Mohammed Elhenawy

An Intersection Game-Theory-Based Traffic Control Algorithm in a Connected Vehicle Environment

Urban traffic congestion is a growing problem that we experience every day. Intersections are one of the major bottlenecks that contribute to urban traffic congestion. Traditional traffic control methods, such as traffic signal and stop sign control are not optimal for all demand levels as demonstrated in the literature. Recently, numerous research efforts proposed Intelligent Transportation System (ITS) applications to enhance intersection capacity and hence reduce congestion. In this paper we propose a game-theory-based algorithm for controlling autonomous vehicle movements equipped with Cooperative Adaptive Cruise Control (CACC) systems at uncontrolled intersections. The goal of this research effort is to develop an algorithm capable of using the future autonomous/automated vehicle capabilities to replace the usual state-of-the-practice control systems at intersections (e.g. stop signs, traffic signals, etc.). The proposed algorithm is chicken-game inspired and is efficient for application in real-time. It assumes vehicles can communicate with a central agent at the intersection to provide their instantaneous speeds and locations. The proposed algorithm assumes that vehicles obey the Nash equilibrium solution of the game. The simulation results demonstrated reductions in vehicle travel time and delay relative to an all-way stop sign control in the range of 49 and 89 percent on average respectively.

Modeling driver stop/run behavior at the onset of a yellow indication considering driver run tendency and roadway surface conditions

The ability to model driver stop/run behavior at signalized intersections considering the roadway surface condition is critical in the design of advanced driver assistance systems. Such systems can reduce intersection crashes and fatalities by predicting driver stop/run behavior. The research presented in this paper uses data collected from two controlled field experiments on the Smart Road at the Virginia Tech Transportation Institute (VTTI) to model driver stop/run behavior at the onset of a yellow indication for different roadway surface conditions. The paper offers two contributions. First, it introduces a new predictor related to driver aggressiveness and demonstrates that this measure enhances the modeling of driver stop/run behavior. Second, it applies well-known artificial intelligence techniques including: adaptive boosting (AdaBoost), random forest, and support vector machine (SVM) algorithms as well as traditional logistic regression techniques on the data in order to develop a model that can be used by traffic signal controllers to predict driver stop/run decisions in a connected vehicle environment. The research demonstrates that by adding the proposed driver aggressiveness predictor to the model, there is a statistically significant increase in the model accuracy. Moreover the false alarm rate is significantly reduced but this reduction is not statistically significant. The study demonstrates that, for the subject data, the SVM machine learning algorithm performs the best in terms of optimum classification accuracy and false positive rates. However, the SVM model produces the best performance in terms of the classification accuracy only.

Enhanced modeling of driver stop-or-run actions at a yellow indication

The ability to model driver stop-or-run behavior at signalized intersections is critical in the design of advanced driver assistance systems. Such systems can reduce intersection crashes and fatalities by predicting driver stop-or-run behavior. The research presented in this paper used data collected from a controlled field experiment on the smart road at the Virginia Tech Transportation Institute to model driver stop-or-run behavior at the onset of a yellow indication. The paper offers three contributions. First, it evaluates the importance of various model predictors in the modeling of driver stop-or-run behavior in the vicinity of signalized intersections. Second, this paper introduces a new variable related to driver aggressiveness and demonstrates that this measure enhances the modeling of driver stop-or-run behavior. Third, the paper applies well-known machine learning techniques, including k nearest neighbor (k nn), random forests, and adaptive boosting (AdaBoost) techniques on the data and compares their performance to standard logistic models in an attempt to identify the optimum modeling framework. The experimental work shows that adding the driver aggressiveness predictor to the model increases the model accuracy by approximately 10% for the logistic, random forest, and k nn models and by 7% for the AdaBoost model. The paper also demonstrates that all modeling frameworks produce similar prediction accuracies.

An automated statistically-principled bottleneck identification algorithm (ASBIA)

Bottlenecks are key features of any freeway system. The deployment of stationary sensors and proliferation of mobile vehicle probes provides researchers with a wealth of data that can be used for the automatic identification of active freeway bottlenecks. In this paper we introduce an automated statistically principled algorithm to characterize traffic into two states: free-flow or congested and subsequently identify the spatiotemporal activation of bottlenecks. The proposed algorithm uses speed measurements over short temporal and spatial intervals and segments, respectively to identify the status of a segment while accounting for spatiotemporal correlations and interactions. The outputs of the algorithm are the status of the roadway segment (free-flow or congested) and the confidence level of the test (p-value). The experimental results based on archived data from the northbound Interstate 5 (I-5) corridor in the Portland, Oregon, metropolitan region demonstrates significant improvements over state-of-the-art bottleneck identification algorithms.

Investigating cyclist violations at signal-controlled intersections using naturalistic cycling data

Improving bicycle safety is considered as a growing concern for two reasons. First, in the United States in recent years, about 700 cyclists were killed and about 48,000 were injured in bicycle motor vehicle crashes each year. Regarding crash location, from 2008 to 2012 in the United States, more than 30% of cyclist fatalities occurred at intersections. Furthermore, up to 16% of bicycle-related crashes were due to cyclist violations at intersections. Second, from 2000 to 2011, bicycle commuting rates in the United States has increased: by 80% in large Bicycle Friendly Cities (BFCs), by 32% in non-BFCs, and overall by 47%. Also, cycling as one of the sustainable and eco-friendly modes of transport is receiving more attention than before. In this paper, to investigate factors affecting cyclist behavior at signalized intersections, a naturalistic cycling experiment was designed and conducted. Applying mixed effects generalized regression analysis, movement and presence of other users were found as significant factors that influence the probability of red light violations by cyclists. Moreover, several machine learning algorithms were adopted to develop cyclist violation prediction models at signalized intersections. The violation prediction models were developed based on kinetic information of cyclists approaching the intersection. The results showed a promising performance of the prediction models in terms of high true positive rates and low false positive rates.

An Automatic Traffic Congestion Identification Algorithm Based on Mixture of Linear Regressions

One innovative solution to traffic congestion is to use real-time data and Intelligent Transportation Systems (ITSs) to optimize the existing transportation system. To address this need, we propose an algorithm for real-time automatic congestion identification that uses speed probe data and the corresponding weather and visibility to build a unified model. Based on traffic flow theory, the algorithm assumes three traffic states: congestion, speed-at-capacity, and free-flow. Our algorithm assumes that speed is drawn from a mixture of three components, whose means are functions of weather and visibility and defined using a linear regression of their predictors. The parameters of the model were estimated using three empirical datasets from Virginia, California, and Texas. The fitted model was used to calculate the speed cut-off between congestion and speed-at-capacity by minimizing either the Bayesian classification error or the false positive (congestion) rate. The test results showed promising congestion identification performance.

Automatic Congestion Identification with Two-Component Mixture Models

Automatic identification of traffic congestion is an important component of any intelligent transportation system. These systems need computer algorithms to identify current congestion and to predict the evolution of future congestion. The output of the congestion identification algorithm enables various users to be better informed and make safer, more coordinated, and smarter use of transportation networks. A new automatic congestion identification algorithm was proposed; it assumed that the speed data were drawn from a two-component mixture model. The first component represented the speed distribution in congestion, and the second component was the free-flow speed distribution. The proposed algorithm was first calibrated by using historical speed data in a two-component mixture model. A free-flow speed threshold based on the estimated parameters of the free-flow speed distribution was set. Subsequently, a road segment was identified as having free flow if its speed was greater than the threshold and congested if its speed was less than the threshold. The mixture components considered lognormal and gamma-skewed distributions and normal symmetric distributions. The proposed algorithm was tested by using two real data sets collected from two different roadways and was demonstrated to produce good performance.

Travel Time Modeling using Spatiotemporal Speed Variation and a Mixture of Linear Regressions.

Real-time, accurate travel time prediction algorithms are needed for individual travelers, business sectors, and government agencies. They help commuters make better travel decisions, avert traffic congestion, help the environment by reducing carbon emissions, and improve traffic efficiency. Travel time prediction has begun to attract more attention with the rapid development of intelligent transportation systems (ITSs), and is considered one of the more important elements required for successful ITS subsystems deployment. However, the stochastic nature of travel time makes accurate prediction a difficult task. This paper proposes travel time modeling using a mixture of linear regressions. The proposed model consists of two normal components. The first component models the congested regime while the other models the free-flow regime. The means of the two components are modeled by two linear regression equations. The predictors used in the linear regression equation are selected out of the spatiotemporal speed matrix using a random forest machine-learning algorithm. The proposed model is tested using archived data from a 74.4-mile freeway stretch of I-66 eastbound connecting I-81 and Washington, D.C. The experimental results show the ability of the model to capture the stochastic nature of travel time and to predict travel time accurately.

Modeling bike availability in a bike-sharing system using machine learning

This paper models the availability of bikes at San Francisco Bay Area Bike Share stations using machine learning algorithms. Random Forest (RF) and Least-Squares Boosting (LSBoost) were used as univariate regression algorithms, and Partial Least-Squares Regression (PLSR) was applied as a multivariate regression algorithm. The univariate models were used to model the number of available bikes at each station. PLSR was applied to reduce the number of required prediction models and reflect the spatial correlation between stations in the network. Results clearly show that univariate models have lower error predictions than the multivariate model. However, the multivariate model results are reasonable for networks with a relatively large number of spatially correlated stations. Results also show that station neighbors and the prediction horizon time are significant predictors. The most effective prediction horizon time that produced the least prediction error was 15 minutes.

Network-wide bike availability clustering using the college admission algorithm: A case study of San Francisco Bay area

The significant increase in the use of bike sharing systems (BSSs) causes imbalances in the distribution of bikes, creating logistical challenges and discouraging bike riders who find it difficult to pick up or drop off a bike at their desired location. We investigated this issue by finding the network-wide availability patterns and how these patterns evolve temporally using a novel supervised clustering algorithm based on the College Admission and the K-median algorithms. The proposed approach models the clustering problem as a matching problem between two disjoint sets of agents: centroids and data points. This new view of the clustering problem makes our algorithm a multi-objective algorithm where the impurity and distance in each cluster are minimized simultaneously. The proposed algorithm showed promising performance when applied to BSS data for the San Francisco Bay area. The resultant network-wide availability patterns were used to identify imbalances in the BSS. Using a spatial analysis of these imbalances, we propose potential solutions for decision makers and agencies to improve BSS operations and make it more stable.