Fixed-Time Traffic Control at Superstreet Intersections by Bee Colony Optimization

The superstreet intersection (or restricted crossing U-turn-, J-turn intersection) fixed-time traffic control system was developed in this study. The optimal (or near-optimal) values of cycle length, splits, and offsets were discovered by minimizing the experienced travel time of all network users traveling through the superstreet intersection. The optimization procedure used was based on the bee colony optimization (BCO) metaheuristic. The BCO is a stochastic, random-search, population-based technique, inspired by the foraging behavior of honey bees. The BCO belongs to the class of swarm intelligence methods. A set of numerical experiments was performed. Superstreet intersection configurations that allowed direct left turns from the major street, as well as configurations with no direct left turns, were analyzed within numerical experiments. The obtained results showed that BCO outperformed the traditional Webster approach in the superstreet geometrical configurations considered.

Zero Traffic Conflict (ZTC) Road Networks with Concentric Traffic Paths

U-turns and left turns are sometimes forbidden even though it increases travel distances. The greater travel distances are sometimes outweighed by the improved movement through intersections due to there being fewer conflicting lanes of traffic. One can, further, forbid straight-throughs. Restricting a sufficient number of turns can make intersections free from crossing lanes of traffic (``zero traffic conflict,'' ``ZTC"), though there may still be merging lanes of traffic. It's possible to make \begin{it}all\end{it} intersections in a road \begin{it}network\end{it} ZTC. However, keeping all destinations accessible and travel distances moderate requires careful selection of allowed driving directions and turning directions. We demonstrate through numerical microscopic and macroscopic simulations that there are road networks and ranges of traffic loads for which, in comparison with conventional schemes, ZTC road network can carry approximately $50$\% more vehicular traffic without incurring gridlock.

Driving With Distraction: Measuring Brain Activity and Oculomotor Behavior Using fMRI and Eye-Tracking

IntroductionDriving motor vehicles is a complex task that depends heavily on how visual stimuli are received and subsequently processed by the brain. The potential impact of distraction on driving performance is well known and poses a safety concern – especially for individuals with cognitive impairments who may be clinically unfit to drive. The present study is the first to combine functional magnetic resonance imaging (fMRI) and eye-tracking during simulated driving with distraction, providing oculomotor metrics to enhance scientific understanding of the brain activity that supports driving performance.Materials and MethodsAs initial work, twelve healthy young, right-handed participants performed turns ranging in complexity, including simple right and left turns without oncoming traffic, and left turns with oncoming traffic. Distraction was introduced as an auditory task during straight driving, and during left turns with oncoming traffic. Eye-tracking data were recorded during fMRI to characterize fixations, saccades, pupil diameter and blink rate.ResultsBrain activation maps for right turns, left turns without oncoming traffic, left turns with oncoming traffic, and the distraction conditions were largely consistent with previous literature reporting the neural correlates of simulated driving. When the effects of distraction were evaluated for left turns with oncoming traffic, increased activation was observed in areas involved in executive function (e.g., middle and inferior frontal gyri) as well as decreased activation in the posterior brain (e.g., middle and superior occipital gyri). Whereas driving performance remained mostly unchanged (e.g., turn speed, time to turn, collisions), the oculomotor measures showed that distraction resulted in more consistent gaze at oncoming traffic in a small area of the visual scene; less time spent gazing at off-road targets (e.g., speedometer, rear-view mirror); more time spent performing saccadic eye movements; and decreased blink rate.ConclusionOculomotor behavior modulated with driving task complexity and distraction in a manner consistent with the brain activation features revealed by fMRI. The results suggest that eye-tracking technology should be included in future fMRI studies of simulated driving behavior in targeted populations, such as the elderly and individuals with cognitive complaints – ultimately toward developing better technology to assess and enhance fitness to drive.

Left Turns by Older Drivers With Vision Impairment: A Naturalistic Driving Study

Background and Objectives Older drivers are over-represented in collisions at intersections while making left turns across oncoming traffic. Using naturalistic driving methods, we evaluated the association between vision impairment and their left turn characteristics. Research Design and Methods In this prospective, observational study, vision impairment as defined by visual acuity, contrast sensitivity, visual processing speed, visual field sensitivity and motion perception was assessed in drivers ≥70 years old. Data acquisition systems were installed in their personal vehicles recording video and vehicle kinematics. Driving during everyday life was recorded for six months. Data analysts evaluated a temporal data window surrounding randomly selected left turns at four-way intersections. Left turn traversals and turning behavior were evaluated in terms of age-adjusted associations with vision impairment. Results The sample consisted of 151 older drivers. The number of turns studied was 473; 265 turns were rated as unsafe traversals, and 201 as problematic turning behavior. Drivers with slowed visual processing speed and visual field impairment were less likely to exhibit unsafe traversals (p< 0.05); those with worse contrast sensitivity, slowed visual processing speed, and visual field impairment were less likely to exhibit problematic turning behavior (p <0.05). Discussion and Implications Using naturalistic driving, our study suggests older drivers with vision impairment exhibit better performance in making left turns than those without deficits, which contradicts older driver studies on left turns using driving simulators and on-road driving evaluations. Our findings suggest more cautious and self-regulatory behavior which are consistent with older visually impaired drivers’ commonly expressed concerns about their driving difficulties. Translational Significance In contrast to previous older driver studies using driving simulators and on-road driving evaluations, our study suggests that in everyday life older visually-impaired drivers exercise caution in left-turn behavior across oncoming traffic as compared to those who are normally sighted.

Modelling of Three-Dimensional Intersection Sight Distance

Intersection sight distance(ISD) is an important design element. Each intersection has a potential for several different types of vehicular conflicts that can be greatly reduced through the provision of proper sight distance. Current guidelines do not adequately address sight distance requirements for intersections located on horizontal curves alone or horizontal curves combined with vertical alignments. In many practical situations, however, sight distance is required to be checked for an existing or proposed three-dimensional(3D) intersection alignments. In this thesis, models were developed to check sight (2001) were considered on 3D alignment: (1)Departure from stop-control minor-road and (2) Left-turns from major-road. For stop-control intersections, several cases were addressed. These include Case 1(a): Intersection and approaching vehicle (object) lie on the curve, Case 2: Intersection lies on the tangent and object lies on the curve. For both cases (1) and (2), obstruction may lie inside or outside the horizontal curve and the intersection and object can be anywhere with respect to the vertical alignment. In many practical situations, however, sight distance is required to be checked for an existing or proposed three-dimensional(3D) intersection alignments. In this thesis, models were developed to check sight (2001) were considered on 3D alignment: (1)Departure from stop-control minor-road and (2) Left-turns from major-road. For stop-control intersections, several cases were addressed. These include Case 1(a): Intersection and approaching vehicle (object) lie on the curve, Case 2: Intersection lies on the tangent and object lies on the curve. For both cases (1) and (2), obstruction may lie inside or outside the horizontal curve and the intersection and object can be anywhere with respect to the vertical alignment. Design aids for required minimum lateral clearance (from the minor and major roads) are presented for different radii of intersections located on horizontal curves, guidelines are presented for offsetting opposing left-turn lanes to provide unobstructed required sight distance. Applications of the methodologies are illustrated using numerical examples.

Modelling of Three-Dimensional Intersection Sight Distance

Intersection sight distance(ISD) is an important design element. Each intersection has a potential for several different types of vehicular conflicts that can be greatly reduced through the provision of proper sight distance. Current guidelines do not adequately address sight distance requirements for intersections located on horizontal curves alone or horizontal curves combined with vertical alignments. In many practical situations, however, sight distance is required to be checked for an existing or proposed three-dimensional(3D) intersection alignments. In this thesis, models were developed to check sight (2001) were considered on 3D alignment: (1)Departure from stop-control minor-road and (2) Left-turns from major-road. For stop-control intersections, several cases were addressed. These include Case 1(a): Intersection and approaching vehicle (object) lie on the curve, Case 2: Intersection lies on the tangent and object lies on the curve. For both cases (1) and (2), obstruction may lie inside or outside the horizontal curve and the intersection and object can be anywhere with respect to the vertical alignment. In many practical situations, however, sight distance is required to be checked for an existing or proposed three-dimensional(3D) intersection alignments. In this thesis, models were developed to check sight (2001) were considered on 3D alignment: (1)Departure from stop-control minor-road and (2) Left-turns from major-road. For stop-control intersections, several cases were addressed. These include Case 1(a): Intersection and approaching vehicle (object) lie on the curve, Case 2: Intersection lies on the tangent and object lies on the curve. For both cases (1) and (2), obstruction may lie inside or outside the horizontal curve and the intersection and object can be anywhere with respect to the vertical alignment. Design aids for required minimum lateral clearance (from the minor and major roads) are presented for different radii of intersections located on horizontal curves, guidelines are presented for offsetting opposing left-turn lanes to provide unobstructed required sight distance. Applications of the methodologies are illustrated using numerical examples.

Plan recognition in dynamic 2-D environments

We look at the relatively unexplored problem of plan recognition applied to motion in 2-D environments where all moving objects are modelled as circles. Golog is a well-known high level logical language for solving planning problems and specifying agent controllers. Few studies have applied Golog to plan recognition. We use some of the features of this language, but its standard interpreter is adapted to solving plan recognition problems. This thesis makes several other contributions. First, plan recognition procedures are formulated as finite automata and expressed as Golog programs. Second, we elaborate a logical formalism for reasoning about depth and motion from an observer's viewpoint. We not only expand on this situation calculus based formalism, but also apply it to tackle plan recognition problems in the traffic domain. The proposed approach is implemented and thoroughly tested on recognizing simple behaviours such as left turns, right turns, and overtaking.

Plan recognition in dynamic 2-D environments

We look at the relatively unexplored problem of plan recognition applied to motion in 2-D environments where all moving objects are modelled as circles. Golog is a well-known high level logical language for solving planning problems and specifying agent controllers. Few studies have applied Golog to plan recognition. We use some of the features of this language, but its standard interpreter is adapted to solving plan recognition problems. This thesis makes several other contributions. First, plan recognition procedures are formulated as finite automata and expressed as Golog programs. Second, we elaborate a logical formalism for reasoning about depth and motion from an observer's viewpoint. We not only expand on this situation calculus based formalism, but also apply it to tackle plan recognition problems in the traffic domain. The proposed approach is implemented and thoroughly tested on recognizing simple behaviours such as left turns, right turns, and overtaking.

Risk to Bicyclists in a Separated Path from Left Turns across Multiple Lanes: A Case for Protected-Only Left Turns

At signalized intersections, permitted left turns (i.e., on a green ball, after yielding) across multiple through lanes and across a separated bike lane or bike path present a threat to bicyclist safety. A conflict study of two such intersections with a bidirectional bike path found that when cyclists cross while a vehicle is ready to turn left and there is no opposing through traffic to block it, the chance of the left-turning motorist yielding safely was only 9%, and the chance of their yielding at all—including yielding only after beginning the turn, then stopping in the opposing through lanes—was still only 37%. Motorist non-yielding rates were worse toward bikes arriving during green, toward bikes approaching from the opposite direction (i.e., riding on the right side of the road), and toward bikes facing a queue with multiple left turning vehicles. Of 112 cyclists who arrived on green when there was at least one left-turning car, but no opposing through traffic blocking it, 73 had to slow or stop to avoid a collision. Although these conflicts could be essentially eliminated using protected-only left turn phasing (turn on green arrow), common existing criteria prefer permitted left turns to reduce vehicular delay. A case study shows how, by considering multiple signalization alternatives, it can be possible to convert left turns to protected-only phasing without imposing a substantial delay burden on vehicles or other road users.