Proceedings of the Canadian Society for Civil Engineering Annual Conference 2023, Volume 2

Pavement surface distresses can be indicators of underlying structural issues on roads but require over-the-road or field surveys which can limit monitoring on extensive road networks to multiyear cycles. Advances in remote sensing imaging and image processing offer considerable potential to complement existing road-based surface distress measurement. If pavement surface distress observed from aerial or satellite imagery can be quantified, and the assessment is comparable to that obtained from field surveys, it may be possible to automate provincial or national-level network assessments from this imagery. This paper presents the results of a preliminary assessment of various types of aerial and satellite imagery for their potential to be used to identify road surface distresses. Qualitative assessments in terms of potential applicability for remote sensing (e.g. high, low, and not applicable) were conducted by exploring road surface distresses according to ASTM D6433-18 and included three different road functional classes in New Brunswick. Field survey data for pavement surface condition was obtained from the New Brunswick Department of Transportation and Infrastructure (NBDTI) for two National Highway System routes. The images were assessed according to ASTM-D6433-18 and compared to NBDTI condition data, and a pilot test was conducted on enhancing the highest-resolution imagery available. Select distresses were only observable at high severity in the highest-resolution imagery (SNB Orthoimagery at 0.10 m resolution); no distresses could be distinguished from any of the satellite imagery. Additional image processing was conducted on existing SNB Orthoimages at 0.10 m resolution; cracks that could be barely seen in the original aerial orthoimage can be detected by the Laplacian edge detector. While there have been advancements in this field since the time this work was originally completed in 2019, research continues to focus on segment-by-segment evaluation. There continues to be a need for network-wide applications of results. Artificial Intelligence (AI) can play an important role in the process.

Modeling pedestrian collisions has been difficult due to the rarity of such events and the high number of locations with zero collision frequency. Traditional collision data is reactive and may not identify risks until after injuries or fatalities have occurred. The supplement of the traditional collision data analysis approach with user perceptions of safety is necessary to provide a proactive approach to pedestrian safety. This study aims to take a step forward and collect data on pedestrian safety perceptions using virtual reality (VR) technology. Seventy participants were recruited and had to view pre-recorded 360-degree videos of intersections using a VR headset and report their perceptions of intersection design, traffic, and environmental factors from a safety point of view. Safety levels were analyzed on a Likert scale, and two-way analysis of variance (ANOVA) was performed to understand the variations in the average safety perceptions of intersections between participants during the day and night, as well as between men and women respondents. The study found that participants’ perceptions of safety varied depending on the time of day at the intersection, with the majority feeling less safe at night. The results suggest the importance of understanding people's perceptions of the walking environment and collision risk in designing cost-effective pedestrian safety initiatives that match both perceived and actual levels of safety on the road. Overall, the study demonstrates the potential benefits of using user-centered approaches and timely feedback on changes in pedestrian safety, which can provide a wealth of data to support proactive pedestrian safety planning.

Pavement deterioration is an inevitable part of the road network lifecycle that typically results from several natural and artificial factors. Non-destructive evaluation methods such as visual inspections, sensors, and others are used to evaluate road network conditions and are usually presented by the pavement condition index (PCI). However, the unique conditions of different roads and the scarcity of data introduce challenges in predicting the future PCI of a road. This paper introduces a new method for developing a pavement network's transition probability matrix (TPM) using Markovian chain to predict roads’ future conditions. Data from the Long-Term Pavement Performance database on 331 roads in the United States were used to develop the TPM. A maintenance, repair, and replacement (MRR) schedule was developed by optimizing the transition probabilities in the generated TPM. The current PCI, budget, and cost of different maintenance strategies are leveraged to create the MRR schedule. This work also introduces a new approach to generating the MRR schedule by combining targeted and random rehabilitation strategies with the model's TPM. The costs of targeted and random rehabilitation strategies are introduced to the model, allowing it to choose between both options depending on the outcome. By doing so, the probable outcomes of the schedule increase, allowing the model to produce an optimized MRR schedule that maximizes the overall average PCI of the network while sticking to budget constraints. The predicted PCIs from the developed TPM showed a reduced error of 45%, compared to previous works.

The emerging connected vehicle (CV) technology offers new ways for improving transit service performance in mixed-traffic corridors without having adverse impacts on the general traffic. One promising strategy, CV-enabled driver advisory system (DAS) for transit, aims at advising each bus driver of target travel speed and dwell time values to allow the bus to arrive at the downstream signalized intersection during the green phase, thus minimizing stoppage instances. In this paper, four different DAS algorithms with different objectives were tested with two space priority strategies in the form of exclusive bus lanes (EBLs) or dynamic bus lanes (DBLs). To evaluate the performance of the DAS algorithms, a traffic simulation model of the Eglinton East corridor in Toronto, Canada, was built in Aimsun Next for the morning peak period. The simulation results showed that DAS is a promising strategy that allows the bus to travel at or near the maximum allowable speed with 50% reduction in the total number of stops and 15 to 20% reduction in the bus energy consumption rate (for e-buses). Adding headway regularity as another objective of DAS shows significant benefits, with the regularity level of service (LOS) rising from LOS F to LOS C.

This work adapts a pedestrian-based modeling framework to represent cyclists’ behavior when crossing at a roundabout approach. The original pedestrian model structure captured pedestrian movement using a series of discrete choice models. This work expands on the previous model by applying a similar structure of discrete choice models to the behavior of cyclists as they cross the approach. Expectedly, the cyclists will behave in a different manner than pedestrians interacting with each other, thus requiring a new modeling framework. To account for these differences in behavior, additional modifications are introduced to the pedestrian-only model to capture cyclists’ choice. The proposed modeling structure will improve on conventional traffic microsimulation which uses a series of vehicle following models or simple rule-based approaches that cannot fully capture the dynamics of cyclist’s decision-making. The trajectories dataset used in this model are publicly available as the rounD dataset which is a real-world trajectory dataset extracted from drone videos. The trajectories were further analyzed to extract positional choices made by cyclists every half a second. A Multinomial Logit model is proposed to capture the choice process that describes cyclist's crossing. The model parameters are estimated in this study using maximum likelihood estimation. Comments on the statistical significance and goodness of fit are presented which were generally acceptable. The improvements in predicting cyclists’ behavior that this modeling structure promises will have significant potential benefits in improving the realism of multimodal traffic microsimulation, cycling infrastructure design and cyclists’ safety.

Connected vehicles have many promising applications which improve road safety. Thanks to vehicle-to-vehicle and vehicle-to-infrastructure communications, motorists can receive in-advance road information to better respond to hazardous situations and prevent severe collisions. As connected vehicles will likely share the road with human-driven vehicles in a transition period until full implementation, a continuous understanding of complex driver interactions with connected vehicles is needed. This study investigates the effect of introducing a connected cruise control-equipped vehicle supplemented with a crash warning system on the driver’s speed choice using a driving simulator. Participants were recruited to run the simulations in connected and non-connected environments. Advisory and warning messages via various delivery types were shared with the participants at different vehicle-to-vehicle connectivity lengths. The speed change ratio between the connected and non-connected environments was modeled using ordinal logistic regression. It was found that connectivity length, gender, education level, and years of driving experience have statistically significant effects on the speed change ratio at the 90% confidence level. Although the results are only valid to the study participant sample due to limitations in the collected sample, they draw insights into the human response to connected vehicles.

The province of Ontario has invested several billion dollars on improving the transportation infrastructure in the Greater Toronto and Hamilton Area (GTHA). The existing infrastructure is mainly focused on managing daily commuter trips. However, discretionary travel generates many trips in a modern urban transportation system. To promote sustainable modes of travel for discretionary purposes, understanding the factors that affect discretionary trip-making behavior is important. This paper investigates the mode choice behavior of home-based discretionary (HBD) trips in the GTHA, considering the land-use attributes and sociodemographics present among the post-secondary students of the study area. Multinomial logit models, which predict the mode choice of the individuals, are developed using two cross-sectional datasets from 2015 and 2019. The results reveal significant differences in mode choice behavior based on residence location and age. A marginal effects analysis also reveals that the propensity of generating transit and automobile trips decreases if a respondent lives downtown, whereas the propensity for active modes of travel increases. Moreover, discretionary travel is found to be less sensitive to travel time and cost, unlike commuter trips. The study’s findings indicate that the mode choice determinants vary significantly from commuter trips.

Due to the daily increasement in traffic, the total delay of roads increases which will decrease the roads level of service. Accordingly, there is a need to optimize the traffic signals to adapt to the variation in traffic at different times throughout the day. This paper focuses on solving the problem of urban congestion by optimizing the effective green time ratio of a given intersection under oversaturated conditions. The model employs genetic algorithm as it proved to be beneficial in solving such complex problems in multiple ways. The model was, then, validated using a real case study that was in Mohammed Bin Zayed City, Abu Dhabi, and UAE. The real case study was initially analyzed using “Synchro Traffic,” and the green timing, delays, and level of service were calculated. The results were, then, compared against the optimized green time, delay, and level of service obtained from the proposed genetic algorithm (GA) model, and recommendations and conclusion are highlighted.

In this research, the impact of traffic analysis zone (TAZ) characteristics on the likelihood of pedestrian–vehicle collisions in the City of Ottawa is investigated. Data was gathered from 422 TAZs. Generalized linear modeling (GLM) is used to analyze the data. This model utilized a negative binomial (NB) data distribution. Two methods, the half–half distribution method and the geo-process method, were used to allocate the data from the boundary of adjacent TAZs. According to this study, there is a statistically significant relationship between pedestrian collisions and the overall number of signalized intersections within a TAZ. Additionally, pedestrian collisions were less frequent on roads with speed limits of less than 40 km/h. This indicates that lower speed limits may be more favorable to pedestrian safety. The study also found that for roadways with a speed limit of 70 km/h, pedestrian-involved collisions were inversely related to roadway length. The model also revealed that pedestrian–vehicle collisions can be impacted by the presence of schools. Pedestrian-involved collisions are also more likely to happen on collector roadways than on major collector roadways. In addition, it was found that sociodemographic factors, including employment and the number of households, are linked to the frequency of pedestrian-involved collisions. This research opens the gate for proactively identifying and addressing potential road safety issues, which can assist planners and safety practitioners in reducing the frequency of pedestrian collisions.

Large earthquake events may damage civil infrastructure systems, cascading to impact community emergency resource access. For example, after the (M) 7.8 and (M) 7.5 Kahramanmaraş Earthquakes in Turkey (2023), many households with water service outages were forced to travel across damaged transportation networks to access emergency food and water supplies. Transportation system performance is often measured in minimized economic cost across a network. However, this metric may not include the social implications of infrastructure system disruptions. For example, vulnerable households often suffer disproportionate hardships from infrastructure interruptions and disparities in accessing emergency resources post-disaster. Here, we develop a household-level metric of single-mode-destination accessibility loss to estimate the median percentage (%) increase in travel time for emergency food and water resource access in households impacted by water service outages post-disaster. We present an application of the metric in a case study of the HayWired earthquake scenario, a hypothetical M_w 7.05 earthquake on the Hayward Fault in the San Francisco Bay Area (USA). We estimate a median percentage (%) increase in travel time (over a baseline) for emergency household food and water access using an origin–destination matrix analysis. Our results identified areas with disproportionate increases in travel time for emergency household food and water resource access in several communities. The metric could be applied to help civil engineers rapidly prioritize emergency road and water distribution repairs that may disproportionately impact vulnerable communities as they access life-sustaining emergency food and water resources.

The lateral adjacent distance (LAD) between a bicycle and a passing vehicle is an important parameter that reflects the level of safety and comfort experienced by the cyclist. This LAD can be influenced by a variety of geometric design decisions. This study presents an automated technique to measure a LAD using deep learning (DL). Specifically, this paper studies the use of deep neural networks (DNN) to automatically measure bicycle–vehicle lateral distance as they pass each other on urban streets in the City of Ottawa. The sensor used is a camera, set-up at a relatively low altitude. Instance Segmentation (IS) is a technique that aims to accurately detect the outlines of objects of interest with varying appearances. IS has been successfully used in applications beyond road safety analysis, e.g., autonomous vehicles and medical scan diagnosis. In this study, 14 classes of road users are defined based on their functional attributes rather than their unique appearances. After the annotation of training examples for each class, the data is used to train a DNN till an acceptable learning rate is achieved. Subsequently, a lateral distance is measured between polygons of interest of the detected IS output object of interest and showed promising results. The developed model is an IS multi-class multi-object detector and classifier. The trained DNN is validated based on manually measured distances from multi-day video observation. Low-altitude cameras provide a challenge due to a higher level of occlusion compared to higher-altitude cameras. However, IS demonstrated the potential to overcome this challenge. A limitation of this study is that it does not automatically detect adjacency but, instead, measures the distance between a vehicle and a bicycle when the adjacency is observed. This model can help researchers in analyzing objective as well as subjective cyclist safety in urban areas.

This paper evaluates the effectiveness of several traffic control devices to improve safety and capacity conditions of a rural all-way stop controlled intersection between Victoria Avenue and Canborough Road in the Town of Pelham, Ontario. Several measures of effectiveness are used to rank the proposed solutions, including safety, cost, level of service, and environmental impacts. The proposed alternatives are do-nothing, two-way stop controlled (TWSC), signalized intersection, and roundabout. The paper takes a phased approach and evaluates the proposed alternatives for the design years 2031 (Phase I) and 2051 (Phase II). For Phase I, the alternatives are do-nothing and TWSC, while signalized intersection and roundabout were evaluated for Phase II. The best alternative for each phase was then determined using a scoring method that included weighted criteria; each alternative was graded on a scale from 1 to 5. The results show that for Phase I, AWSC scored 3.15, while TWSC scored 1.50. For Phase II, the signalized intersection scored 2.63, while the roundabout scored 3.63. Taking a phased approach to the improvements ensures the optimization of the cost, capacity, and safety variables affecting the intersection.

Travel time is a performance index for measuring users’ perception of the quality of service of a roadway. This paper investigated the extent of bus rapid transit (BRT) travel time differentials caused by mixed traffic flow ‘with BRT’ in Cape Town, South Africa. Travel time differential in the context of this study is simply the difference between the actual time required by both the drivers of BRT buses and other vehicles to traverse a roadway section under mixed traffic flow ‘with BRT’ and the corresponding travel time under mixed traffic flow ‘without BRT’. Based on the hypothesis that a mixed traffic flow involving BRT buses and other vehicles would result in differentials in traffic stream characteristics, this study therefore estimated the extent of BRT travel time differentials between three scenarios namely: BRT dedicated lane scenario, ‘without BRT’ scenario, and ‘with BRT’ scenario. Consequently, ‘with and without’ BRT impact study was carried out at four selected sites on regional route R27, in Cape Town, South Africa. Traffic characteristics of vehicles such as flow, speed, and density, moving on both the BRT dedicated lanes and their adjoining lanes, were logged using two automatic traffic counters (ATC) per road section for a period of twelve weeks, while flow-density models were calibrated to obtain the free flow speeds. The survey data was supported by design information harvested from the Western Cape Department of Transport and Public Works (WCDTPW), in Cape Town. Subsequently the US Bureau of Public Roads (US-BPR) travel time model equation was used to predict travel time over the roadway segments under the three scenarios and the outcomes compared. Results showed a minor speed reduction of 13.5% under the mixed traffic flow ‘with BRT’, which triggered a corresponding 11% minor increase in travel time. The paper concluded that mixed traffic flow ‘with BRT’ affects travel time, caused by speed changes.

Soil arching phenomenon plays an important role in distributing the stresses through earth embankments in the presence of stiff deep foundation system. The mass of soil at the lower stiffness zone tends to yield and deform, while the stiffer zone will oppose this deformation causing additional soil pressure. The type of soil arching can be distinguished depending on the compressibility of the structure and the surrounding soil. The current study investigates the performance of two embankments rested over the ground surface underlain two combined pile caps spaced by 32.0 m center-to-center that are used to support the vertical loadings of two piers of a roadway bridge. The soil arching effect is evaluated at short-term and long-term conditions due to the presence of a layer of soft compressible soil in the native ground. Three-dimensional numerical analysis was conducted on 5.5 m high embankments rested over zones with different vertical stiffnesses. The purpose of this study is to evaluate the performance of the embankments through long-term deformations in addition to the additional induced axial loads on the deep foundation system resulting from soil arching phenomenon in the embankments. The results demonstrates that the vertical arching factor over the combined pile caps increased from 1.69 just after construction to 1.82 at the end of 90% consolidation of the soft layer. In addition, the serviceability limit state of the earth embankment was controlled by the longitudinal settlement trough of the embankments to meet the current design codes limitations.

Worldwide, railroads are essential for transporting passengers and cargo. Any failure during this infrastructure can have a significant social, environmental, and economic impacts. Therefore, a better understanding of this failure and prediction of railroad deterioration can lead to more efficient maintenance and rehabilitation, as well as safer operations. Recent advances in railway data collection techniques provide an opportunity to develop more accurate models for predicting track deterioration using machine-learning techniques. This research uses machine-learning models to forecast track deterioration to handle large and automatically collected railroad condition datasets. The study was developed in four steps: The open dataset of track characteristics and defects made available by the “INFORMS 2015 Railway Applications Section Problem Solving Competition” was used to validate the models. Defect length, amplitude, type, and tag were set as targets in separate models. Random forest and XGBoost were compared for each. The machine-learning models developed were able to determine current tag with an accuracy of 99% and predict future tags with 78% accuracy. As defects are predicted and maintained before they exceed certain thresholds, these models can improve track performance, reduce costly downtime and ensure continued safe operations.