agents/corporate_entities/385

Full item page

Publication Search Results

Now showing 1 - 10 of 47

Emergency Vehicle Preemption Strategies using Machine Learning to Optimize Traffic Operations

(Georgia Institute of Technology, 2023-04-30) Roy, Somdut

Emergency-Response-Vehicles (ERVs) operate with the purpose of saving lives and mitigating property damage. Emergency-response Vehicle Preemption (EVP) is implemented to provide the right-of-way to ERVs by displaying the green indications along the ERV route. Two EVP strategies were developed as part of this effort. First, a strategy was developed, defined as “Dynamic-Preemption” (DP), that utilizes Connected-Vehicle (CV) technology to detect, in real time, the need for preemption prior to the ERV reaching the vicinity of an intersection. The DP strategy is based on several generalized traffic demand and simplified traffic flow assumptions. Second, a machine learning approach was utilized to develop an EVP call strategy that sought to (1) preemptively clear queues at intersections prior to ERV arrival, (2) create a "delay-free" path for the ERV, and (3) minimize excess delay to the conflicting traffic in the event of an EVP call. The ML approach utilizes currently available vehicle detection data streams and is trained based on simulated EVP scenarios. Existing field strategies and the developed strategies were tested under varying scenarios, on a simulated signalized corridor testbed. It was observed that the proposed methodologies showed tangible improvement over the existing baseline algorithms for EVP, both in terms of ERV travel time and delay to the conflicting movements. In summary, this research is expected to lay the foundation for use of novel computational approaches in solving the EVP problem in traffic ecosystems with limited CV penetration, with the aid of microsimulation.
Development of a real-time connected corridor data-driven digital twin and data imputation methods

(Georgia Institute of Technology, 2020-07-09) Saroj, Abhilasha Jairam

Smart cities -- equipped with connected infrastructure -- receive significant real-time traffic data. The simulation platform developed in this research leverages these high-frequency connected data streams to derive meaningful insights on the current traffic state by providing near real-time corridor performance measures. A data-driven traffic simulation model, i.e. digital twin, capable of providing environmental and traffic performance measures in near real-time is developed for a connected corridor. The data streams driving the developed simulation model are traffic volumes and signal indications. The research demonstrates the feasibility of the overall connected corridor simulation approach. In addition, investigation of the real-time data streams from the connected corridor revealed the presence of data gaps. Such data gaps can impact the simulation generated performance measures. This research investigates the sensitivity of simulated performance measures to data loss and data imputations developed to infill the detector stream gaps. The impact of data stream gaps on the simulated performance measures is seen, in part, to be dependent on the combination of intersection approaches experiencing data loss. This combination effect can be attributed to both the vehicle volumes observed at these approaches and the ability of the approaches to process additional vehicles. The corridor location of the intersection approaches that have missing data, as well as the travel path of interest, also influence performance measure accuracy. The research demonstrates that to successfully leverage real-time high frequency connected corridor data streams for (near) real-time applications, it is crucial to develop data imputation methodologies that can both learn from historically available data and adapt to recent data trends. In this research, a Long Short Term Memory Recurrent Neural Network layers approach, modeling univariate and multivariate time series data, is developed for data imputation. Experiments are conducted to compare the performance of the univariate and multivariate models and to investigate the impact of these imputation approaches on the simulation performance measures. The findings show the potential advantages of using a multivariate model approach for imputations over a univariate model under atypical traffic conditions. Results also suggest better performance of the univariate model to impute missing data under typical traffic conditions. Future work includes additional development of the model using increased training and validation data along with hyper parameter tuning to increase robustness of the model performance.
ESTIMATING THE EFFECT OF VEHICLE SPEEDS ON BICYCLE AND PEDESTRIAN SAFETY ON THE GEORGIA ARTERIAL ROADWAY NETWORK

(Georgia Institute of Technology, 2020-07-09) Arias, Daniel F.

Despite a decreasing trend in overall crashes, bicyclist and pedestrian fatalities have increased steadily since 2009 in the United States (Cicchino & Hu 2016). A large body of research suggests vehicle speeds are a key contributing factor for crashes (Elvik et al. 2019). Furthermore, vehicle impact speed has been identified as the principal determinant of severity and death in the event of a pedestrian crash (Tefft 2013). However, there have been few studies of bicycle or pedestrian crash probability that incorporate detailed vehicle speed data. Newly available probe vehicle data in the state of Georgia makes it possible to study the relationship between bicycle and pedestrian crashes and speed across the network of Georgia arterial roadways. The analysis uses INRIX® speed data and the Georgia DOT crash database and relates these data in a Negative Binomial crash count model for the year 2017. Models using speed percentiles (85th, 50th and 15th) and models using speed differences (85th - 50th and 50th - 15th percentile) are compared. A small set of covariates are included. This study shows that the high speed difference (85th - 50th percentile) is a robust indicator of bicycle and pedestrian crash frequency on Georgia arterial roadways. The high speed difference outperformed the low speed difference (50th - 15th percentile), suggesting that the high end of the distribution is more important to crash prediction than the low end. Additionally, speed percentile models showed no clear, intuitive relationship to bicycle and pedestrian crashes. In light of these results, planners and policymakers should identify arterial roadways with high speeds, high spread of speeds at the top end of the distribution, and high bicyclists and pedestrian activity. To do so, a complete bicycle and pedestrian count data collection effort is needed. These target roadways should be considered for treatments which prioritize the reduction of the fastest speeds and limitation of exposure for unprotected road users. Finally, the practice of setting the speed limit at the 85th percentile speed (NTSB 2017) should end. Road user safety must supplant vehicle throughput and access to create a sustainable, equitable and just transportation system in Georgia.
Modeling the impact of road grade on vehicle operation, vehicle energy consumption, and emissions

(Georgia Institute of Technology, 2018-08-08) Liu, Haobing

Motor vehicle emissions and their impacts on local air pollutant concentrations are a primary concern in cities. Properly quantifying energy and emissions is the key step in identifying the major sources of air pollution, evaluating whether transportation activities are consistent with air quality goals, and providing decision makers with reference for implementation of new policies for sustainable development. Mathematical models are commonly used to predict vehicle energy consumption and emissions. Vehicle-specific power (VSP) is widely used in such models to evaluate engine load, and it is represented as a function of vehicle mass, vehicle dynamic parameters (rolling/drag coefficient), driving behavior (speed and acceleration) and road conditions (gravitational acceleration and road gradient). In the U.S. Environmental Protection Agency’s (USEPA’s) MOVES (MOtor Vehicle Emission Simulator) model, speed and VSP levels are tied to vehicle energy consumption and emission rates. Detailed and accurate speed-acceleration joint distributions (SAJDs, also known as Watson plots) can be used to reflect onroad activity required for calculating the distribution of activities in MOVES VSP and speed bins, and thus for estimating vehicle energy consumption and emissions. Road grade is also a critical variable that affects engine operations, as uphill grades require that the engine perform additional work against gravity in the direction of vehicle motion (while downhill grades obtain an energy benefit). Real-world vehicle speed and acceleration can be easily collected using low-cost global positioning system (GPS) data loggers, on-board diagnostics (OBD) system data loggers, and smartphones apps. But, The effect of road grade is usually ignored in emission modeling. On the other hand, very little attention has been paid to the interaction between real-world road grade and onroad activity patterns and the resulting impact on energy use and emissions. However, road grade is expected to impact vehicle operations due to drivers’ response to uphill and downhill driving, or due to vehicle mechanical performance. It is currently unclear that how speed and accelerations vary across different road grade levels, and how the interaction of driver behavior and road grade affect engine power, energy consumption, and emissions modeling. This study is directed at answering two issues: 1): how road grade impacts vehicle speed and acceleration distributions, and how such distributions vary across vehicle types, roadway types, traffic conditions, etc., and 2): how significant the impact of integrating grade interactions is with respect to energy, emissions, and air quality modeling.
Safety analysis of centerline rumble strips along rural two-lane undivided highways in Georgia

(Georgia Institute of Technology, 2017-12-14) Pena, Marisha S.

Vehicle crashes involving crossing over the roadway centerlines are among the most severe types of collisions nationwide. To address this issue, the Georgia Department of Transportation (GDOT) started implementing centerline rumble strips (CLRS) in rural locations across Georgia in 2005 and 2006. CLRS produce both an audible and tactile warning to alert drivers of impending lane departure into the lane of oncoming traffic. As of 2015, approximately 200 miles of CLRS have been installed by GDOT as a countermeasure for crossover crashes along rural two-lane undivided highways. This study evaluates the safety impacts of CLRS deployments in Georgia by analyzing two years of before and two years of after periods to evaluate the safety impacts associated with nine treatment sites and a control group of comparison sites with similar traffic and physical characteristics. The study dataset consisted of 154 target crashes along 126.46 miles of CLRS treatment sites and 1,391 crashes along control group sites. The empirical Bayes method was used to develop a crash modification factor for CLRS of 0.66, indicating a 34% reduction in crashes involving centerline crossings associated with the installation of centerline rumble strips. The sample size of fatal and injury crashes was too small to obtain separate crash modification factors for fatal crashes and injury crashes. The favorable crash modification factor (0.66) found in this study supports wider use of centerline rumble strips as a safety measure to address crashes involving vehicles that cross the centerline of the roadway. In addition to the safety analysis, this study also provided insights into the crash reporting process by conducting a comprehensive manual review of more than 17,000 crash reports. Approximately 6% of target crashes were found to be misclassified due to coding errors.
Analysis of trucking variability in roadway network energy using basic safety message data

(Georgia Institute of Technology, 2017-12-11) Bolen, John

This experiment uses VISSIM to replicate the message broadcasted by connected vehicles and plugs it into an energy calculator in order to determine how the energy usage of a vehicle fleet changes as the truck percentage of the fleet changes. By replicating the connected vehicle message, it also allows researchers to determine the extent to which connected vehicle data can be used in future experiments. This experiment began with the building of a microscopic simulation traffic model the North Avenue Corridor in Atlanta, GA, modeling the signal timing, traffic volumes, and overall characteristics of all 19 signalized intersections within the three mile corridor. With this done, the model was run ten different times for each of seven different fleet compositions, each with a different percentage of single unit delivery trucks and tractor trailers. The data files directly outputted into VISSIM were then processed in such a way that they mimicked the standardized message broadcasted by connected vehicles. After this, the processed files were run through the energy calculator in order to determine the energy for each vehicle type as well as for the entire fleet. From this experiment, it was determined that adding more trucks to a vehicle fleet has a small but definite change in the per-vehicle energy for passenger cars. The per-vehicle change for trucks was larger than that of cars, but due to extreme variability in the truck results, the extent to which increasing truck percentage affects trucks is inconclusive. Future research into this topic should include much larger sample sizes than ten runs per fleet composition, and should include more fleet compositions in the range of 10% trucks to 50% trucks. Future research may also include sampling the connected vehicle replica data to determine the expected sample error from various connected vehicle market penetration rates.
Sensitivity analysis of operational performance under conventional diamond interchange and diverging diamond interchange

(Georgia Institute of Technology, 2017-12-11) Park, Sung Jun

Rapidly growing traffic volumes and changes in traffic patterns over time have forced many intersections and interchanges into sub-optimal operation. Diverging diamond interchange (DDI) is one of many innovative interchange designs currently being proposed and implemented to better accommodate these changes. This study compares the operational performance of a conventional diamond interchange (CDI) and a DDI at different traffic volumes and turning movement combinations, and explores conditions for which one interchange design may be more advantageous over the other. To achieve this objective, traffic simulation models built using the microscopic simulation software, VISSIM, and procedures involving the Critical Lane Volume (CLV) method were used to conduct sensitivity analyses at different traffic conditions and to explore differences in delay, travel time, queue length, number of stops, and volume-to-capacity ratio between the two interchange designs studied. The results of the study show that the DDI will have better operational performance at high cross street traffic volumes with high left-turn ratio (above 50%), while the CDI will perform better at low cross street traffic volumes with low left-turn ratio (below 30%). The through/left proportion where the CDI and DDI has similar performance is dependent to the cross street cross sections. This study is one of the first to examine in detail the parameters and conditions that are best accommodated by the DDI related to conventional interchanges. Findings from this study can support planning and decision making processes associated with the implementation of DDIs.
Developing freeway merging calibration techniques for analysis of ramp metering In Georgia through VISSIM simulation

(Georgia Institute of Technology, 2016-05-03) Whaley, Michael T.

Freeway merging VISSIM calibration techniques were developed for the analysis of ramp metering in Georgia. An analysis of VISSIM’s advanced merging and cooperative lane change settings was undertaken to determine their effects on merging behavior. Another analysis was performed to determine the effects of the safety reduction factor and the maximum deceleration for cooperative braking parameter on the simulated merging behavior. Results indicated that having both the advanced merging and cooperative lane change setting active produced the best results and that the safety reduction factor had more influence on the merging behavior than the maximum deceleration for cooperative braking parameter. Results also indicated that the on-ramp experienced unrealistic congestion when on-ramp traffic was unable to immediately find an acceptable gap when entering the acceleration lane. These vehicles would form a queue at the end of the acceleration lane and then be unable to merge into the freeway lane due to the speed differential between the freeway and the queued ramp traffic. An Incremental Desired Speed algorithm was developed to maintain an acceptable speed differential between the merging traffic and the freeway traffic. The Incremental Desired Speed algorithm resulted in a smoother merging behavior. Lastly, a ramp meter was introduced and an increase in both the freeway throughput and overall speeds was found. Implications of these findings on the future research is discussed.
Optimal Ramp Metering of Freeway Corridors

(Georgia Institute of Technology, 2015-11-19) Chilukuri, Bhargava Rama

Ramp meters have been used for congestion management on freeways since the 1960s to maximize freeway capacity by controlling on-ramp flows. Traditionally, the focus has been to develop rule-based algorithms and optimal control case studies. This led to a host of algorithms and methods which cannot be proven to provide an optimal control and the case studies does not provide a systematic understanding of the characteristics of optimal control and its influence on traffic dynamics. Moreover, optimal is not easy to achieve in practice due to the limited storage on the on-ramps. Towards this end, this dissertation systematically studies the optimality conditions for the case of unlimited storage and spatiotemporal evolution of control and its corresponding traffic dynamics on freeway and ramps under queue constraint, carefully taking the traffic dynamics into account. A Kinematic Wave model of the freeway-ramps system is optimized for minimal total delay. The optimality conditions for the case of unlimited ramp storage are studied using Moskowitz functions that provide several interesting insights for different scenarios, including the case of limited storage. This dissertation shows that the current problem posed as a nonlinear coupled PDE system with a nonlinear merge model cannot be solved analytically. This study also shows that the discrete-time nonlinear formulation solved with simulation-based optimization does not converge in reasonable time. To overcome this, the problem is reposed as a LP formulation that includes capacity drop. For discrete formulation, this study develops an error-free solution to the KW model with a source term that enhanced the quality of the numerical solution. This study identifies four distinct regions in the state surface with distinct metering patterns. Explicit modeling of ramps enabled correlating the initialization and termination times of the metering patterns with the evolution of traffic dynamics on the freeways and ramps. Using these results, this dissertation presents a hybrid isolated ramp metering algorithm that outperforms existing methods.
Evaluating comprehension of temporary traffic control

(Georgia Institute of Technology, 2015-11-18) Greenwood, Aaron T.

There are over 5 million reported motor vehicle collisions annually in the United States, and while crash rates and fatality rates have declined in the past decades, rates in work zones are disproportionately high. There are strict standards for evaluating the crashworthiness of temporary traffic control devices, but not for evaluating drivers’ comprehension of existing or novel device deployments. This dissertation presents a series of three experiments evaluating driver comprehension for existing and novel traffic control devices conducted in a work zone setting. This evaluation is further expanded by decomposing the task of comprehending traffic control into the three subtasks of detection, localization, and identification. Methods are proposed for conducting a computer-based experiment with still image stimuli to measure participant performance at each of these subtasks. Next, procedures for categorizing localization responses and accounting for variation in participants physical responses are explored. Lastly, an application of Item Response Theory toward the evaluation and comparison of participant comprehension is demonstrated. It is hoped that these methods and procedures can be used by future researchers and experimenters to compare novel temporary traffic control devices and systems to inform future design.