VK
V.L. Knoop
info
Please Note
<p>This page displays the records of the person named above and is not linked to a unique person identifier. This record may need to be merged to a profile.</p>
19 records found
1
Highway congestion costs European road networks an estimated €100 billion annually. Highway traffic is characterised by congestion shock waves, arising when density exceeds a critical threshold density. These shock waves cause a heterogeneous density distribution across the network, reducing its overall efficiency. Crucially, this heterogeneity is path-dependent: the network operates less efficiently during recovery from congestion than during its build-up, a phenomenon known as hysteresis. Current network-level traffic models cannot reproduce this, as they rely on a memoryless speed–density relation.
This thesis extends the classical bathtub model — a network-level traffic model driven by average density — with an explicit, path-dependent congestion variable representing the spatial extent of congestion in the network. To calibrate the model, loop detector data from 124 weekday mornings on the A10 ring road in Amsterdam (January–June 2018) were used to construct empirical congestion measures. Three candidate congestion measures were compared — density spread, an unweighted congestion fraction, and a density–weighted variant — on their ability to characterise the network congestion state. The density–weighted measure was selected based on superior fit to observed network speeds (R2 = 0.960). Empirical analysis established a critical density threshold ρcrit ≈ 17.2 veh km−1 above which congestion onset and recovery follow asymmetric paths, captured by a dynamical system with separate build-up and recovery rates. Calibration confirmed that congestion builds approximately 31% faster than it dissipates. Forward simulation of the calibrated model reproduces the hysteresis loops observed on the A10.
The model offers a tractable basis for traffic management applications such as ramp metering, variable speed limits, and congestion pricing, enabling prediction of the asymmetric dynamics of morning rush congestion. A current limitation is that inflow is treated as exogenous, producing an unrealistically sharp gridlock sensitivity. Incorporating observed inflow data is identified as the most important direction for future work. ...
This thesis extends the classical bathtub model — a network-level traffic model driven by average density — with an explicit, path-dependent congestion variable representing the spatial extent of congestion in the network. To calibrate the model, loop detector data from 124 weekday mornings on the A10 ring road in Amsterdam (January–June 2018) were used to construct empirical congestion measures. Three candidate congestion measures were compared — density spread, an unweighted congestion fraction, and a density–weighted variant — on their ability to characterise the network congestion state. The density–weighted measure was selected based on superior fit to observed network speeds (R2 = 0.960). Empirical analysis established a critical density threshold ρcrit ≈ 17.2 veh km−1 above which congestion onset and recovery follow asymmetric paths, captured by a dynamical system with separate build-up and recovery rates. Calibration confirmed that congestion builds approximately 31% faster than it dissipates. Forward simulation of the calibrated model reproduces the hysteresis loops observed on the A10.
The model offers a tractable basis for traffic management applications such as ramp metering, variable speed limits, and congestion pricing, enabling prediction of the asymmetric dynamics of morning rush congestion. A current limitation is that inflow is treated as exogenous, producing an unrealistically sharp gridlock sensitivity. Incorporating observed inflow data is identified as the most important direction for future work. ...
Highway congestion costs European road networks an estimated €100 billion annually. Highway traffic is characterised by congestion shock waves, arising when density exceeds a critical threshold density. These shock waves cause a heterogeneous density distribution across the network, reducing its overall efficiency. Crucially, this heterogeneity is path-dependent: the network operates less efficiently during recovery from congestion than during its build-up, a phenomenon known as hysteresis. Current network-level traffic models cannot reproduce this, as they rely on a memoryless speed–density relation.
This thesis extends the classical bathtub model — a network-level traffic model driven by average density — with an explicit, path-dependent congestion variable representing the spatial extent of congestion in the network. To calibrate the model, loop detector data from 124 weekday mornings on the A10 ring road in Amsterdam (January–June 2018) were used to construct empirical congestion measures. Three candidate congestion measures were compared — density spread, an unweighted congestion fraction, and a density–weighted variant — on their ability to characterise the network congestion state. The density–weighted measure was selected based on superior fit to observed network speeds (R2 = 0.960). Empirical analysis established a critical density threshold ρcrit ≈ 17.2 veh km−1 above which congestion onset and recovery follow asymmetric paths, captured by a dynamical system with separate build-up and recovery rates. Calibration confirmed that congestion builds approximately 31% faster than it dissipates. Forward simulation of the calibrated model reproduces the hysteresis loops observed on the A10.
The model offers a tractable basis for traffic management applications such as ramp metering, variable speed limits, and congestion pricing, enabling prediction of the asymmetric dynamics of morning rush congestion. A current limitation is that inflow is treated as exogenous, producing an unrealistically sharp gridlock sensitivity. Incorporating observed inflow data is identified as the most important direction for future work.
This thesis extends the classical bathtub model — a network-level traffic model driven by average density — with an explicit, path-dependent congestion variable representing the spatial extent of congestion in the network. To calibrate the model, loop detector data from 124 weekday mornings on the A10 ring road in Amsterdam (January–June 2018) were used to construct empirical congestion measures. Three candidate congestion measures were compared — density spread, an unweighted congestion fraction, and a density–weighted variant — on their ability to characterise the network congestion state. The density–weighted measure was selected based on superior fit to observed network speeds (R2 = 0.960). Empirical analysis established a critical density threshold ρcrit ≈ 17.2 veh km−1 above which congestion onset and recovery follow asymmetric paths, captured by a dynamical system with separate build-up and recovery rates. Calibration confirmed that congestion builds approximately 31% faster than it dissipates. Forward simulation of the calibrated model reproduces the hysteresis loops observed on the A10.
The model offers a tractable basis for traffic management applications such as ramp metering, variable speed limits, and congestion pricing, enabling prediction of the asymmetric dynamics of morning rush congestion. A current limitation is that inflow is treated as exogenous, producing an unrealistically sharp gridlock sensitivity. Incorporating observed inflow data is identified as the most important direction for future work.
Modelling changing travel behaviour in response to road capacity reductions
The impact of reallocating road space in urban areas on car use
As cities increasingly prioritize sustainable and low-car urban environments, road capacity reductions have become a widely applied policy measure. While traditional transport models predict severe congestion resulting from such interventions, empirical studies suggest that traffic volumes tend to decrease, a phenomenon known as ‘disappearing traffic.’ This study develops a model-based approach to estimate the behavioural adaptations underlying disappearing traffic and their implications for travel demand and network performance.
Using a three-step iterative framework, the model adjusts the traffic demand based on observed travel times in response to capacity reductions. The process accounts for modal shifts, destination changes and trip frequency reduction. The methodology is applied to a real-world case study where significant road space reallocation is implemented, allowing for an evaluation of the extent and mechanisms of disappearing traffic.
Findings indicate that a considerable share of the traffic does not reroute but instead disappears due to behavioural changes. The extent of disappearing traffic depends on the severity of capacity reductions, the existing congestion levels on both intervening and alternative routes, and the availability of viable transport alternatives. The findings highlight the importance of incorporating adaptive travel behaviour in transport models to better support decision-making in sustainable urban mobility planning.
This research contributes to the growing body of knowledge on the effects of road capacity reductions and provides a modelling framework for policymakers and urban planners to anticipate and manage disappearing traffic effects effectively.
...
Using a three-step iterative framework, the model adjusts the traffic demand based on observed travel times in response to capacity reductions. The process accounts for modal shifts, destination changes and trip frequency reduction. The methodology is applied to a real-world case study where significant road space reallocation is implemented, allowing for an evaluation of the extent and mechanisms of disappearing traffic.
Findings indicate that a considerable share of the traffic does not reroute but instead disappears due to behavioural changes. The extent of disappearing traffic depends on the severity of capacity reductions, the existing congestion levels on both intervening and alternative routes, and the availability of viable transport alternatives. The findings highlight the importance of incorporating adaptive travel behaviour in transport models to better support decision-making in sustainable urban mobility planning.
This research contributes to the growing body of knowledge on the effects of road capacity reductions and provides a modelling framework for policymakers and urban planners to anticipate and manage disappearing traffic effects effectively.
...
As cities increasingly prioritize sustainable and low-car urban environments, road capacity reductions have become a widely applied policy measure. While traditional transport models predict severe congestion resulting from such interventions, empirical studies suggest that traffic volumes tend to decrease, a phenomenon known as ‘disappearing traffic.’ This study develops a model-based approach to estimate the behavioural adaptations underlying disappearing traffic and their implications for travel demand and network performance.
Using a three-step iterative framework, the model adjusts the traffic demand based on observed travel times in response to capacity reductions. The process accounts for modal shifts, destination changes and trip frequency reduction. The methodology is applied to a real-world case study where significant road space reallocation is implemented, allowing for an evaluation of the extent and mechanisms of disappearing traffic.
Findings indicate that a considerable share of the traffic does not reroute but instead disappears due to behavioural changes. The extent of disappearing traffic depends on the severity of capacity reductions, the existing congestion levels on both intervening and alternative routes, and the availability of viable transport alternatives. The findings highlight the importance of incorporating adaptive travel behaviour in transport models to better support decision-making in sustainable urban mobility planning.
This research contributes to the growing body of knowledge on the effects of road capacity reductions and provides a modelling framework for policymakers and urban planners to anticipate and manage disappearing traffic effects effectively.
Using a three-step iterative framework, the model adjusts the traffic demand based on observed travel times in response to capacity reductions. The process accounts for modal shifts, destination changes and trip frequency reduction. The methodology is applied to a real-world case study where significant road space reallocation is implemented, allowing for an evaluation of the extent and mechanisms of disappearing traffic.
Findings indicate that a considerable share of the traffic does not reroute but instead disappears due to behavioural changes. The extent of disappearing traffic depends on the severity of capacity reductions, the existing congestion levels on both intervening and alternative routes, and the availability of viable transport alternatives. The findings highlight the importance of incorporating adaptive travel behaviour in transport models to better support decision-making in sustainable urban mobility planning.
This research contributes to the growing body of knowledge on the effects of road capacity reductions and provides a modelling framework for policymakers and urban planners to anticipate and manage disappearing traffic effects effectively.
Keep Traffic Moving: Safeguarding Traffic Accessibility during Temporary Roadworks
A Microsimulation-based Case Study Complemented with Expert Validation
In this research a dynamic time and energy model is constructed to simulate free flow utility cycling. Using a Monte Carlo simulation and distributions of rider and bicycle characteristics a population of cyclists is modelled to find the difference in time and energy expenditure on six different bicycle routes. The riders use realistic like power inputs and braking behaviour based on the rider and the route characteristics. While travel times are almost always lower for all cyclists when comparing routes, there are cyclists for who a shorter route distance does not equal a lower energy expenditure. For simpler routes (less elevation difference, traffic lights and shorter distance) the standard deviation of both time and energy decreases, showing that slow cyclist have a relative higher gain. For e-bike users there is even almost no difference in energy expenditure between the six evaluated routes that have varying elevation, traffic signals and routing, while travel times show a similar trend as for regular bicycles. With this model difference in routes can be quantified in matters of time and energy expenditure for a population of cyclists giving an objective picture of the differences between routes, which can be a useful tool for city planning and evaluation of bicycle infrastructure.
...
In this research a dynamic time and energy model is constructed to simulate free flow utility cycling. Using a Monte Carlo simulation and distributions of rider and bicycle characteristics a population of cyclists is modelled to find the difference in time and energy expenditure on six different bicycle routes. The riders use realistic like power inputs and braking behaviour based on the rider and the route characteristics. While travel times are almost always lower for all cyclists when comparing routes, there are cyclists for who a shorter route distance does not equal a lower energy expenditure. For simpler routes (less elevation difference, traffic lights and shorter distance) the standard deviation of both time and energy decreases, showing that slow cyclist have a relative higher gain. For e-bike users there is even almost no difference in energy expenditure between the six evaluated routes that have varying elevation, traffic signals and routing, while travel times show a similar trend as for regular bicycles. With this model difference in routes can be quantified in matters of time and energy expenditure for a population of cyclists giving an objective picture of the differences between routes, which can be a useful tool for city planning and evaluation of bicycle infrastructure.
Autonomous vehicles rely on prediction modules, in order to plan collision-free trajectories. Vehicle trajectory prediction models are multimodal, to account for the multiple route options and the inherent uncertainty in human behavior. The state-of-the-art prediction models are deep-learning models, which are susceptible to mode collapse, a phenomenon in which the model fails to output the full distribution of modes and only predicts the most likely one. Mode collapsing poses safety concerns for autonomous driving, as missed predictions could result in collisions. Most works have focused on addressing this issue by generating diverse predictions that cover various route options at the environmental level. However, there are no metrics for mode-collapse. Furthermore, little attention has been given to generating diversity in the interaction modes among agent trajectories. Additionally, the traditional distance-based metrics are heavily dependent on datasets and do not evaluate interactions between agents. To this end, we propose a novel evaluation framework that assesses the interaction modes of joint trajectory predictions, focusing only on the safety-critical interactions in a dataset. We introduce a metric for mode-collapse and time-based metrics for mode correctness and coverage, shedding light on the temporal dimension of the predictions. We test four multi-agent trajectory prediction models on the widely used nuScenes dataset and conclude that mode collapse happens. While the rate of correctly predicted interaction modes increases closer to the interaction event, there are still cases where the models are unable to predict the interaction mode even right before the interaction happens. With the introduction of our novel framework, researchers can now benchmark their models’ performance in predicting critical interactions. This provides new insights and perspectives, helping the holistic evaluation and interpretation of a model’s performance. Additionally, our work offers a new developmental direction for prediction models, aiming for greater consistency and accuracy in predicting agent interactions, thereby advancing the safety of autonomous driving systems. Our evaluation framework is available online at: https://github.com/MaartenHugenholtz/InteractionEval
...
Autonomous vehicles rely on prediction modules, in order to plan collision-free trajectories. Vehicle trajectory prediction models are multimodal, to account for the multiple route options and the inherent uncertainty in human behavior. The state-of-the-art prediction models are deep-learning models, which are susceptible to mode collapse, a phenomenon in which the model fails to output the full distribution of modes and only predicts the most likely one. Mode collapsing poses safety concerns for autonomous driving, as missed predictions could result in collisions. Most works have focused on addressing this issue by generating diverse predictions that cover various route options at the environmental level. However, there are no metrics for mode-collapse. Furthermore, little attention has been given to generating diversity in the interaction modes among agent trajectories. Additionally, the traditional distance-based metrics are heavily dependent on datasets and do not evaluate interactions between agents. To this end, we propose a novel evaluation framework that assesses the interaction modes of joint trajectory predictions, focusing only on the safety-critical interactions in a dataset. We introduce a metric for mode-collapse and time-based metrics for mode correctness and coverage, shedding light on the temporal dimension of the predictions. We test four multi-agent trajectory prediction models on the widely used nuScenes dataset and conclude that mode collapse happens. While the rate of correctly predicted interaction modes increases closer to the interaction event, there are still cases where the models are unable to predict the interaction mode even right before the interaction happens. With the introduction of our novel framework, researchers can now benchmark their models’ performance in predicting critical interactions. This provides new insights and perspectives, helping the holistic evaluation and interpretation of a model’s performance. Additionally, our work offers a new developmental direction for prediction models, aiming for greater consistency and accuracy in predicting agent interactions, thereby advancing the safety of autonomous driving systems. Our evaluation framework is available online at: https://github.com/MaartenHugenholtz/InteractionEval
The increasing prevalence of micromobility vehicles in urban environments has raised concerns about safety in shared cycling spaces. This study examines the overtaking behavior of e-scooter and e-bike riders to inform traffic management strategies and infrastructure development. A controlled experiment was conducted using strategically placed cameras to track vehicle trajectories and inertial measurement units (IMUs) to capture roll data. Extensive data processing ensured accuracy and synchronization of trajectory and IMU information. Key findings reveal that e-bikes overtaking e-scooters initiate maneuvers from greater distances but maintain smaller lateral distances compared to e-bikes overtaking e-bikes.
Lateral position differences showed a stronger correlation with speed difference than longitudinal position differences. The highest roll rates and angles occurred during the overtaking phase. Pre-overtaking, higher roll rates and angles were observed when e-bikes overtook other e-bikes, indicating greater control adjustments. No significant gender differences were found in overtaking behavior. However, in non-interactive scenarios, male e-scooter riders traveled at higher speeds than females, while no gender differences were observed among e-bike riders. These results provide insights into the complex interactions between different types of micromobility vehicles during overtaking maneuvers. The findings underscore the need for targeted safety interventions and infrastructure improvements to mitigate risks associated with shared cycling spaces, ensuring safer coexistence of micromobility users and conventional cyclists in urban environments. ...
Lateral position differences showed a stronger correlation with speed difference than longitudinal position differences. The highest roll rates and angles occurred during the overtaking phase. Pre-overtaking, higher roll rates and angles were observed when e-bikes overtook other e-bikes, indicating greater control adjustments. No significant gender differences were found in overtaking behavior. However, in non-interactive scenarios, male e-scooter riders traveled at higher speeds than females, while no gender differences were observed among e-bike riders. These results provide insights into the complex interactions between different types of micromobility vehicles during overtaking maneuvers. The findings underscore the need for targeted safety interventions and infrastructure improvements to mitigate risks associated with shared cycling spaces, ensuring safer coexistence of micromobility users and conventional cyclists in urban environments. ...
The increasing prevalence of micromobility vehicles in urban environments has raised concerns about safety in shared cycling spaces. This study examines the overtaking behavior of e-scooter and e-bike riders to inform traffic management strategies and infrastructure development. A controlled experiment was conducted using strategically placed cameras to track vehicle trajectories and inertial measurement units (IMUs) to capture roll data. Extensive data processing ensured accuracy and synchronization of trajectory and IMU information. Key findings reveal that e-bikes overtaking e-scooters initiate maneuvers from greater distances but maintain smaller lateral distances compared to e-bikes overtaking e-bikes.
Lateral position differences showed a stronger correlation with speed difference than longitudinal position differences. The highest roll rates and angles occurred during the overtaking phase. Pre-overtaking, higher roll rates and angles were observed when e-bikes overtook other e-bikes, indicating greater control adjustments. No significant gender differences were found in overtaking behavior. However, in non-interactive scenarios, male e-scooter riders traveled at higher speeds than females, while no gender differences were observed among e-bike riders. These results provide insights into the complex interactions between different types of micromobility vehicles during overtaking maneuvers. The findings underscore the need for targeted safety interventions and infrastructure improvements to mitigate risks associated with shared cycling spaces, ensuring safer coexistence of micromobility users and conventional cyclists in urban environments.
Lateral position differences showed a stronger correlation with speed difference than longitudinal position differences. The highest roll rates and angles occurred during the overtaking phase. Pre-overtaking, higher roll rates and angles were observed when e-bikes overtook other e-bikes, indicating greater control adjustments. No significant gender differences were found in overtaking behavior. However, in non-interactive scenarios, male e-scooter riders traveled at higher speeds than females, while no gender differences were observed among e-bike riders. These results provide insights into the complex interactions between different types of micromobility vehicles during overtaking maneuvers. The findings underscore the need for targeted safety interventions and infrastructure improvements to mitigate risks associated with shared cycling spaces, ensuring safer coexistence of micromobility users and conventional cyclists in urban environments.
This project aims to relieve traffic pressure and enhance the parking experience for attendees during planned special events (PSEs). The objective is to develop an optimal strategy for efficiently allocating parking spaces during PSEs in parking lots.
PSEs, such as football games or large concerts, typically result in concentrated vehicle arrivals within a limited time period, leading to increased traffic flow, potential disruptions, elevated emissions, and safety concerns in nearby areas. By optimizing parking space allocation strategies in the parking lot, this project seeks to improve overall traffic management and relieve these challenges.
To achieve this, a linear programming (LP) algorithm and a simulation-based genetic algorithm (GA) are employed to search for the optimal solution. While the LP model offers computational efficiency, it has limitations in incorporating different route conditions. To address this, an agent-based simulation is constructed to depict the interaction and movement of vehicles within the parking lot. The simulation-based GA utilizes objective values derived from the simulation, providing a more comprehensive basis for finding the optimal solution. The allocation process considers factors such as parking lot layout, vehicle entry time step, and specific parking rules including road directions within the parking lot.
Results demonstrate that the optimal strategy obtained from the simulation-based GA outperforms comparison groups. The simulation-based GA showcases its ability to converge on the optimal solution within a large solution area. The optimal strategy saving time for all vehicles, particularly during periods of high demand. Effective parking is achieved by allocating parking spaces according to the arrival order and positioning vehicles on the left or right based on their arrival order and parking space location.
By employing these methods, this project offers a valuable contribution to the field of parking space allocation in the parking lot during PSEs, enhancing the overall parking experience for event attendees. ...
PSEs, such as football games or large concerts, typically result in concentrated vehicle arrivals within a limited time period, leading to increased traffic flow, potential disruptions, elevated emissions, and safety concerns in nearby areas. By optimizing parking space allocation strategies in the parking lot, this project seeks to improve overall traffic management and relieve these challenges.
To achieve this, a linear programming (LP) algorithm and a simulation-based genetic algorithm (GA) are employed to search for the optimal solution. While the LP model offers computational efficiency, it has limitations in incorporating different route conditions. To address this, an agent-based simulation is constructed to depict the interaction and movement of vehicles within the parking lot. The simulation-based GA utilizes objective values derived from the simulation, providing a more comprehensive basis for finding the optimal solution. The allocation process considers factors such as parking lot layout, vehicle entry time step, and specific parking rules including road directions within the parking lot.
Results demonstrate that the optimal strategy obtained from the simulation-based GA outperforms comparison groups. The simulation-based GA showcases its ability to converge on the optimal solution within a large solution area. The optimal strategy saving time for all vehicles, particularly during periods of high demand. Effective parking is achieved by allocating parking spaces according to the arrival order and positioning vehicles on the left or right based on their arrival order and parking space location.
By employing these methods, this project offers a valuable contribution to the field of parking space allocation in the parking lot during PSEs, enhancing the overall parking experience for event attendees. ...
This project aims to relieve traffic pressure and enhance the parking experience for attendees during planned special events (PSEs). The objective is to develop an optimal strategy for efficiently allocating parking spaces during PSEs in parking lots.
PSEs, such as football games or large concerts, typically result in concentrated vehicle arrivals within a limited time period, leading to increased traffic flow, potential disruptions, elevated emissions, and safety concerns in nearby areas. By optimizing parking space allocation strategies in the parking lot, this project seeks to improve overall traffic management and relieve these challenges.
To achieve this, a linear programming (LP) algorithm and a simulation-based genetic algorithm (GA) are employed to search for the optimal solution. While the LP model offers computational efficiency, it has limitations in incorporating different route conditions. To address this, an agent-based simulation is constructed to depict the interaction and movement of vehicles within the parking lot. The simulation-based GA utilizes objective values derived from the simulation, providing a more comprehensive basis for finding the optimal solution. The allocation process considers factors such as parking lot layout, vehicle entry time step, and specific parking rules including road directions within the parking lot.
Results demonstrate that the optimal strategy obtained from the simulation-based GA outperforms comparison groups. The simulation-based GA showcases its ability to converge on the optimal solution within a large solution area. The optimal strategy saving time for all vehicles, particularly during periods of high demand. Effective parking is achieved by allocating parking spaces according to the arrival order and positioning vehicles on the left or right based on their arrival order and parking space location.
By employing these methods, this project offers a valuable contribution to the field of parking space allocation in the parking lot during PSEs, enhancing the overall parking experience for event attendees.
PSEs, such as football games or large concerts, typically result in concentrated vehicle arrivals within a limited time period, leading to increased traffic flow, potential disruptions, elevated emissions, and safety concerns in nearby areas. By optimizing parking space allocation strategies in the parking lot, this project seeks to improve overall traffic management and relieve these challenges.
To achieve this, a linear programming (LP) algorithm and a simulation-based genetic algorithm (GA) are employed to search for the optimal solution. While the LP model offers computational efficiency, it has limitations in incorporating different route conditions. To address this, an agent-based simulation is constructed to depict the interaction and movement of vehicles within the parking lot. The simulation-based GA utilizes objective values derived from the simulation, providing a more comprehensive basis for finding the optimal solution. The allocation process considers factors such as parking lot layout, vehicle entry time step, and specific parking rules including road directions within the parking lot.
Results demonstrate that the optimal strategy obtained from the simulation-based GA outperforms comparison groups. The simulation-based GA showcases its ability to converge on the optimal solution within a large solution area. The optimal strategy saving time for all vehicles, particularly during periods of high demand. Effective parking is achieved by allocating parking spaces according to the arrival order and positioning vehicles on the left or right based on their arrival order and parking space location.
By employing these methods, this project offers a valuable contribution to the field of parking space allocation in the parking lot during PSEs, enhancing the overall parking experience for event attendees.
The rapid adoption of e-bikes as an alternative mode of transportation to automobiles gives rise to new methods of safety regulations for cyclists. Modern e-bikes feature Internet of Things (IoT) modules capable of collecting and sending cycling data that can be used for traffic safety analysis. This study explores the potential of using cycling data to detect dangerous intersections and then implement local low-speed areas using geofencing.
A natural cycling experiment with 10 participants is conducted, where data is collected similar to the capabilities of e-bike IoT modules (GNSS, power and IMU data) and participants are asked to cycle through Delft, The Netherlands. The data is combined with a traffic accident dataset from the Dutch government, where intersections with accidents are labelled dangerous. The cycling data is separated into separate intersection approaches, based on positional data. Metrics based on power and cadence data show the most significant statistical difference (p-values <0.02) and the largest effect size (Cohen's d >0.80) at dangerous intersections. A binary classification model is trained on the dataset, which can correctly predict whether an intersection is dangerous or safe with an accuracy of 68,2%, a specificity of 50,0%, and a sensitivity of 24,2%. This prediction is based on the metrics of a single intersection approach.
To investigate the viability of using geofencing to implement low-speed areas at designated dangerous intersections, a simulation is carried out to determine the minimum size of a geofence that effectively slows down e-bike cyclists. The minimal effective geofence size to limit a cyclist's speed at an intersection is determined by simulating cyclists approaching a geofence for several geofence perimeters, incline levels, and wind speeds. The e-bike's motor stops supporting when the geofence is entered and the cyclist is cycling above a target speed. The minimal effective geofence size ranges from 50 to 300+ m and depends on the target speed, wind speed, and road incline.
This study shows that cycling data can be used to identify dangerous intersections. Enforcing a slower speed on intersections with local geofencing is not feasible, as the geofences have to be extremely large, showing a lot of overlap on other intersections in cities due to high intersection density. Future work can be done on the intersection risk estimation method and the feasibility of speed-limiting geofences that employ active braking. ...
A natural cycling experiment with 10 participants is conducted, where data is collected similar to the capabilities of e-bike IoT modules (GNSS, power and IMU data) and participants are asked to cycle through Delft, The Netherlands. The data is combined with a traffic accident dataset from the Dutch government, where intersections with accidents are labelled dangerous. The cycling data is separated into separate intersection approaches, based on positional data. Metrics based on power and cadence data show the most significant statistical difference (p-values <0.02) and the largest effect size (Cohen's d >0.80) at dangerous intersections. A binary classification model is trained on the dataset, which can correctly predict whether an intersection is dangerous or safe with an accuracy of 68,2%, a specificity of 50,0%, and a sensitivity of 24,2%. This prediction is based on the metrics of a single intersection approach.
To investigate the viability of using geofencing to implement low-speed areas at designated dangerous intersections, a simulation is carried out to determine the minimum size of a geofence that effectively slows down e-bike cyclists. The minimal effective geofence size to limit a cyclist's speed at an intersection is determined by simulating cyclists approaching a geofence for several geofence perimeters, incline levels, and wind speeds. The e-bike's motor stops supporting when the geofence is entered and the cyclist is cycling above a target speed. The minimal effective geofence size ranges from 50 to 300+ m and depends on the target speed, wind speed, and road incline.
This study shows that cycling data can be used to identify dangerous intersections. Enforcing a slower speed on intersections with local geofencing is not feasible, as the geofences have to be extremely large, showing a lot of overlap on other intersections in cities due to high intersection density. Future work can be done on the intersection risk estimation method and the feasibility of speed-limiting geofences that employ active braking. ...
The rapid adoption of e-bikes as an alternative mode of transportation to automobiles gives rise to new methods of safety regulations for cyclists. Modern e-bikes feature Internet of Things (IoT) modules capable of collecting and sending cycling data that can be used for traffic safety analysis. This study explores the potential of using cycling data to detect dangerous intersections and then implement local low-speed areas using geofencing.
A natural cycling experiment with 10 participants is conducted, where data is collected similar to the capabilities of e-bike IoT modules (GNSS, power and IMU data) and participants are asked to cycle through Delft, The Netherlands. The data is combined with a traffic accident dataset from the Dutch government, where intersections with accidents are labelled dangerous. The cycling data is separated into separate intersection approaches, based on positional data. Metrics based on power and cadence data show the most significant statistical difference (p-values <0.02) and the largest effect size (Cohen's d >0.80) at dangerous intersections. A binary classification model is trained on the dataset, which can correctly predict whether an intersection is dangerous or safe with an accuracy of 68,2%, a specificity of 50,0%, and a sensitivity of 24,2%. This prediction is based on the metrics of a single intersection approach.
To investigate the viability of using geofencing to implement low-speed areas at designated dangerous intersections, a simulation is carried out to determine the minimum size of a geofence that effectively slows down e-bike cyclists. The minimal effective geofence size to limit a cyclist's speed at an intersection is determined by simulating cyclists approaching a geofence for several geofence perimeters, incline levels, and wind speeds. The e-bike's motor stops supporting when the geofence is entered and the cyclist is cycling above a target speed. The minimal effective geofence size ranges from 50 to 300+ m and depends on the target speed, wind speed, and road incline.
This study shows that cycling data can be used to identify dangerous intersections. Enforcing a slower speed on intersections with local geofencing is not feasible, as the geofences have to be extremely large, showing a lot of overlap on other intersections in cities due to high intersection density. Future work can be done on the intersection risk estimation method and the feasibility of speed-limiting geofences that employ active braking.
A natural cycling experiment with 10 participants is conducted, where data is collected similar to the capabilities of e-bike IoT modules (GNSS, power and IMU data) and participants are asked to cycle through Delft, The Netherlands. The data is combined with a traffic accident dataset from the Dutch government, where intersections with accidents are labelled dangerous. The cycling data is separated into separate intersection approaches, based on positional data. Metrics based on power and cadence data show the most significant statistical difference (p-values <0.02) and the largest effect size (Cohen's d >0.80) at dangerous intersections. A binary classification model is trained on the dataset, which can correctly predict whether an intersection is dangerous or safe with an accuracy of 68,2%, a specificity of 50,0%, and a sensitivity of 24,2%. This prediction is based on the metrics of a single intersection approach.
To investigate the viability of using geofencing to implement low-speed areas at designated dangerous intersections, a simulation is carried out to determine the minimum size of a geofence that effectively slows down e-bike cyclists. The minimal effective geofence size to limit a cyclist's speed at an intersection is determined by simulating cyclists approaching a geofence for several geofence perimeters, incline levels, and wind speeds. The e-bike's motor stops supporting when the geofence is entered and the cyclist is cycling above a target speed. The minimal effective geofence size ranges from 50 to 300+ m and depends on the target speed, wind speed, and road incline.
This study shows that cycling data can be used to identify dangerous intersections. Enforcing a slower speed on intersections with local geofencing is not feasible, as the geofences have to be extremely large, showing a lot of overlap on other intersections in cities due to high intersection density. Future work can be done on the intersection risk estimation method and the feasibility of speed-limiting geofences that employ active braking.
Bicycles have an important role to play in the transition towards a more sustainable mobility. In order to achieve the modal shift towards bicycles, more must be done to accommodate cyclists. Although controlled intersections do increase the (perceived) safety of crossings with motorized vehicles, they are seen as major obstacles, and cyclists tend to avoid them when possible. The negative effects of controlled intersections for cyclists may be reduced by new methods of intersection control. This thesis combines the concepts of the connected environment and structure free control, to design an intersection controller that uses a genetic algorithmto determine the optimal signal plan, hereafter referred to as the SFGA controller. The controller is designed for an isolated intersection and considers car drivers and cyclists. Desires of cyclist with regard to controlled intersections, are identified by means of a literature review on the determinants of bicycle use, which are then projected on the controlled intersection. A traffic system model, based on validated models found in literature, is set up and a design for the structure free controller is proposed. A set of control objectives is proposed, including different metrics related to the desires of cyclists and car drivers. Objective function weights can be varied to achieve different levels of cyclist prioritization. A maximumwaiting time of 100 seconds is enforced in order to prevent prioritization of cyclists to result in unreasonable delays for car drivers, because red light running probabilities increase at larger waiting times. The performance of the structure free controller is evaluated for 15, 35 and 45% traffic saturation (percentage of intersection capacity), by means of a simulation based case study. The designed controller is benchmarked to vehicle actuated control (VA). VA has a cyclic, fixed control structure in which green times of movements are flexible and depend on the queue size. SFGA is benchmarkedwith an equalweight for the delay of car drivers and cyclists, and no weight included for the number of stops. The effect of incorporating weights that prioritize the desires of cyclists over those of car drivers is investigated. The SFGA controller results in average delays 1.8, 2.7 and 3.0 times lower than VAC for each of the evaluated traffic saturation levels. The number of stops is 1.9, 2.3 and 3.1 times lower. Including weights in the objective function to explicitly prioritize cyclists, results in even lower average delays and number of stops for cyclists. As is to be expected, this comes at the cost of additional delays for car drivers, especially for higher traffic saturation. The better performance of SFGA is attributed to two main differences between the controllers. First of all, the structure free aspect allows for a larger degree of freedom to choose more effective combinations of traffic lights to show green at the same time, instead of following the fixed sequence of VA. Additionally, the controller allows traffic that otherwise would experience the largest total delay to cross first, even if this means delaying some travellers in close proximity of the traffic light. This contrary to VA, that extends green time based on detected traffic in the active block. Without inclusion of weights that prioritize the desires of cyclist over cars, the controller already tends towards prioritization of the cyclists. This is caused by the controller considering the number of travellers that are influenced by its’ control decisions, combined with the higher traffic densities, that can be expected on bicycle paths in urban areas. Weights to prioritize cyclists can be included to include more priority, for example when bicycle traffic volumes are low. This work implicates that, in order to better serve the cyclists, it is not explicitly required to prioritize cyclists over cars. In areas with large volumes of cyclists, considering the number of travellers and their proximity to the traffic light can already result in cyclists being served better. This work could be used as a starting point or inspiration to design and eventually implement more cyclist oriented intersection controllers. Improvements for the controller and extensions for the research scope are proposed that are required for the controller to be suitable for practical implementation in the real world. If a future version of the controller is to be implemented, it will reduce the negative effects of controlled intersections on cyclists, thereby making the bicycle a more suitable replacement for the car.
...
Bicycles have an important role to play in the transition towards a more sustainable mobility. In order to achieve the modal shift towards bicycles, more must be done to accommodate cyclists. Although controlled intersections do increase the (perceived) safety of crossings with motorized vehicles, they are seen as major obstacles, and cyclists tend to avoid them when possible. The negative effects of controlled intersections for cyclists may be reduced by new methods of intersection control. This thesis combines the concepts of the connected environment and structure free control, to design an intersection controller that uses a genetic algorithmto determine the optimal signal plan, hereafter referred to as the SFGA controller. The controller is designed for an isolated intersection and considers car drivers and cyclists. Desires of cyclist with regard to controlled intersections, are identified by means of a literature review on the determinants of bicycle use, which are then projected on the controlled intersection. A traffic system model, based on validated models found in literature, is set up and a design for the structure free controller is proposed. A set of control objectives is proposed, including different metrics related to the desires of cyclists and car drivers. Objective function weights can be varied to achieve different levels of cyclist prioritization. A maximumwaiting time of 100 seconds is enforced in order to prevent prioritization of cyclists to result in unreasonable delays for car drivers, because red light running probabilities increase at larger waiting times. The performance of the structure free controller is evaluated for 15, 35 and 45% traffic saturation (percentage of intersection capacity), by means of a simulation based case study. The designed controller is benchmarked to vehicle actuated control (VA). VA has a cyclic, fixed control structure in which green times of movements are flexible and depend on the queue size. SFGA is benchmarkedwith an equalweight for the delay of car drivers and cyclists, and no weight included for the number of stops. The effect of incorporating weights that prioritize the desires of cyclists over those of car drivers is investigated. The SFGA controller results in average delays 1.8, 2.7 and 3.0 times lower than VAC for each of the evaluated traffic saturation levels. The number of stops is 1.9, 2.3 and 3.1 times lower. Including weights in the objective function to explicitly prioritize cyclists, results in even lower average delays and number of stops for cyclists. As is to be expected, this comes at the cost of additional delays for car drivers, especially for higher traffic saturation. The better performance of SFGA is attributed to two main differences between the controllers. First of all, the structure free aspect allows for a larger degree of freedom to choose more effective combinations of traffic lights to show green at the same time, instead of following the fixed sequence of VA. Additionally, the controller allows traffic that otherwise would experience the largest total delay to cross first, even if this means delaying some travellers in close proximity of the traffic light. This contrary to VA, that extends green time based on detected traffic in the active block. Without inclusion of weights that prioritize the desires of cyclist over cars, the controller already tends towards prioritization of the cyclists. This is caused by the controller considering the number of travellers that are influenced by its’ control decisions, combined with the higher traffic densities, that can be expected on bicycle paths in urban areas. Weights to prioritize cyclists can be included to include more priority, for example when bicycle traffic volumes are low. This work implicates that, in order to better serve the cyclists, it is not explicitly required to prioritize cyclists over cars. In areas with large volumes of cyclists, considering the number of travellers and their proximity to the traffic light can already result in cyclists being served better. This work could be used as a starting point or inspiration to design and eventually implement more cyclist oriented intersection controllers. Improvements for the controller and extensions for the research scope are proposed that are required for the controller to be suitable for practical implementation in the real world. If a future version of the controller is to be implemented, it will reduce the negative effects of controlled intersections on cyclists, thereby making the bicycle a more suitable replacement for the car.
Traffic signals in a coordinated network normally use a common cycle length which remains constant at all times, including when there is a request for priority from a public transport vehicle. This enables green waves to be maintained effectively but can limit the signals' ability to promptly serve the prioritised vehicle.
To study the effects of momentarily relaxing the constraint of cycle length during Transit Signal Priority (TSP) interventions, a new TSP system is developed for a CRSV halfstarre traffic signal controller, which permits a flexible cycle length during priority interventions. That system is tested using a Vissim microsimulation of a simple fictional network, and compared to the existing fixed-cycle-length TSP system included with the controller.
The new TSP system permits TSP actions as long as it is expected that the signal can return to its normal "in sync" timings within two cycles. During the intervention, the positive and negative impacts on each signal phase are monitored, and "Offset Correction Credits" are distributed, which each represent one second of additional green time. Signal phases which received extra time during the TSP intervention will receive negative OC Credits, and phases which were truncated will receive positive OC Credits. Once the intervention is complete, the signal will execute "offset correction" to return the signal to its normal "In Sync" timings while redeeming OC Credits.
The subject road network consists of fictional road with three coordinated traffic
signals, spaced 150 metres and 400 metres apart. The central intersection is the capacity-critical intersection and also includes a frequent bus line (12 buses per hour per direction) travelling along a median busway perpendicular to the coordinated direction.
In the scenario with a high flexibility to reduce green durations, the average delay for late buses dropped by 59% from 10.7 secondsto 4.4 seconds for the flexible-cycle system compared to the fixed-cycle system. With low flexibility, the average delay for late buses dropped by 78% from 28.0 seconds to 6.2 seconds. The large improvements in performance for buses are due to the flexible-cycle TSP systems being able to execute more TSP actions such as phase insertions which may not fit within a fixed cycle length.
However, the controller’s ability to remain in sync was negatively impacted and the frequency of queues exceeding storage increased by as much as 70% on short roadway links. However on long roadway links, delaysand queue lengths decreased in the coordinated directions thanks to the new TSP system’s green time compensation mechanism.
When the assumed occupancy rate for late buses is 50 passengers (corresponding to a busy but not overcrowded standard bus), there was no significant difference in person-delay between any of the scenarios. Early buses were not included in the calculation for average person-delay. ...
Traffic signals in a coordinated network normally use a common cycle length which remains constant at all times, including when there is a request for priority from a public transport vehicle. This enables green waves to be maintained effectively but can limit the signals' ability to promptly serve the prioritised vehicle.
To study the effects of momentarily relaxing the constraint of cycle length during Transit Signal Priority (TSP) interventions, a new TSP system is developed for a CRSV halfstarre traffic signal controller, which permits a flexible cycle length during priority interventions. That system is tested using a Vissim microsimulation of a simple fictional network, and compared to the existing fixed-cycle-length TSP system included with the controller.
The new TSP system permits TSP actions as long as it is expected that the signal can return to its normal "in sync" timings within two cycles. During the intervention, the positive and negative impacts on each signal phase are monitored, and "Offset Correction Credits" are distributed, which each represent one second of additional green time. Signal phases which received extra time during the TSP intervention will receive negative OC Credits, and phases which were truncated will receive positive OC Credits. Once the intervention is complete, the signal will execute "offset correction" to return the signal to its normal "In Sync" timings while redeeming OC Credits.
The subject road network consists of fictional road with three coordinated traffic
signals, spaced 150 metres and 400 metres apart. The central intersection is the capacity-critical intersection and also includes a frequent bus line (12 buses per hour per direction) travelling along a median busway perpendicular to the coordinated direction.
In the scenario with a high flexibility to reduce green durations, the average delay for late buses dropped by 59% from 10.7 secondsto 4.4 seconds for the flexible-cycle system compared to the fixed-cycle system. With low flexibility, the average delay for late buses dropped by 78% from 28.0 seconds to 6.2 seconds. The large improvements in performance for buses are due to the flexible-cycle TSP systems being able to execute more TSP actions such as phase insertions which may not fit within a fixed cycle length.
However, the controller’s ability to remain in sync was negatively impacted and the frequency of queues exceeding storage increased by as much as 70% on short roadway links. However on long roadway links, delaysand queue lengths decreased in the coordinated directions thanks to the new TSP system’s green time compensation mechanism.
When the assumed occupancy rate for late buses is 50 passengers (corresponding to a busy but not overcrowded standard bus), there was no significant difference in person-delay between any of the scenarios. Early buses were not included in the calculation for average person-delay.
In recent years, the amount of traffic on the highways has increased continuously and in particular freight traffic. Despite solutions to maximize roadway capacity, the drop of capacity after congestions sets in, remains an active field of study. The influence of freight traffic on traffic flow has received little attention, especially regarding the capacity drop. The Kaplan-Meier Product Limit Method was used to estimate capacity and recovery distributions for selected sites to overcome stochastic characteristics of traffic flow and investigate the relation between heavy vehicle share and capacity drop at Dutch highways. Furthermore, a simulation study was executed to investigate increased heavy vehicle share scenarios and changing physical and operational characteristics of heavy vehicles. The empirical results show a connection between heavy vehicle share and the capacity drop, although not statistically significant. Currently, chaotic properties of breakdown flow seem to superimpose the impact. However, it is possible that the effect becomes influential as the current maximum observed share of heavy vehicles during breakdown grows from 9% up to 15%, which appeared to be the worst case scenario in simulation. Besides, concern is raised as a decreasing breakdown capacity was observed on several Dutch highways, even after correction with the PCE values of the increasing heavy vehicle share.
...
In recent years, the amount of traffic on the highways has increased continuously and in particular freight traffic. Despite solutions to maximize roadway capacity, the drop of capacity after congestions sets in, remains an active field of study. The influence of freight traffic on traffic flow has received little attention, especially regarding the capacity drop. The Kaplan-Meier Product Limit Method was used to estimate capacity and recovery distributions for selected sites to overcome stochastic characteristics of traffic flow and investigate the relation between heavy vehicle share and capacity drop at Dutch highways. Furthermore, a simulation study was executed to investigate increased heavy vehicle share scenarios and changing physical and operational characteristics of heavy vehicles. The empirical results show a connection between heavy vehicle share and the capacity drop, although not statistically significant. Currently, chaotic properties of breakdown flow seem to superimpose the impact. However, it is possible that the effect becomes influential as the current maximum observed share of heavy vehicles during breakdown grows from 9% up to 15%, which appeared to be the worst case scenario in simulation. Besides, concern is raised as a decreasing breakdown capacity was observed on several Dutch highways, even after correction with the PCE values of the increasing heavy vehicle share.
Speed limits and their effect on freeway capacity
An investigation of two lane freeway bottlenecks
Master thesis
(2020)
-
William van Lindonk, S.P. Hoogendoorn, V.L. Knoop, S. van Cranenburgh, H. Taale
In this thesis an investigation is performed into the effect of different speed limits on freeway capacity. From literature, much is known about the variety of factors that affect capacity, but the exact effect of the speed limit on capacity is not yet clear. In recent years, several speed limits changes have taken place at multiple two-lane freeway bottlenecks throughout The Netherlands, which makes it possible to compare effects of different speed limits at the same location. To evaluate the effect of the speed limit on capacity, the Product Limit Method has been applied to identify breakdown flows and generate capacity distributions, which could subsequently be compared for different limits. In the comparison of capacity distributions under different speed limits, it was found that significant changes in capacity had occurred, but that no uniform direction of the effect could be found. Subsequently, to control for location specific factors and other variables, Fixed Effects regression has been used to determine the effect of the speed limit on the breakdown flow. It was found that the breakdown flow under the 120 km/h limit was significantly higher than under the 130 km/h limit (in the range of 60 to 190 vehicles per hour) and that the breakdown flow under the 100 km/h limit was, in some cases, also higher than under the 130 km/h limit. In addition to this, it was found that a significant positive relation exists between the height of the speed limit and the fraction of flow in the passing lane. Moreover, it was shown that the relation between the fraction of flow in the passing lane and the level of breakdown flow was best represented by a quadratic relation, which could indicate that an “optimal” distribution of flows may exist. Given the results of this thesis, it is posed that a change in the speed limit is likely to affect capacity primarily through altering the lane flow distribution and that it will depend on the layout of a freeway location what the optimal lane flow distribution is and which speed limit leads to this optimal lane flow distribution.
...
In this thesis an investigation is performed into the effect of different speed limits on freeway capacity. From literature, much is known about the variety of factors that affect capacity, but the exact effect of the speed limit on capacity is not yet clear. In recent years, several speed limits changes have taken place at multiple two-lane freeway bottlenecks throughout The Netherlands, which makes it possible to compare effects of different speed limits at the same location. To evaluate the effect of the speed limit on capacity, the Product Limit Method has been applied to identify breakdown flows and generate capacity distributions, which could subsequently be compared for different limits. In the comparison of capacity distributions under different speed limits, it was found that significant changes in capacity had occurred, but that no uniform direction of the effect could be found. Subsequently, to control for location specific factors and other variables, Fixed Effects regression has been used to determine the effect of the speed limit on the breakdown flow. It was found that the breakdown flow under the 120 km/h limit was significantly higher than under the 130 km/h limit (in the range of 60 to 190 vehicles per hour) and that the breakdown flow under the 100 km/h limit was, in some cases, also higher than under the 130 km/h limit. In addition to this, it was found that a significant positive relation exists between the height of the speed limit and the fraction of flow in the passing lane. Moreover, it was shown that the relation between the fraction of flow in the passing lane and the level of breakdown flow was best represented by a quadratic relation, which could indicate that an “optimal” distribution of flows may exist. Given the results of this thesis, it is posed that a change in the speed limit is likely to affect capacity primarily through altering the lane flow distribution and that it will depend on the layout of a freeway location what the optimal lane flow distribution is and which speed limit leads to this optimal lane flow distribution.
Flooding events not only cause physical damage to the road infrastructure and roadside systems, yet they potentially lead to significant traffic disruptions over the network which have a large impact on the sustainability of urban cities. According to several climate change scenarios, flooding events as a result of intense precipitation are expected to continue into the future becoming more frequent and severe. Therefore, it is essential that engineers design and manage transport networks as efficiently as possible under adverse weather conditions in order to build more resilient transport systems. To do so, professionals need to understand and analyse the impacts of flooding on transport networks. Modelling the impacts of flooding events on the road network is nowadays possible by the use of integrated simulation techniques that combine a flood simulation model with a microscopic traffic model. With these modelling techniques, the geographic location, severity and other impacts of flooding on the road network can be predicted and studied. Nevertheless, in order to perform a consistent traffic simulation under flooding conditions, a thorough analysis of driving behaviour under such conditions needs first to be accomplished, which is a research field that has been scarcely investigated thus far. Therefore, the main contribution of this thesis is to provide a systematic methodology for studying driving adaptation effects under flooding conditions at two complementary levels, namely microscopic and macroscopic level. This consists on a multilevel approach that uses microscopic traffic data of vehicles driving through flooding to extend the study to a macroscopic level. A video analysis procedure called the `The 3-Step Video Analysis Approach´ (3SVAA) is developed by the author to extract microscopic traffic parameters from video recording of vehicles driving under different flood depths. This methodology is implemented through an empirical study that analyses two videos of vehicles crossing waterlogged stretches part of the A94 and A93 major roads in Scotland (UK). First, the 3SVAA is introduced as a suitable data collection technique to extract microscopic traffic variables (i.e. vehicle's speed, time headways and spacing) from video recording of vehicles driving through flooding. More specifically, these parameters are obtained from the vehicle's trajectories in the space (m)-time(s) diagram, which is the main outcome of this video analysis approach. Subsequently, by applying a multilevel analysis, the study can be extended to a macroscopic level by correlating macroscopic variables (i.e. average speed, flow and density) with their microscopic counterparts through the fundamental relationship of traffic flow. This allows to estimate part of the fundamental diagrams of vehicles travelling under different flood depths, and therefore to study flood impacts on free flow speed and capacity.
...
Flooding events not only cause physical damage to the road infrastructure and roadside systems, yet they potentially lead to significant traffic disruptions over the network which have a large impact on the sustainability of urban cities. According to several climate change scenarios, flooding events as a result of intense precipitation are expected to continue into the future becoming more frequent and severe. Therefore, it is essential that engineers design and manage transport networks as efficiently as possible under adverse weather conditions in order to build more resilient transport systems. To do so, professionals need to understand and analyse the impacts of flooding on transport networks. Modelling the impacts of flooding events on the road network is nowadays possible by the use of integrated simulation techniques that combine a flood simulation model with a microscopic traffic model. With these modelling techniques, the geographic location, severity and other impacts of flooding on the road network can be predicted and studied. Nevertheless, in order to perform a consistent traffic simulation under flooding conditions, a thorough analysis of driving behaviour under such conditions needs first to be accomplished, which is a research field that has been scarcely investigated thus far. Therefore, the main contribution of this thesis is to provide a systematic methodology for studying driving adaptation effects under flooding conditions at two complementary levels, namely microscopic and macroscopic level. This consists on a multilevel approach that uses microscopic traffic data of vehicles driving through flooding to extend the study to a macroscopic level. A video analysis procedure called the `The 3-Step Video Analysis Approach´ (3SVAA) is developed by the author to extract microscopic traffic parameters from video recording of vehicles driving under different flood depths. This methodology is implemented through an empirical study that analyses two videos of vehicles crossing waterlogged stretches part of the A94 and A93 major roads in Scotland (UK). First, the 3SVAA is introduced as a suitable data collection technique to extract microscopic traffic variables (i.e. vehicle's speed, time headways and spacing) from video recording of vehicles driving through flooding. More specifically, these parameters are obtained from the vehicle's trajectories in the space (m)-time(s) diagram, which is the main outcome of this video analysis approach. Subsequently, by applying a multilevel analysis, the study can be extended to a macroscopic level by correlating macroscopic variables (i.e. average speed, flow and density) with their microscopic counterparts through the fundamental relationship of traffic flow. This allows to estimate part of the fundamental diagrams of vehicles travelling under different flood depths, and therefore to study flood impacts on free flow speed and capacity.
This study aims at a possible solution to predict the fuel consumption of heavy duty diesel trucks, particularly, the tractor semitrailer for their long haul operations using various machine learning techniques. It intends to provide a possible alternative to simulation or physics based models, which often are very complicated. The stringent laws on emission control set by the Paris Agreement and the fact that heavy duty trucks contribute to almost 27% of CO2 emissions from road transport and their dependence on diesel for operations (in long haul) makes it the need of the hour to first, have an estimate on the emissions being produced and second, to develop technologies to reduce those emissions.
This study focuses specifically on the first part i.e., estimating the amount of fuel consumed by heavy duty trucks in the European Union and thereby determine the emissions being produced. The main objective is to examine whether an approach of machine learning could be a viable option to predict fuel consumption. This thesis is part of the AEROFLEX project and was done in collaboration with TNO, which provided all the data-sets required for the study.
The idea was to explore the regime of machine learning for one time step ahead prediction of fuel consumption. Furthermore, this study also focused on the development of another model by not using any variables affected by the driver as input into the training model. This exclusion was necessary to make sure the model remained adaptive to new routes and new trucks, especially because large scale on-road testing of the newly developed trucks is impossible and also because this way would help predict the fuel consumed by a truck without the necessity of it driving on a road. The study concludes with a comparison with an existing simulation model at TNO and provide an alternative machine learning solution. It also provides a comparison between different machine learning techniques and suggest the most accurate one.
It was found that machine learning could potentially be used to predict the amount of fuel consumed by a long haul heavy duty truck driving on a motorway. It was also found that engine torque was the variable that affected the fuel consumption of the truck the most. Furthermore, Neural Network was the most potent algorithm among all the other learning techniques for both the models developed in this study with it performing better than the simulation tool by a factor of approximately 3.8 in the model where the driver/drive influenced inputs were not considered in the training data-set. The results obtained from this work at a sampling frequency of 10 Hz. (i.e., 0.1 seconds) are comparable to the ones reported by other sources at a sampling rate of 0.016 Hz. (i.e., 1 minute) or 0.0016 Hz. (i.e., 10 minutes). This goes on to say that the machine learning algorithms are also potent at much higher sampling frequencies.
...
This study focuses specifically on the first part i.e., estimating the amount of fuel consumed by heavy duty trucks in the European Union and thereby determine the emissions being produced. The main objective is to examine whether an approach of machine learning could be a viable option to predict fuel consumption. This thesis is part of the AEROFLEX project and was done in collaboration with TNO, which provided all the data-sets required for the study.
The idea was to explore the regime of machine learning for one time step ahead prediction of fuel consumption. Furthermore, this study also focused on the development of another model by not using any variables affected by the driver as input into the training model. This exclusion was necessary to make sure the model remained adaptive to new routes and new trucks, especially because large scale on-road testing of the newly developed trucks is impossible and also because this way would help predict the fuel consumed by a truck without the necessity of it driving on a road. The study concludes with a comparison with an existing simulation model at TNO and provide an alternative machine learning solution. It also provides a comparison between different machine learning techniques and suggest the most accurate one.
It was found that machine learning could potentially be used to predict the amount of fuel consumed by a long haul heavy duty truck driving on a motorway. It was also found that engine torque was the variable that affected the fuel consumption of the truck the most. Furthermore, Neural Network was the most potent algorithm among all the other learning techniques for both the models developed in this study with it performing better than the simulation tool by a factor of approximately 3.8 in the model where the driver/drive influenced inputs were not considered in the training data-set. The results obtained from this work at a sampling frequency of 10 Hz. (i.e., 0.1 seconds) are comparable to the ones reported by other sources at a sampling rate of 0.016 Hz. (i.e., 1 minute) or 0.0016 Hz. (i.e., 10 minutes). This goes on to say that the machine learning algorithms are also potent at much higher sampling frequencies.
...
This study aims at a possible solution to predict the fuel consumption of heavy duty diesel trucks, particularly, the tractor semitrailer for their long haul operations using various machine learning techniques. It intends to provide a possible alternative to simulation or physics based models, which often are very complicated. The stringent laws on emission control set by the Paris Agreement and the fact that heavy duty trucks contribute to almost 27% of CO2 emissions from road transport and their dependence on diesel for operations (in long haul) makes it the need of the hour to first, have an estimate on the emissions being produced and second, to develop technologies to reduce those emissions.
This study focuses specifically on the first part i.e., estimating the amount of fuel consumed by heavy duty trucks in the European Union and thereby determine the emissions being produced. The main objective is to examine whether an approach of machine learning could be a viable option to predict fuel consumption. This thesis is part of the AEROFLEX project and was done in collaboration with TNO, which provided all the data-sets required for the study.
The idea was to explore the regime of machine learning for one time step ahead prediction of fuel consumption. Furthermore, this study also focused on the development of another model by not using any variables affected by the driver as input into the training model. This exclusion was necessary to make sure the model remained adaptive to new routes and new trucks, especially because large scale on-road testing of the newly developed trucks is impossible and also because this way would help predict the fuel consumed by a truck without the necessity of it driving on a road. The study concludes with a comparison with an existing simulation model at TNO and provide an alternative machine learning solution. It also provides a comparison between different machine learning techniques and suggest the most accurate one.
It was found that machine learning could potentially be used to predict the amount of fuel consumed by a long haul heavy duty truck driving on a motorway. It was also found that engine torque was the variable that affected the fuel consumption of the truck the most. Furthermore, Neural Network was the most potent algorithm among all the other learning techniques for both the models developed in this study with it performing better than the simulation tool by a factor of approximately 3.8 in the model where the driver/drive influenced inputs were not considered in the training data-set. The results obtained from this work at a sampling frequency of 10 Hz. (i.e., 0.1 seconds) are comparable to the ones reported by other sources at a sampling rate of 0.016 Hz. (i.e., 1 minute) or 0.0016 Hz. (i.e., 10 minutes). This goes on to say that the machine learning algorithms are also potent at much higher sampling frequencies.
This study focuses specifically on the first part i.e., estimating the amount of fuel consumed by heavy duty trucks in the European Union and thereby determine the emissions being produced. The main objective is to examine whether an approach of machine learning could be a viable option to predict fuel consumption. This thesis is part of the AEROFLEX project and was done in collaboration with TNO, which provided all the data-sets required for the study.
The idea was to explore the regime of machine learning for one time step ahead prediction of fuel consumption. Furthermore, this study also focused on the development of another model by not using any variables affected by the driver as input into the training model. This exclusion was necessary to make sure the model remained adaptive to new routes and new trucks, especially because large scale on-road testing of the newly developed trucks is impossible and also because this way would help predict the fuel consumed by a truck without the necessity of it driving on a road. The study concludes with a comparison with an existing simulation model at TNO and provide an alternative machine learning solution. It also provides a comparison between different machine learning techniques and suggest the most accurate one.
It was found that machine learning could potentially be used to predict the amount of fuel consumed by a long haul heavy duty truck driving on a motorway. It was also found that engine torque was the variable that affected the fuel consumption of the truck the most. Furthermore, Neural Network was the most potent algorithm among all the other learning techniques for both the models developed in this study with it performing better than the simulation tool by a factor of approximately 3.8 in the model where the driver/drive influenced inputs were not considered in the training data-set. The results obtained from this work at a sampling frequency of 10 Hz. (i.e., 0.1 seconds) are comparable to the ones reported by other sources at a sampling rate of 0.016 Hz. (i.e., 1 minute) or 0.0016 Hz. (i.e., 10 minutes). This goes on to say that the machine learning algorithms are also potent at much higher sampling frequencies.
Master thesis
(2019)
-
Ting Wei Wu, Lorant Tavasszy, Alessandro Bombelli, Bilge Atasoy, Milan Janić, Victor Knoop
Congestion in the landside air cargo supply chain occurs for different concurring reasons. Lack of coordination between freight forwarders, as example, might create truck congestion on the ground handler side. Horizontal collaboration between forwarders can be introduced and modeled via mathematical programming to mitigate congestion. Resulting models, which are generally variations of Pickup and Delivery Problem with Time Windows (PDPTW), can be solved to optimality only for small-size instances, and the computation is generally time consuming. We therefore propose a simulated annealing (SA)-embedded adaptive large neighborhood search (ALNS) heuristic to address truck route planning in the landside air cargo supply chain. In this work, we allow the search to visit infeasible time-dependent solutions. Accordingly, the objective function minimizes the feasible solution, where total travel distance cost, total travel time cost and unassigned shipments cost, and the time-dependent violation costs. Computational results are reported for 10 instances that were also solved with a mathematical programming approach. Results shows that the meta-heuristic method performs equally or better than the mathematical model given a computational limit for the latter of 2 hours. In addition, the meta-heuristic method was able to find a feasible solution for those cases where the exact model failed to identify a feasible solution.
...
Congestion in the landside air cargo supply chain occurs for different concurring reasons. Lack of coordination between freight forwarders, as example, might create truck congestion on the ground handler side. Horizontal collaboration between forwarders can be introduced and modeled via mathematical programming to mitigate congestion. Resulting models, which are generally variations of Pickup and Delivery Problem with Time Windows (PDPTW), can be solved to optimality only for small-size instances, and the computation is generally time consuming. We therefore propose a simulated annealing (SA)-embedded adaptive large neighborhood search (ALNS) heuristic to address truck route planning in the landside air cargo supply chain. In this work, we allow the search to visit infeasible time-dependent solutions. Accordingly, the objective function minimizes the feasible solution, where total travel distance cost, total travel time cost and unassigned shipments cost, and the time-dependent violation costs. Computational results are reported for 10 instances that were also solved with a mathematical programming approach. Results shows that the meta-heuristic method performs equally or better than the mathematical model given a computational limit for the latter of 2 hours. In addition, the meta-heuristic method was able to find a feasible solution for those cases where the exact model failed to identify a feasible solution.
New Intersection Control for Conventional and Automated Vehicles without Traffic Lights
A combination of individual control and self-regulation
This thesis aims to develop a novel control strategy that efficiently controls traffic with different penetration rates by relying purely on wireless communication to integrate the control of automated and human drivers. The new strategy uses the concept of induced platooning to use automated vehicle to indirectly control conventional vehicles. The control of these induced platoons is based on the combination of self-regulation and individual trajectory control. Results show that the new strategy is able to efficiently coordinate vehicles at an intersection and it can be considered as candidate solution to manage traffic during the transition phase where conventional and automated vehicles coexist.
...
This thesis aims to develop a novel control strategy that efficiently controls traffic with different penetration rates by relying purely on wireless communication to integrate the control of automated and human drivers. The new strategy uses the concept of induced platooning to use automated vehicle to indirectly control conventional vehicles. The control of these induced platoons is based on the combination of self-regulation and individual trajectory control. Results show that the new strategy is able to efficiently coordinate vehicles at an intersection and it can be considered as candidate solution to manage traffic during the transition phase where conventional and automated vehicles coexist.
How to calibrate a pedestrian simulation model
An investigation into how the choices of scenarios and metrics influence the calibration
This research investigates the question how the choice of scenarios and metrics influences the calibration of pedestrian simulation models. By calibrating a pedestrian model using different combinations of scenarios and metrics and comparing the results it was concluded that it matters what scenarios and metrics are included during the calibration. The implication of this is that one should calibrate a pedestrian model based on its intended usage and that this should be leading in the choice of which scenarios and metrics to use.
...
This research investigates the question how the choice of scenarios and metrics influences the calibration of pedestrian simulation models. By calibrating a pedestrian model using different combinations of scenarios and metrics and comparing the results it was concluded that it matters what scenarios and metrics are included during the calibration. The implication of this is that one should calibrate a pedestrian model based on its intended usage and that this should be leading in the choice of which scenarios and metrics to use.
Car-Following Model using Machine Learning Techniques
Approach at Urban Signalized Intersections with Traffic Radar Detection
Master thesis
(2017)
-
Ignasi Echaniz Soldevila, Serge Hoogendoorn, Victor Knoop, Jeroen Steenbakkers, Javier Alonso Mora
This master thesis aims to gain new empirical insights into longitudinal driving behavior by means of the enumeration of a new hybrid car-following (CF) model which combines parametric and non parametric formulation. On one hand, the model, which predicts the drivers acceleration given a set of variables, benefits from innovative machine learning techniques such as Gaussian process regression (GPR) to make predictions when there exist correlation between new input and the training dataset. On the other hand, it uses existent traditional parametric CF models to predict acceleration when no similar situations are found in the training dataset. This formulation guarantees a complete and continues model and deals with the challenges of new available types of dataset in the transport field: noisy and incomplete yet with large amount of data. Multiple models have been trained using the Optimal Velocity Model (OVM) as a basis parametric model and a dataset collected in the PPA project in Amsterdam by traffic radar detection in stop and go traffic conditions. The other main innovation of this thesis is that variables rarely included in any CF model such as the status and the distance of drivers to the traffic light are also analyzed. Results show that the GPR model formulation is robust as the model performs better than OVM alone according to the main KPI, but still collisions occasionally occur. Moreover, results depict that traffic light status actively influences driver behavior. Overall, this thesis gives insights into new powerful mathematical techniques that can be applied to describe longitudinal driving behavior or any modeled process.
...
This master thesis aims to gain new empirical insights into longitudinal driving behavior by means of the enumeration of a new hybrid car-following (CF) model which combines parametric and non parametric formulation. On one hand, the model, which predicts the drivers acceleration given a set of variables, benefits from innovative machine learning techniques such as Gaussian process regression (GPR) to make predictions when there exist correlation between new input and the training dataset. On the other hand, it uses existent traditional parametric CF models to predict acceleration when no similar situations are found in the training dataset. This formulation guarantees a complete and continues model and deals with the challenges of new available types of dataset in the transport field: noisy and incomplete yet with large amount of data. Multiple models have been trained using the Optimal Velocity Model (OVM) as a basis parametric model and a dataset collected in the PPA project in Amsterdam by traffic radar detection in stop and go traffic conditions. The other main innovation of this thesis is that variables rarely included in any CF model such as the status and the distance of drivers to the traffic light are also analyzed. Results show that the GPR model formulation is robust as the model performs better than OVM alone according to the main KPI, but still collisions occasionally occur. Moreover, results depict that traffic light status actively influences driver behavior. Overall, this thesis gives insights into new powerful mathematical techniques that can be applied to describe longitudinal driving behavior or any modeled process.