Showing papers in "Transportation Research Part C-emerging Technologies in 2022"

Decision making of autonomous vehicles in lane change scenarios: Deep reinforcement learning approaches with risk awareness

Abstract: • A lane change decision making framework based on deep reinforcement learning is proposed. • A probabilistic-model based risk assessment method is proposed to assess the driving risk. • A risk-aware strategy with the minimum expected risk is developed for autonomous driving. • Our proposed methods have superior lane change driving performance in both static and moving scenarios. • Our proposed methods can be applied for safe autonomous driving in dangerous situations. Driving safety is the most important element that needs to be considered for autonomous vehicles (AVs). To ensure driving safety, we proposed a lane change decision-making framework based on deep reinforcement learning to find a risk-aware driving decision strategy with the minimum expected risk for autonomous driving. Firstly, a probabilistic-model based risk assessment method was proposed to assess the driving risk using position uncertainty and distance-based safety metrics. Then, a risk aware decision making algorithm was proposed to find a strategy with the minimum expected risk using deep reinforcement learning. Finally, our proposed methods were evaluated in CARLA in two scenarios (one with static obstacles and one with dynamically moving vehicles). The results show that our proposed methods can generate robust safe driving strategies and achieve better driving performances than previous methods.

Estimating key traffic state parameters through parsimonious spatial queue models

Abstract: As an active performance evaluation method, the fluid-based queueing model plays an important role in traffic flow modeling and traffic state estimation problems. A critical challenge in the application of traffic state estimation is how to utilize heterogeneous data sources in identifying key interpretable model parameters of freeway bottlenecks, such as queue discharge rates, system-level bottleneck-oriented arrival rates, and congestion duration. Inspired by Newell’s deterministic fluid approximation model, this paper proposes a spatial queue model for oversaturated traffic systems with time-dependent arrival rates. The oversaturated system dynamics can be described by parsimonious analytical formulations based on polynomial functional approximation for virtual arrival flow rates. With available flow, density and end-to-end travel time data along traffic bottlenecks, the proposed modeling framework for estimating the key traffic queueing state parameters is able to systematically map various measurements to the bottleneck-level dynamics and queue evolution process. The effectiveness of the developed method is demonstrated based on three case studies with empirical data in different metropolitan areas, including New York, Los Angeles, and Beijing.

A survey on demand-responsive public bus systems

Abstract: When demand for transportation is low or highly variable, traditional public bus services tend to lose their efficiency and typically frustrate (potential) passengers. In the literature, a large number of demand-responsive systems, that promise improved flexibility, have therefore been developed. At present, however, a comprehensive survey of these systems is lacking. In this paper, we fill this gap by presenting a unifying framework that classifies all demand-responsive public bus systems. The classification is mainly based on three degrees of responsiveness: dynamic online, dynamic offline, and static. For each system we discuss the specific optimization problem modeled, whether realistic data is considered, and the size of the instances used for testing. Where possible, we try to draw conclusions on the current state of the literature and try to identify potential avenues for future research. Different tables are included to structure and summarize the information of all papers.

A literature review of Artificial Intelligence applications in railway systems

Abstract: Nowadays it is widely accepted that Artificial Intelligence (AI) is significantly influencing a large number of domains, including railways. In this paper, we present a systematic literature review of the current state-of-the-art of AI in railway transport. In particular, we analysed and discussed papers from a holistic railway perspective, covering sub-domains such as maintenance and inspection, planning and management, safety and security, autonomous driving and control, revenue management, transport policy, and passenger mobility. This review makes an initial step towards shaping the role of AI in future railways and provides a summary of the current focuses of AI research connected to rail transport. We reviewed about 139 scientific papers covering the period from 2010 to December 2020. We found that the major research efforts have been put in AI for rail maintenance and inspection, while very limited or no research has been found on AI for rail transport policy and revenue management. The remaining sub-domains received mild to moderate attention. AI applications are promising and tend to act as a game-changer in tackling multiple railway challenges. However, at the moment, AI research in railways is still mostly at its early stages. Future research can be expected towards developing advanced combined AI applications (e.g. with optimization), using AI in decision making, dealing with uncertainty and tackling newly rising cybersecurity challenges.

Comfortable and energy-efficient speed control of autonomous vehicles on rough pavements using deep reinforcement learning

Abstract: Rough pavements cause ride discomfort and energy inefficiency for road vehicles. Existing methods to address these problems are time-consuming and not adaptive to changing driving conditions on rough pavements. With the development of sensor and communication technologies, crowdsourced road and dynamic traffic information become available for enhancing driving performance, particularly addressing the discomfort and inefficiency issues by controlling driving speeds. This study proposes a speed control framework on rough pavements, envisioning the operation of autonomous vehicles based on the crowdsourced data. We suggest the concept of ‘maximum comfortable speed’ for representing the vertical ride comfort of oncoming roads. A deep reinforcement learning (DRL) algorithm is designed to learn comfortable and energy-efficient speed control strategies. The DRL-based speed control model is trained using real-world rough pavement data in Shanghai, China. The experimental results show that the vertical ride comfort, energy efficiency, and computation efficiency increase by 8.22%, 24.37%, and 94.38%, respectively, compared to an optimization-based speed control model. The results indicate that the proposed framework is effective for real-time speed controls of autonomous vehicles on rough pavements.

DDP-GCN: Multi-graph convolutional network for spatiotemporal traffic forecasting

Abstract: Traffic speed forecasting is one of the core problems in transportation systems. For a more accurate prediction, recent studies started using not only the temporal speed patterns but also the spatial information on the road network through the graph convolutional networks. Even though the road network is highly complex due to its non-Euclidean and directional characteristics, previous approaches mainly focused on modeling the spatial dependencies using the distance only. In this paper, we identify two essential spatial dependencies in traffic forecasting in addition to distance, direction and positional relationship, for designing basic graph elements as the fundamental building blocks. Using the building blocks, we suggest DDP-GCN (Distance, Direction, and Positional relationship Graph Convolutional Network) to incorporate the three spatial relationships into deep neural networks. We evaluate the proposed model with two large-scale real-world datasets, and find positive improvements for long-term forecasting in highly complex urban networks. The improvement can be larger for commute hours, but it can be also limited for short-term forecasting.

Real-time joint traffic state and model parameter estimation on freeways with fixed sensors and connected vehicles: State-of-the-art overview, methods, and case studies

Abstract: This paper addresses real-time joint traffic state and model parameter estimation on freeways using data from fixed sensors and connected vehicles. It investigates how the combined usage of both types of sensing data improves the performance of traffic state estimation (TSE) and what role the online model parameter estimation (OMPE) plays therein. The paper first presents a state-of-the-art overview for freeway TSE with mixed sensing, focusing on a few critical issues such as filtering methods, Eulerian and Lagrangian formulation for traffic flow modeling/sensing/estimation, OMPE, and fusion of disparate sensing data, to determine the strengths and weaknesses of various technical paths, and figure out a viable roadmap for future studies. Three representative approaches to the design of freeway traffic state estimators using mixed sensing data are then investigated, which are based on a first-order, a second-order traffic flow model, and a speed-uniformity assumption, respectively. The paper intends to check if the gradual richness of mobile sensing data (in the era of connected vehicles) would compensate the deficiency of first-order models (as compared to second-order models) for TSE; if OMPE would still be essential for TSE in the mixed sensing case compared to the fixed sensing case; if the increasing usage of mobile sensing data would reduce the necessity of OMPE for TSE? The designed traffic state estimators have been evaluated thoroughly using NGSIM data, with the above questions answered.

Do autonomous vehicles drive like humans? A Turing approach and an application to SAE automation Level 2 cars

Abstract: • If AVs drove like humans, they would reduce interaction problems with drivers and passengers. • The ability of AVs not to be distinguished from a human driver was tested through a Turing approach. • A real on the road experiment with 550 university students was performed in Italy. • In most cases the Artificial Intelligence (AI) was indistinguishable from the human driver. • Artificial Intelligence of the cruise control is less recognizable than that of the lane keeping. Fully automated vehicles (AVs) are set to become a reality in future decades and changes are to be expected in user perceptions and behavior. While AV acceptability has been widely studied, changes in human drivers’ behavior and in passengers’ reactions have received less attention. It is not yet possible to ascertain the risk of driver behavioral changes such as overreaction, and the corresponding safety problems, in mixed traffic with partially AVs. Nor has there been proper investigation of the potential unease of car occupants trained for human control, when exposed to automatic maneuvers. The conjecture proposed in this paper is that automation Level 2 vehicles do not induce potentially adverse effects in traditional vehicle drivers’ behavior or in occupants’ reactions, provided that they are indistinguishable from human-driven vehicles. To this end, the paper proposes a Turing approach to test the “humanity” of automation Level 2 vehicles. The proposed test was applied to the results of an experimental campaign carried out in Italy: 546 car passengers were interviewed on board Level 2 cars in which they could not see the driver. They were asked whether a specific driving action (braking, accelerating, lane keeping) had been performed by the human driver or by the automatic on-board software under different traffic conditions (congestion and speed). Estimation results show that in most cases the interviewees were unable to distinguish the Artificial Intelligence (AI) from the human driver by observing random responses with a 95% significance level (proportion of success statistically equal to 50%). However, in the case of moderate braking and lane keeping at >100 km/h and in high traffic congestion, respondents recognized AI control from the human driver above pure chance, with 62–69% correct response rates. These findings, if confirmed in other case studies, could significantly impact on AVs acceptability, also contributing to their design as well as to long-debated ethical questions. AI driving software could be designed and tested for “humanity”, as long as safety is guaranteed, and autonomous cars could be allowed to circulate as long as they cannot be distinguished from human-driven vehicles in recurrent driving conditions.

How many are enough?: Investigating the effectiveness of multiple conflict indicators for crash frequency-by-severity estimation by automated traffic conflict analysis

Abstract: • Multivariate Extreme value models based on Gumbel Copulas are used to estimate crash frequency-by-severity from traffic conflict indicators. • Traffic conflict indicators are extracted from videos by applying an automated computer vision technique. • The accuracy and precision of crash predictions are not proportional to the number of conflict indicators used in the extreme value models. • Modified Time to Collision (MTTC) and Deceleration Rate to Avoid a Crash (DRAC) is the best combination of indicators for rear-end crash frequency estimation. • A trivariate model with MTTC, DRAC and Delta-V efficiently estimates crash frequency-by-severity. Traffic conflict techniques are a viable alternative to crash-based safety assessments and are particularly well suited to evaluating emerging technologies such as connected and automated vehicles for which crash data are sparsely available. Recently, the use of multiple traffic conflict indicators has become common in methodological studies, yet it is often difficult to determine which conflict indicators are appropriate given the application context, and the net benefit, in terms of improved crash prediction accuracy, of considering additional conflict indicators. Addressing these concerns, this study investigates the potential benefits of multiple conflict indicators for conflict-based crash estimation models by using a multivariate extreme value modeling framework (with Gumbel-Hougaard copulas) to estimate crash frequency by severity. The selected conflict indicators include Modified Time-To-Collision (MTTC), Deceleration Rate to Avoid a Collision (DRAC), Proportion of Stopping Distance (PSD) and expected post-collision change in velocity (Delta-V). The proposed framework was applied to estimate the total, severe (Maximum Abbreviated Injury Scale ≥ 3; MAIS3+), and non-severe (MAIS < 3) rear-end crash frequencies at three four-legged signalized intersections in Brisbane, Australia. Rear-end traffic conflicts were extracted from video data using state-of-the-art Computer Vision analytics. Results show that the prediction performance improvements are not necessarily proportional to the number of conflict indicators used in extreme value models. MTTC and DRAC, combined with the severity indicator Delta-V, were the most suitable predictors of rear-end crashes at signalized intersections. Results suggest that instead of adding more and more conflict indicators, careful selection of compatible conflict indicators (considering their functional differences and empirical correlations) is the best way to enhance the predictive performance of conflict-based models.

Driver distraction detection based on vehicle dynamics using naturalistic driving data

Abstract: Distracted driving such as phone use during driving is risky, as it increases the probability of severe crashes. Detecting distraction using Naturalistic Driving Studies was attempted in existing studies, and most of them used facial motions, which would be highly influenced by light conditions and algorithm effectiveness, still could not fully indicate auditory and physical distractions. This study aims to optimize Long Short-Term Memory (LSTM) model for phone usage detection based on vehicle dynamics sensor data from Shanghai Naturalistic Driving Study (SH-NDS), China. A total of 1244 phone use events were extracted from videos of SH-NDS, and analyzed against focus driving baseline. Performance attributes included speed, longitudinal acceleration, lateral acceleration, lane offset, and steering wheel rate. Their mean, standard deviation, and predicted error (PE) were calculated, and derived 15 indicators. A Bidirectional layer and attention mechanism were added to the LSTM model for higher accuracy. Results showed that besides the mean and standard deviation of steering wheel rate, all the other 13 indicators were significant and effective in the model. The Bidirectional Long Short-Term Memory (Bi-LSTM) model reached a promising result of approximately 91.2% accuracy using 5-fold cross validation, which was better than other machine learning methods such as recurrent neural network, support vector machine, k-nearest neighbor, and adaptive boosting. This Bi-LSTM model with attention mechanism could potentially be applied in advanced driving assistant systems to warn driver and reduce phone involved distracted driving.

TrajGAT: A map-embedded graph attention network for real-time vehicle trajectory imputation of roadside perception

Abstract: With the increasing deployment of roadside sensors, vehicle trajectories can be collected for driving behavior analysis and vehicle-highway automation systems. However, due to dynamic occlusions, vehicles are often lost from the view of roadside sensors, strongly affecting the data availability. To address this issue, we propose a novel deep learning framework to impute missing Trajectory data called map-embedded Graph ATtention network (TrajGAT). The framework splits the problem into two subtasks, a trajectory forecasting task based on historical data and an imputation task based on the forecasting results and real-time incomplete observational data. Temporal features are extracted and fused following an encoder-decoder architecture. To model dynamic spatial patterns, we introduce a sparse heterogeneous graph construct technique via vectorized lane-level map and a rule-based graph attention network, which can effectively capture remote dependencies and highlight key adjacency relationships. Numerical experiments based on the Argoverse imputation dataset and Lyft dataset are conducted to compare our TrajGAT and other state-of-the-art models. The results indicate that our model performs best based on various evaluation indicators and has strong robustness with different missing trajectory rates. The learned dynamic interaction can further help achieve a better understanding of the spatiotemporal dependency of vehicles in complex traffic scenarios.

A joint optimisation model for charger locating and electric bus charging scheduling considering opportunity fast charging and uncertainties

Abstract: The continuous improvement of charging technologies makes fast charging applicable to overcoming the limited driving range of electric vehicles by providing en-route opportunity charging. Fast chargers can be deployed at bus stops to top up the battery for an electric bus during its dwelling time when passengers board and alight, which allows the usage of small size battery and avoids deadheading trips to depot or charging station for recharging. This paper examines the joint optimisation problem of locating fast chargers at selected bus stops and developing charging schedule to support electric bus operation considering time-varying electricity price and penalty cost if passengers’ extra waiting time, the delay time of starting a trip or the excessive travel time of completing a trip is caused by charging activities. To handle the uncertainties associated with the travel time and passenger demand, a robust model is proposed and formulated. The proposed methods are applied to multiple bus routes in Sydney in the numerical studies. The numerical results show that frequent charging activities occur at both intermediate and last stops during bus service trips, indicating that it is beneficial for buses to utilise passengers’ boarding and alighting time at intermediate stops and resting time at the last stop between two adjacent trips to top up battery. Sensitivity analysis on a series of system parameters, such as the acquisition cost of chargers, amortized battery price and charging power, is conducted. The numerical results verify the feasibility of deployment of fast chargers and corresponding top-up charging activities at both intermediate and last stops. Electric buses outperform the conventional diesel buses regarding energy consumption cost, especially considering the increasing diesel fuel price and its costly environmental impacts. Moreover, charging activities at intermediate stops will become more common when passenger demand (the volume of boarding and alighting passengers) increases at intermediate stops. • En-route top-up charging schedule at selected bus stops for electric buses. • Optimal deployment of fast chargers at selected bus stops for electric buses. • System cost minimisation via charger location, battery size and charging schedule. • Uncertain passenger demand and travel time between stops during bus operation. • Sensitivity analysis of various factors on the charging locations and schedule.

Processing, assessing, and enhancing the Waymo autonomous vehicle open dataset for driving behavior research

Abstract: Recently released Autonomous Vehicle (AV) trajectory datasets can potentially catalyze research progress on AV-oriented traffic flow analysis. This paper aims to comprehensively and systematically process and assess one of the AV-oriented open datasets, i.e., Waymo Open Dataset, with a focus on car following paired trajectories. First, the original dataset has been processed into a user-friendly format which contains all important information related to the behavior of AV and surrounding objects. Second, the data quality has been assessed in terms of internal consistency, jerk values and trajectory completeness. Results show that the extracted trajectories are all incomplete but generally they have better quality than that of Next Generation Simulation program (NGSIM) dataset. Third, the trajectory data has been further enhanced by using an optimization-based outlier removal method and a wavelet denoising method. Additionally, we have tested the impact of data outliers and noise on IDM calibration, and revealed significant differences in parameter values for desired time gap T and maximum acceleration a.

Cooperative signal-free intersection control using virtual platooning and traffic flow regulation

Abstract: The emerging technologies of connectivity and automation enable the potential for signal-free intersection control. In this context, virtual platooning is posited to be an innovative, decentralized control strategy that maps two-dimensional vehicle movements onto a one-dimension virtual platoon to enable intersection operations. However, the effectiveness of virtual platooning-based control can be limited or degraded by parametric inaccuracies and unparameterized disturbances in vehicle dynamics, heavy traffic congestion, and/or uncoordinated platoons in multi-lane intersections. To explicitly address these limitations, this study proposes a hybrid cooperative intersection control framework consisting of microscopic-level virtual platooning control and macroscopic-level traffic flow regulation for traffic environments with connected autonomous vehicles. In virtual platooning control, vehicles approaching an intersection are organized into coordinated independent virtual platoons to avoid potential conflicts triggered by platoon formation changes. Through coordination, vehicles in a platoon are grouped into compatible passing sets to maintain desired safe spacing when proceeding through the intersection. We propose a distributed adaptive sliding mode controller (DASMC) which uses the backstepping control method and model reference adaptive control method to address parametric inaccuracies, and the sliding mode control method to consistently suppress the negative effects of the unparameterized disturbances. Each vehicle approaching the intersection utilizes the kinematic information from neighboring vehicles to implement the DASMC in a distributed manner such that vehicles within the same virtual platoon can achieve consensus safely. However, virtual platooning control cannot preclude excessive traffic from approaching the intersection, which can cause undesired spillbacks and degrade intersection control performance. To address this issue, traffic flow regulation is integrated with the virtual platooning control using an iterative feedback loop mechanism. In each iteration of the iterative feedback loop, a constrained finite-time optimal control (CFTOC) problem is solved to determine the optimal input flow permitted to proceed through the intersection, and the virtual platooning control provides feedback on the queue status to the CFTOC to initiate the next iteration. The effectiveness of the proposed intersection control framework is evaluated through numerical experiments. The results indicate that the proposed virtual platooning DASMC controller can mitigate the effects of parametric inaccuracies and unparameterized disturbances to achieve consensus for approaching vehicles, as well as guarantee string stability. Further, the proposed framework can alleviate traffic spillbacks and travel delays effectively through traffic flow regulation.

Analysis of cooperative driving strategies at road network level with macroscopic fundamental diagram

Abstract: • This paper investigates the key ways for automated vehicles (AVs) to improve the traffic efficiency of two-dimensional urban network traffic. • Different cooperative driving strategies are investigated in a systematic and comprehensive way, and we find that the passing order has a dominant impact on the network traffic efficiency. • This paper compares the differences and characteristics of the impact of the passing orders and the car-following gaps on the urban network traffic. • By using the macroscopic fundamental diagram (MFD) as the performance metric, we find that a better passing order can broaden the free flow region and improve the maximum flow, which is of great benefit to the network traffic efficiency. Cooperative driving, especially the passing order, is the critical link to improve the traffic efficiency of the road network by using automated vehicles (AVs). However, most studies have only considered the performance of the passing order at the isolated intersection and have not yet investigated its impact on the road network. In this paper, we will focus on the performance of the passing orders derived from different cooperative driving strategies on the network traffic through a series of simulation experiments. Meanwhile, we will compare the impacts of the passing order at intersections and the car-following gap in straight links on the network traffic efficiency. The experiments results show that the passing order has a dominant impact on the network traffic efficiency, and a better order can significantly raise the curve of the macroscopic fundamental diagram (MFD); due to the inevitable conflicts in the two-dimensional traffic, the choice of the car-following gap within a reasonable range has a relatively small improvement on the network traffic efficiency. The findings in this paper have instructive significance for the rising research on network-wide cooperative driving and provide a systematical perspective for network traffic control.

Evaluation of e-scooters as transit last-mile solution

Abstract: E-scooters are an alternative for short trips and are particularly suitable for solving the last-mile transit problem, yet their impact on transit is not well understood. There is a need to understand the e-scooter demand patterns and users’ characteristics to develop adequate policies and regulations. In this research, we consider the problem of modeling the interaction of e-scooters and bus transit services and provide an overview of e-scooter trips and user characteristics. We use a revealed-preference survey to evaluate the e-scooter usage in one of the highest-demand areas in the City of Austin, corresponding to a university campus. We explore population characteristics, mode shift, and mode interaction. Then, using publicly available datasets, we provide a causal analysis to evaluate the nature of the relationship between e-scooter and transit trips in the whole city. Assessing this relationship is challenging because several factors affect the demand of both types of trips (e.g., location of attractive zones), known as confounding variables. We develop a methodological framework to isolate the effects of confounding variables on transit trips using a two-stage regression procedure. The first stage aims to isolate confounding variables using a gradient boosting regression. The second stage models first and last-mile trips using a negative binomial and a zero-inflated negative binomial count model. The university survey indicated that 12 percent of the e-scooter users employed transit to complement their trips. Although small in magnitude, the data modeling results show that a statistically significant relationship was found on the university campus and downtown areas, supporting the survey results and extending the analysis to other areas of the city. However, the overall interaction between the two modes has a small magnitude. The proposed methodology can be used to identify areas with potential e-scooter and transit interaction. • We develop a methodological framework to model e-scooters and transit trips • It allows isolating confounding variables using a two-stage regression procedure • The university population is surveyed to understand the e-scooter demand in this area • Results show a statistically significant relationship in certain areas • This methodology can identify areas with potential e-scooter and transit interaction

A control strategy for merging a single vehicle into a platoon at highway on-ramps

Abstract: An important topic of research regarding cooperative platoons is merging vehicles into a platoon at highway on-ramps. This paper proposes a control strategy for the merging of a single cooperative automated vehicle into a platoon of vehicles at highway on-ramps. The proposed strategy can handle large differences in initial positions and velocities, sensor noise, and disturbances caused by the platoon leader. Furthermore, the required controller transitions are designed such that the switch between regular platooning and the merging maneuver can easily be made by all vehicles. The proposed strategy is validated using simulations. In the simulation environment communication delays, sensor noise, and disturbances of the platoon leader have been included. The proposed strategy is compared to a traditional strategy and shows a clear improvement in terms of noise handling. Furthermore, the proposed strategy behaves satisfactory considering safety, efficiency, passenger comfort, and disturbance handling.

Unsupervised hierarchical methodology of maritime traffic pattern extraction for knowledge discovery

Abstract: Owing to the space–air–ground integrated networks (SAGIN), seaborne shipping has attracted increasing interest in the research on the motion behavior knowledge extraction and navigation pattern mining problems in the era of maritime big data for improving maritime traffic safety management. This study aims to develop a novel unsupervised methodology for feature extraction and knowledge discovery based on automatic identification system (AIS) data, allowing for seamless knowledge transfer to support trajectory data mining. The unsupervised hierarchical methodology is constructed from three parts: trajectory compression, trajectory similarity measure, and trajectory clustering. In the first part, an adaptive Douglas–Peucker with speed (ADPS) algorithm is created to preserve critical features, obtain useful information, and simplify trajectory information. Then, dynamic time warping (DTW) is utilized to measure the similarity between trajectories as the critical indicator in trajectory clustering. Finally, the improved spectral clustering with mapping (ISCM) is presented to extract vessel traffic behavior characteristics and mine movement patterns for enhancing marine safety and situational awareness. Comprehensive experiments are conducted and implemented in the Chengshan Jiao Promontory in China to verify the feasibility and effectiveness of the novel methodology. Experimental results show that the proposed methodology can effectively compress the trajectories, determine the number of clusters in advance, guarantee the clustering accuracy, and extract useful navigation knowledge while significantly reducing the computational cost. The clustering results are further explored and follow the Gaussian mixture distribution, which can help provide new discriminant criteria for trajectory clustering.

A Markov Decision Process framework to incorporate network-level data in motion planning for connected and automated vehicles

Abstract: Autonomy and connectivity are expected to enhance safety and improve fuel efficiency in transportation systems. While connected vehicle-enabled technologies, such as coordinated cruise control, can improve vehicle motion planning by incorporating information beyond the line of sight of vehicles, their benefits are limited by the current short-sighted planning strategies that only utilize local information. In this paper, we propose a framework that devises vehicle trajectories by coupling a locally-optimal motion planner with a Markov decision process (MDP) model that can capture network-level information. Our proposed framework can guarantee safety while minimizing a trip’s generalized cost, which comprises of its fuel and time costs. To showcase the benefits of incorporating network-level data when devising vehicle trajectories, we conduct a comprehensive simulation study in three experimental settings, namely a circular track, a highway with on- and off-ramps, and a small urban network. The simulation results indicate that statistically significant efficiency can be obtained for the subject vehicle and its surrounding vehicles in different traffic states under all experimental settings. This paper serves as a proof-of-concept to showcase how connectivity and autonomy can be leveraged to incorporate network-level information into motion planning.

Anticipated Collision Time (ACT): A two-dimensional surrogate safety indicator for trajectory-based proactive safety assessment

Abstract: Surrogate Safety Measures (SSMs) are widely used to assess potential crash risk proactively. Notably, most of the existing safety indicators are fundamentally designed to capture the rear-end collision risk. However, in reality, the traffic dynamics involve the simultaneous interaction of multiple vehicles on a 2-dimensional (2D) surface, which results in a broad spectrum of collision patterns, such as head-on, side-swipe, rear-end, and angled collisions. This study proposes a novel 2D surrogate safety indicator called Anticipated Collision Time (ACT) to capture the risk pattern corresponding to all collision types using a single measure. We also devised a procedure to automatically detect conflict situations and extract ACT profiles as well as crash-type information from the trajectory data. The primary inputs to the ACT estimation are the shortest distance between the vehicles and the closing-in rate. This study also introduces another safety indicator derived from the ACT profile called Time of Evasive Action (TEA). TEA primarily captures the time at which a vehicle commences to respond, in terms of deceleration, when it encounters an unsafe situation. Such a measure helps to understand the response pattern of different vehicles/drivers to a potential collision. We also derived Time Exposed ACT (TE-ACT) and Time Integrated ACT (TI-ACT) from the ACT profile to capture crash exposure and severity. To highlight the potential of ACT, the Powered Two Wheeler (PTW) safety in an urban environment was analyzed with trajectory data collected from a busy urban midblock section. The results emphasize the capabilities of ACT and the other derived indicators to capture crash risk proactively. It is incontrovertible from the analyses that the ACT is opening up a new avenue for a comprehensive investigation of the safety of various transport facilities, irrespective of geometry and traffic scenario. • Developed a novel surrogate safety indicator called Anticipated Collision Time (ACT). • ACT captures the overall crash risk for any road, vehicle, or the traffic condition. • Another safety indicator, Time of Evasive Action (TEA) was extracted from the ACT profile. • TEA captures the driver’s evasive actions ahead of a potential conflict. • ACT and its derivatives can capture the overall crash risk.

Bayesian extreme value analysis of kinematic-based surrogate measure of safety to detect crash-prone conditions in connected vehicles environment: A driving simulator experiment

Abstract: Connected Vehicles (CVs) technology has provided large-scale driving database embedded in Basic Safety Messages (BSMs). This valuable data source can shed more light on tracking individual driving behaviors to detect crashes. This study delves into the possibility of detecting crashes using BSMs in controlled high-fidelity driving simulator experiments. To this end, two driving simulator scenarios were designed to simulate Run-off-Road (ROR), and Rear-End (RE) crashes. Twenty-four professional truck drivers were recruited to drive the scenarios. In each scenario, crash and non-crash cases were identified from vehicles’ trajectories, resulting in four study cases. Drivers’ behaviors were quantified by characterizing two Kinematic-based Surrogate Measures of Safety (K-SMoS), namely Absolute value of Derivative of Instantaneous Acceleration (ADInstAccel) and Absolute value of Derivative of Steering (ADSteering). Extreme defensive driving volatilities under crash and non-crash cases were modeled by extreme value analysis of K-SMoS and fitting their associated Generalized Extreme Value (GEV) distributions under Bayesian inference. Accordingly, for each K-SMoS, the crash detection was formulated as a binary classification between two K-SMoS GEV continuous distributions under crash and corresponding non-crash conditions. Qualitative uncertainty analysis of joint posterior density distributions of GEVs’ parameters revealed a higher uncertainty of extreme driving behaviors in crash conditions. Regardless, notable relative increases in the central tendency of extreme K-SMoS in crash compared to non-crash conditions were found, implying the possibility of crash detection by tracking extreme drivers’ behaviors using trajectory-level observations. This visual inference was affirmed by the result of binary classification of GEV distributions associated with K-SMoS. Depending on the crash type and K-SMoS, 71% to 81% accuracy in crash detection was obtained, where ADSteering outperformed ADInstAccel in terms of the discriminative ability. Besides, using sensitivity–specificity analysis, the optimal threshold of 1.24 (rad/s) and 1.31 (m/s3), respectively, for ADSteering and ADInstAccel, were identified to detect crashes. These findings can potentially enhance CVs' automation level in spatiotemporally identifying crash-prone conditions to disseminate distress notifications. Furthermore, the introduced methodology can be a complementary one to what has been followed in the crash detection domain.

Long-term 4D trajectory prediction using generative adversarial networks

Abstract: Four-dimensional trajectory prediction is one of the key technologies of air traffic management (ATM) and plays a considerably significant role in enhancing air traffic safety, accelerating air traffic flow and improving ATM efficiency. In this work, we propose a novel long-term 4D trajectory prediction model based on generative adversarial network (GAN). First, trajectory data is preprocessed. Then, three deep generation models for trajectory prediction are designed based on one-dimensional convolution neural network (Conv1D-GAN), two-dimensional convolution neural network (Conv2D-GAN), and long short-term memory neural network (LSTM-GAN). Finally, the models are trained and tested using historical 4D trajectory data from Beijing to Chengdu, China. The results demonstrate that the Conv1D-GAN is the most suitable generative adversarial network for long-term aircraft trajectory prediction.

A novel reinforced dynamic graph convolutional network model with data imputation for network-wide traffic flow prediction

Abstract: Traffic data missing issues due to unpredictable equipment failure, extreme weather, and other reasons have brought great challenges to traffic flow prediction modeling. In this paper, a novel reinforced dynamic graph convolutional network model is proposed to simultaneously conduct data imputation and network-wide traffic flow prediction. First, a multi-graph convolutional fusion network is proposed for data imputation by using the graph convolutional network to analyze the propagation law of traffic states between traffic flow detection stations in both time and space dimensions. Second, to enhance the robustness of network-wide traffic flow prediction, a dynamic graph learning method based on deep reinforcement learning is proposed to adaptively generate the graph adjacency matrix to represent the dynamic spatiotemporal dependencies between the stations. Finally, experimental results on two real-world traffic datasets show that the proposed method outperforms other baseline methods and can effectively extract the data missing features and spatiotemporal dependence features between the stations. The visualization results of the graph adjacency matrix indicate that the proposed method can effectively identify the influential traffic stations in the process of traffic flow prediction, and the extracted dependencies between the stations are interpretable. The proposed model has strong generalization in tackling network-wide traffic flow prediction tasks with different data missing rates and missing patterns, and can be extended to assist decision-makers in enhancing traffic management and mitigating traffic congestion.

Ego-efficient lane changes of connected and automated vehicles with impacts on traffic flow

Abstract: Connected and automated vehicles (CAVs) enabled by wireless communication and vehicle automation are believed to revolutionize the form and operation of road transport in the next decades. This paper addresses traffic flow effects of CAVs, and focuses on their lane-changing impacts on the mixed traffic flow of CAVs and human-driven vehicles (HVs). At present technical paths towards the development and deployment of CAVs are still uncertain. With CAV technologies getting matured, CAVs are supposed to provide rides of higher efficiency than HVs, beyond improved safety and comfort. In heavy traffic, this would only be achievable via agile and flexible lane changes of CAVs, because longitudinal acceleration would be unhelpful or even impossible in mixed traffic. Such lane changes are expected to be ego-efficient in that they serve solely CAVs’ interests without much considering surrounding vehicles, as long as safety constraints are not violated. As road resources are limited, the growth of the CAV population adopting such ego-efficient lane-changing strategies would probably lead to renowned “Tragedy of the Commons”. In this context, this paper considers three important prospective questions: A: How to determine an ego-efficient lane-changing strategy for CAVs? B: With increasingly more CAVs introduced each adopting the ego-efficient lane-changing strategy, what is the impact on traffic flow? C: How to determine a system-efficient lane-changing strategy for CAVs? These forward-looking issues are addressed from the perspectives of microscopic traffic simulation and reinforcement learning. Without any constraint on the lane-changing incentive, the developed lane-changing strategy was found to be beneficial for CAVs and the entire traffic flow only if the market penetration rate (MPR) of CAVs is less than 50%. With an appropriate constraint placed, however, the lane-changing strategy was found to become consistently beneficial for the entire traffic flow at any MPR. These findings suggest that CAVs may not simply be a magic cure for traffic problems that the society is currently facing, unless some upper-level coordination may be proposed for CAVs to benefit not only themselves but also the entire traffic. This is also consistent with what “Tragedy of the Commons” suggests.

Integrating public transit signal priority into max-pressure signal control: Methodology and simulation study on a downtown network

Abstract: Max-pressure signal control has been analytically proven to maximize the network throughput and stabilize queue lengths whenever possible. Since there are many transit lines operating in the metropolis, the max-pressure signal control should be extended to multi-modal transportation systems to achieve more widespread usage. The standard max-pressure controller is more likely to actuate phases during high-demand approaches, which may end up ignoring the arrival of buses, especially in bus rapid transit. In this paper, we propose a novel max-pressure signal control that considers transit signal priority of bus rapid transit systems to achieve both maximum stability for private vehicles and reliable transit service. This study revises the original max-pressure control to include constraints that provide priority for buses. Furthermore, this policy is decentralized which means it only relies on it relies only on the local conditions of each intersection. We set the simulation on the real-world road network with bus rapid transit systems. Numerical results show that the max-pressure signal control which considers transit signal priority can still achieve maximum stability compared with other signal control integrated with transit signal priority. Furthermore, the max-pressure control reduces private vehicle travel time and bus travel time compared to the current signal control. • This study combines the max-pressure control with transit signal priority for the first time. • This study designs dynamic queueing models for bus rapid transit systems and private vehicles. • Rigorous proof of maximum stability by implementing proposed signal control method is provided. • The proposed signal control method is still a decentralized method, which relies on the local conditions of each intersection. • Simulations implemented in realistic network highlight the advantage of the proposed method.

GPS data-based mobility mode inference model using long-term recurrent convolutional networks

Abstract: • Long-term Recurrent Convolutional Networks is proposed to infer mobility modes. • Using only GPS data, motion characteristics of each mode are captured. • Advantage of LRCN for sequential behavior of mobility mode selection is identified. Recently, identifying and estimating distributions of mobility modes have become essential tasks for establishing traffic management strategies. The primary purpose of this study is to develop a mobility mode inference model able to consider the sequential behaviors of choosing a mobility mode by using long-term recurrent convolutional networks (LRCNs). Only GPS profile data are used for the mobility mode classification. The modes are categorized into walk, bike, motorcycle, bus, driving, train, and Segway. The proposed LRCN architecture applies the sequencing mode transition concept, a novel concept in mobility mode inference research. We conducted a data preprocessing procedure to normalize the input data size, capture the motion behaviors from the GPS points, and refine the data. We then identified an optimal convolutional neural network (CNN) model by considering the number of layers, layer-order pattern, and number of filters. The CNN model was established by applying an ensemble CNN concept to a single optimal CNN model. Furthermore, we integrated the optimal CNN model and a long short-term memory (LSTM) network as an LRCN, so as to consider the sequential behaviors in choosing the mobility mode over time. Consequently, we established an optimal LRCN–bi model with the highest performance among the existing LRCN architectures. By comparing the confusion matrices of each of the two best models in the CNN and LRCN approaches, we confirmed that considering the sequential behaviors in choosing the mobility mode enhances the model’s performance in inferring the mobility mode. Furtherm ore, we confirmed that the LRCN approach outperforms approaches from previous studies.

AdapGL: An adaptive graph learning algorithm for traffic prediction based on spatiotemporal neural networks

Abstract: With well-defined graphs, graph convolution based spatiotemporal neural networks for traffic prediction have achieved great performance in numerous tasks. Compared to other methods, the networks can exploit the latent spatial dependencies between nodes according to the adjacency relationship. However, as the topological structure of the real road network tends to be intricate, it is difficult to accurately quantify the correlations between nodes in advance. In this paper, we propose a graph convolutional network based adaptive graph learning algorithm (AdapGL) to acquire the complex dependencies. First, by developing a novel graph learning module, more possible correlations between nodes can be adaptively captured during training. Second, inspired by the expectation maximization (EM) algorithm, the parameters of the prediction network module and the graph learning module are optimized by alternate training. An elaborate loss function is leveraged for graph learning to ensure the sparsity of the generated affinity matrix. In this way, the expectation maximization of one part can be realized under the condition that the other part is the best estimate. Finally, the graph structure is updated by a weighted sum approach. The proposed algorithm can be applied to most graph convolution based networks for traffic forecast. Experimental results demonstrated that our method can not only further improve the accuracy of traffic prediction, but also effectively exploit the hidden correlations of the nodes. The source code is available at https://github.com/goaheand/AdapGL-pytorch.

Traffic congestion propagation inference using dynamic Bayesian graph convolution network

Abstract: Congestion, whether recurrent or non-recurrent, propagates through the road network. The process of congestion propagation from a particular road to its neighbors can be regarded as a kind of message passing with a directed relationship. Existing methods have created a solid foundation for characterizing congestion propagation; however, they are either built upon simplified assumptions in traffic flow theory or predefined relationships among road sections, which would lead to downgraded accuracy in practice. This paper proposes a dynamic Bayesian graph convolutional network (DBGCN), which integrates Bayesian inference into a deep learning framework. Therefore, the rules of congestion propagation in the network can be actively learned from the observed data instead of predefining them based on prior knowledge. Experimental results on 971 testbeds in a regional road network in Beijing demonstrate that DBGCN outperforms the state-of-the-art models in inferring the congestion propagation spatiotemporal coverage and reveals variations in congestion propagation patterns according to the road network structure. Furthermore, the proposed model can simulate the congestion propagation process in customized scenarios by learning the latent congestion propagation rules. The results in different scenarios show that the change of congestion source location leads to distinct congestion magnitude, and the propagation of congestion will eventually stop at the road sections with strong shunting effect.

Online parking assignment in an environment of partially connected vehicles: A multi-agent deep reinforcement learning approach

Abstract: The advent of connected vehicles (CVs) provides new opportunities to address urban parking issues due to the widespread application of online parking assignment (OPA) services. However, before CVs fully replace non-connected vehicles (NCVs), it is envisioned that CVs and NCVs will coexist for a long time. This brings challenges for OPA because of the availability constraints imposed by the uncertain arrivals and departures of NCVs. This paper proposes a multi-agent deep reinforcement learning framework to generate efficient OPA strategies with partial observations of parking demand. Specifically, we create two agents, one for measuring the impact of NCVs and the other for exploring the parking characteristics of CVs. A value decomposition method is adopted to solve the multi-agent learning problem, and a modified exploration strategy is designed to direct agent training and avoid unnecessary trials. To verify the performance of the proposed approach, we derive the baselines of the total time expenditure in a parking area based on the widely adopted first-come-first-served strategy and a hypothetical system optimum strategy, respectively. Also, we present a dynamic assignment model with forecasting as a comparison of the proposed approach with the same demand information. Two typical parking scenarios are selected to conduct comparative experiments with actual operating data. The experimental results show that the proposed learning-based approach can effectively allocate parking resources. Provided with user parking information of CVs short in advance, our approach can achieve up to 15% improvement in assignment performance compared with other baselines.

E-scooter sharing and bikesharing systems: An individual-level analysis of factors affecting first-use and use frequency

Abstract: Shared micromobility modes have increasingly penetrated the mobility environment of cities in the U.S. and the world over. At the same time, to best integrate these emerging modes within the fabric of the existing (and larger) transportation ecosystem, it is critical to understand how individuals may respond and “who” the likely users of these relatively new modes may be. In this paper, we develop a model to analyze first-use and use frequency of two micromobility modes: E-scooter sharing systems (ESS) and Bike sharing systems (BSS). The model employs psycho-social constructs, built environment attributes, as well as individual-level demographics as determinants. In doing so, we explicitly recognize the role played by awareness/first-use of new technologies as a cognitive antecedent to subsequent frequency decisions. The main data source for this analysis is drawn from a 2019 survey of Austin, Texas area residents. Our results highlight the importance of considering psycho-social attitudes to both gain better insights into the behavioral process leading up to ESS/BSS adoption/use as well as ensure an accurate data fit. In particular, there are distinctive pathways of adoption/use frequency for each of the ESS and BSS modes, but also complementary processes and behavioral spillover effects at play that warrant a joint modeling of the ESS and BSS modes. Our results suggest that addressing safety concerns of micromobility modes should be the top priority of providers and public agencies. Efforts solely directed toward extoling the “green” virtues of micromobility modes is likely to have limited returns.