Next Generation Simulation Research Articles

A data-driven traffic shockwave speed detection approach based on vehicle trajectories data

Traffic shockwaves demonstrate the formation and spreading of traffic fluctuation on roads. Existing methods mainly detect the shockwaves and their propagation by estimating traffic density and flow, which presents weaknesses in applications when traffic data is only partially or locally collected. This paper proposed a four-step data-driven approach that integrates machine learning with the traffic features to detect shockwaves and estimate their propagation speeds only using partial vehicle trajectory data. Specifically, we first denoise the speed data derived from trajectory data by the Fast Fourier Transform (FFT) to mitigate the effect of spontaneous random speed fluctuation. Next, we identify trajectory curves’ turning points where a vehicle runs into a shockwave and its speed presents a high standard deviation within a short interval. Furthermore, the Density-based Spatial Clustering of Applications with Noise algorithm (DBSCAN) combined with traffic flow features is adopted to split the turning points into different clusters, each corresponding to a shockwave with constant speed. Last, the one-norm distance regression method is used to estimate the propagation speed of detected shockwaves. The proposed framework was applied to the field data collected from the I-80 and US-101 freeway by the Next Generation Simulation (NGSIM) program. The results show that this four-step data-driven method could efficiently detect the shockwaves and their propagation speeds without estimating the traffic densities and flows nearby. It performs well for both homogenous and nonhomogeneous road segments with trajectory data collected from total or partial traffic flow.

Virtual reality simulation as a training tool for perfusionists in extracorporeal circulation: Establishing face and content validity.

We conducted a prospective study to assess the face and content validity of a new virtual reality (VR) extracorporeal circulation simulator (ECC) developed for perfusionists to facilitate training and practice. We evaluated the opinions of students and staff members about the feasibility of the simulation. The 2 groups consisted of experts (qualified perfusionists) and novices (trainee perfusionists). Perfusionists (n=12 experts and n=11 trainees) received instructions on how to use the VR simulator and then proceeded to perform the start of cardiopulmonary bypass in the VR environment. Participants then completed a Usefulness, Satisfaction, and Ease of Use Questionnaire. The questions were rated on a 5-point Likert scale, ranging from 1 (fully disagree) to 5 (fully agree), to assess the face validity and content validity of this simulator. Participants reported a predominantly positive experience with the VR-ECC simulator, with 96% (n=22) agreeing that the simulator was a useful way of training ECC scenarios. All participants found it easy to interact with the software (100%, n=23), and 82% of students (n=9) believed it helped them remember the steps involved with initiating ECC. Finally, (87% [n=20]) of participants believed the image quality and depth perception were good. Our next-generation simulator was valid for face and content constructs, and almost all participants found it to be a useful way of training for ECC scenarios. This simulator represents a first step toward truly blended digital learning and a new interactive, flexible, and innovative modality for perfusion training.

Spatio-temporal interactive graph convolution network for vehicle trajectory prediction

Vehicle trajectory prediction is crucial in achieving safe and reliable autonomous driving decision-making. The accuracy of the prediction is affected by many different factors, such as the integrity and efficiency of vehicle-to-vehicle (V2V) data transfer, the complex environmental factors of the surrounding roads and the perception range of vehicle sensors. However, most existing methods cannot capture the dynamic interactive information of vehicles at different time steps. In this paper, we propose a Spatio-Temporal Interactive Graph Convolutional Network (STI-GCN) that predicts future trajectories by acquiring spatiotemporal features of vehicles. In the spatial dimension, we construct a kernel function based on spatial autocorrelation and use it as prior knowledge to describe the degree of mutual influence between vehicles in real traffic scenarios. In addition, the Gated Recurrent Unit (GRU) is used to dynamically capture the spatial features of vehicles to solve the dynamic making graph problem of vehicles in the real traffic scene. In the temporal dimension, we use a Convolutional Neural Network (CNN) to extract the temporal feature in the historical trajectory of the vehicle. Finally, we experimented with our method and existing methods on the public Next Generation Simulation (NGSIM) dataset. The experimental results show that the error of our model is reduced by about 10% compared with the state-of-the-art model. And it also improves about 10 times in two key metrics, namely the number of parameters and inference time.

Anti-Disturbance Self-Supervised Reinforcement Learning for Perturbed Car-Following System

This paper proposes an anti-disturbance car-following strategy for attenuating (i) exogenous disturbances from preceding traffic oscillations and (ii) endogenous disturbances in vehicular control systems (e.g., wind gust, ground friction, and rolling resistance). Firstly, it employs a modified robust controller to generate an expert car-following control experience. Subsequently, it imitates the expert behaviors via the behavioral cloning (BC) technique, thereby developing the anti-disturbance ability. Lastly, the obtained policy is optimized using the self-supervised reinforcement learning (RL) approach. The simulation experiments, comprising both training and evaluation phases, are performed via Python. To simulate car-following scenarios, we utilize the ground-truth data from the Next Generation Simulation (NGSIM) datasets. Through recursive interactions with the perturbed car-following environment, self-supervised RL drives stable policy improvement. The proposed anti-disturbance self-supervised RL (ADSSRL) policy presents a smooth and almost monotonously increasing reward curve. Further evaluation of disturbance dampening performance suggests that at least a 44.5% reduction in control efficiency cost and a 10.1% reduction in driving comfort cost are achieved compared with baselines.

A physics-informed Transformer model for vehicle trajectory prediction on highways

Autonomous Vehicles (AVs) have made remarkable developments and are anticipated to replace human drivers. In transitioning from human-driven vehicles to fully AVs, one crucial task is to predict the trajectories of the subject vehicle and its surrounding vehicles in real time. Most existing methods of vehicle trajectory prediction on highways are based on physical models or purely data-driven models. However, they either yield unsatisfactory prediction performance or lack model interpretability and physical implications. This paper proposes a Physics-Informed Deep Learning framework that fully leverages the advantages of data-driven and physics-based models to go beyond the existing models. We use the Transformer neural network architecture with self-attention as Physics-Uninformed Neural Network (PUNN) and Intelligent Driver Model (IDM) as physical model to construct of Physics-Informed Transformer-Intelligent Driver Model (PIT-IDM). Extensive experiments have been conducted on two datasets with different traffic environments, i.e., Next Generation SIMulation (NGSIM) data in the US and the Ubiquitous Traffic Eyes (UTE) data in China, to verify model accuracy and efficiency. Compared with the three kinds of baselines by relative and absolute measures of effectiveness, the best performing PIT-IDM reduces longitudinal trajectory prediction errors for long horizons by 5%-50%, some even reduced up to 70%. Extensive empirical analyses have been carried out to verify its excellent spatio-temporal transferability and explore the physics-informed mechanism underlying this deep learning method. The training and inference time analysis indicates that although it takes longer to train PIT-IDM, it requires fewer calls and accumulates fewer errors with less computation time in real-world applications. The overall results further validate the efficacy of this Physics-Informed Deep Learning framework in enhancing model accuracy, interpretability, and transferability.

Roadside LiDAR Sensor Configuration Assessment and Optimization Methods for Vehicle Detection and Tracking in Connected and Automated Vehicle Applications

This paper develops an assessment and optimization model for configuring roadside LiDAR (Line Detection and Ranging) installation. More specifically, an analytic and a simulation model have been developed to analyze the detection blind zones and their impact on vehicle detection and tracking capabilities in Connected and Automated Vehicle (CAV) applications. The proposed model can derive the area and height of the detection blind zones from a given roadside LiDAR location and road geometry. Evaluation metrics are also proposed to assess the severity of the blind zone including laser beam density, blind zone height and duration, vehicle trajectory missing rate and duration. The simulation model can be used to evaluate and identify optimal configurations for different installation scenarios. To validate the proposed model, the 15-min US101 NGSIM (Next Generation SIMulation) dataset was used for validating the proposed model. Different configuration settings were simulated and compared. The evaluation results demonstrate the capabilities of the proposed models in planning for optimal roadside LiDAR sensor installation for vehicle detection and tracking.

Fault Tolerance analysis of an Adaptive Neuro-Fuzzy Inference System for mandatory lane changing decisions in automated driving

Past research has developed a binary decision model for mandatory lane changes based on the Adaptive Neuro-Fuzzy Inference System (ANFIS). This ANFIS Decision Model (simply called ADEM), developed and tested with the Next Generation Simulation (NGSIM) data, mimics the sensory inputs and decisions of human drivers. This research assumed that ADEM will be implemented as part of the automated lane changing system in Automated Vehicles (AVs). The system in AVs will depend on active radar sensors to make measurements. The sensor outputs will be converted into the input parameter values of ADEM. This research tested ADEM’s performance when the sensors could only measure the distance of surrounding vehicles within 50 m , and when one of the sensors malfunctions. The original NGSIM test data set was modified to simulate the sensors’ detection range limit in Scenario 0, plus six other scenarios in which each sensor took turns to fail and assumed either the minimum or maximum possible output values. The results show that: (i) ADEM performs in a safer manner when considering the sensors’ limited detection range; (ii) the minimum value of 0 m should be used as the default sensor output when a sensor fails, so that ADEM makes safer mandatory lane changing decisions; and (iii) the most critical sensors, by which failure of any of them would cause the greatest degradation to ADEM’s performance, are the two sensors that measure the distances to the preceding and the following vehicles in the target lane.

Anisotropy safety potential field model under intelligent and connected vehicle environment and its application in car-following modeling

Potential field theory, as a theory that can also be applied to vehicle control, is an emerging risk quantification approach to accommodate the connected and self-driving vehicle environment. Vehicles have different risk impact effects on other road participants in each direction under the influence of road rules. This variability exhibited by vehicles in each direction is not considered in the previous potential field model. Therefore, this paper proposed a potential field model that takes the anisotropy of vehicle impact into account: (1) introducing equivalent distances to separate the potential field area in the different directions before and after the vehicle; (2) introducing co-virtual forces to characterize the effect of the side-by-side travel phenomenon on vehicle car-following travel; (3) introducing target forces and lane resistance, which regress the control of desired speed to control the acceptable risk of drivers. The Next Generation Simulation (NGSIM) dataset is used in this study to create the model's initial parameter values based on the artificial swarm algorithm. The simulation findings indicate that when the vehicle is given the capacity to perceive the surrounding traffic environment, the suggested the anisotropic safety potential field model (ASPFM) performs better in terms of driving safety.

Uncertainty Estimation of Connected Vehicle Penetration Rate

Knowledge of the connected vehicle (CV) penetration rate is crucial for realizing numerous beneficial applications during the prolonged transition period to full CV deployment. A recent study described a novel single-source data penetration rate estimator (SSDPRE) for estimating the CV penetration rate solely from CV data. However, despite the unbiasedness of the SSDPRE, it is only a point estimator. Consequently, given the typically nonlinear nature of transportation systems, model estimations or system optimizations conducted with the SSDPRE without considering its variability can generate biased models or suboptimal solutions. Thus, this study proposes a probabilistic penetration rate model for estimating the variability of the results generated by the SSDPRE. An essential input for this model is the constrained queue length distribution, which is the distribution of the number of stopping vehicles in a signal cycle. An exact probabilistic dissipation time model and a simplified constant dissipation time model are developed for estimating this distribution. In addition, to improve the estimation accuracy in real-world situations, the braking and start-up motions of vehicles are considered by constructing a constant time loss model for use in calibrating the dissipation time models. VISSIM simulation demonstrates that the calibrated models accurately describe constrained queue length distributions and estimate the variability of the results generated by the SSDPRE. Furthermore, applications of the calibrated models to the next-generation simulation data set and a simple CV-based adaptive signal control scheme demonstrate the readiness of the models for use in real-world situations and the potential of the models to improve system optimizations. Funding: This work was supported by The University of Hong Kong [Francis S Y Bong Professorship in Engineering and Postgraduate Scholarship] and by the Council of the Hong Kong Special Administrative Region, China [Grants 17204919 and 17205822]. Supplemental Material: The online appendices are available at https://doi.org/10.1287/trsc.2023.1209 .

Naturalistic Driving Data-Based Anomalous Driving Behavior Detection Using Hypertuned Deep Autoencoders

Autonomous driving is predicted to play a large part in future transportation systems, providing benefits such as enhanced road usage and mobility schemes. However, self-driving cars must be perceived as safe drivers by other road users and contribute to traffic safety in addition to being operationally safe. Despite efforts to develop machine learning algorithms and solutions for the safety of automated vehicles, researchers have yet to agree upon a single approach to categorizing and accurately detecting safe and unsafe driving behaviors. This paper proposes a modified Z-score method-based autoencoder for anomalous behavior detection using multiple driving indicators. The experiments are performed on the benchmark Next Generation Simulation (NGSIM) vehicle trajectories and supporting datasets to discover anomalous driving behavior to assess our proposed approach’s performance. The experiments reveal that the proposed approach detected 81 anomalous driving behaviors out of 1031 naturalistic driving behavior instances (7.86%) with an accuracy of 96.31% without early stopping. With early stopping, our method successfully detected 147 anomalous driving behaviors (14.26%) with an accuracy of 95.25%. Overall, the proposed approach provides promising results for detecting anomalous driving behavior in automated vehicles using multiple driving indicators.

Modeling Autonomous Vehicles’ Altruistic Behavior to Human-Driven Vehicles in the Car following Events and Impact Analysis

To explore the impact of autonomous vehicles (AVs) on human-driven vehicles (HDVs), a solution for AV to coexist harmoniously with HDV during the car following period when AVs are in low market penetration rate (MPR) was provided. An extension car following framework with two possible soft optimization targets was proposed in this article to improve the experience of HDV followers with different following strategies by deep deterministic policy gradient (DDPG) algorithm. The pretreated Next Generation Simulation (NGSIM) dataset was used for the experiments. 1027 car following events with being redefined were extracted from it, in which 600 of the events were used for training and 427 of the events were used for testing. The different driving strategies obtained from the classical car following models were embedded into virtual environment built by OpenAI gym. The reward function combined safety, efficiency, jerk, and stability was used to encourage the agent with DDPG algorithm to maximize it. The final result reveals that disturbance of HDV followers decreases by 2.362% (strategy a), 8.184% (strategy b), and 13.904% (strategy c), respectively. The disturbance of HDV follower decreases by 14.961% (strategy a), 12.020% (strategy b), and 13.425% (strategy c), respectively. HDV followers with different strategies get less jerk in both soft optimizations. AV passengers get a loss on jerk and efficiency, but safety is enhanced. Also, AV car following performs better than HDV car following in both soft and brutal optimizations. Moreover, two possible solutions for harmonious coexistence of HDVs and AVs when AVs are in low MPR are proposed.

VRR-Net: Learning Vehicle–Road Relationships for Vehicle Trajectory Prediction on Highways

Vehicle trajectory prediction is an important decision-making and planning basis for autonomous driving systems that enables them to drive safely and efficiently. To accurately predict vehicle trajectories, the complex representations and dynamic interactions among the elements in a traffic scene are abstracted and modelled. This paper presents vehicle–road relationships net, a deep learning network that extracts features from vehicle–road relationships and models the effects of traffic environments on vehicles. The introduction of geographic highway information and the calculation of spatiotemporal distances with a reference not only unify heterogeneous vehicle–road relationship representations into a time series vector but also reduce the requirement for sensing transient changes in the surrounding area. A hierarchical long short-term memory network extracts environmental features from two perspectives—social interactions and road constraints—and predicts the future trajectories of vehicles by their manoeuvre categories. Accordingly, vehicle–road relationships net fully exploits the contributions of historical trajectories and integrates the effects of road constraints to achieve performance that is comparable to or better than that of state-of-the-art methods on the next-generation simulation dataset.

Impact of Autonomous Vehicles on the Car-Following Behavior of Human Drivers

The past few years have been witness to an increase in autonomous vehicle (AV) development and testing. However, even with a fully developed and comprehensively tested AV technology, AVs are anticipated to share the roadway network with human drivers for the unforeseeable future. In such a mixed environment, we use naturalistic driving data from the next generation simulation (NGSIM) and Lyft Level 5 (Lyft L5) prediction data sets to investigate whether the existence of AVs influences the car-following behavior of human drivers. We use time headway time series as a proxy to capture the car-following behavior of human drivers. We then develop a nested fixed model to find possible changes in behavior when human drivers are following different types of vehicles (i.e., human-driven vehicles or AVs). The factors included in this model are the platoon structure (a legacy vehicle following a legacy vehicle, and a legacy vehicle following an autonomous vehicle), road type (freeway and urban), time period (morning and afternoon), lane (right, middle, and left), and the source of the data (NGSIM and Lyft L5). Results indicate a statistically significant difference between the car-following behavior of drivers when they follow a human-driven vehicle compared to an AV. This change in the car-following behavior of drivers is manifested in the form of a reduction in the mean and variance of time headways when human drivers follow an AV. These findings can bridge the gap between anticipated and real-world impacts of AVs on traffic streams and roadway capacity, providing meaningful insights on the influence of AVs on the driving behavior of humans in a naturalistic driving environment.

An Evolutionary Learning Framework of Lane-Changing Control for Autonomous Vehicles at Freeway Off-Ramps

This paper proposes a lateral control strategy for autonomous vehicles (AVs) and develops an evolutionary learning framework for off-ramps. Random forest (RF) and back-propagation neural network (BPNN) integrated with model predictive control (MPC) algorithm are respectively used to capture the decision-making and trajectory characteristics during the lane-changing maneuver based on the Next Generation Simulation (NGSIM) dataset. Then, a running cost function is calculated to optimize the trajectory dataset. Finally, the numerical simulation is conducted to investigate the characteristics of the proposed framework. Simulation results indicate that the performance of our method is much better than some other methods in lane-changing gap choice and trajectory execution. Moreover, the traffic system controlled by the evolutionary algorithms reaches the highest capacity and safest level when all vehicles are equipped with cooperative adaptive cruise control (CACC) systems. On the contrary, the scenario with 50% CACC vehicles shows the lowest travel efficiency and the worst safety because of the CACC vehicles’ degradation. Furthermore, three iterations and 500 vehicle trajectories at each optimization cycle are recommended for the application in the off-ramp traffic control.

Improved Car-Following Model for Connected Vehicles Considering Backward-Looking Effect and Motion Information of Multiple Vehicles

To alleviate traffic congestion, an improved car-following model for connected vehicles is proposed in this paper by considering backward-looking effect and motion information of multiple vehicles. The backward-looking effect is common in traffic and motion information of multiple vehicles and is beneficial for alleviating traffic congestion. The linear and nonlinear stability of the proposed model is analyzed, and the stability condition and modified Korteweg-de Vries equations of the proposed model are derived. The theoretical analysis results prove the effectiveness of the proposed model in alleviating traffic congestion and improving the stability of the traffic system. Further, numerical simulation is designed to verify the promoting effect of the introduced parameters on the traffic stability and to test the effect of the improved model on large-scale car-following queues. Finally, the next-generation simulation (NGSIM) data sets are used to calibrate the parameters of the improved model. The results show that the improved model can effectively avoid traffic congestion and enhance the stability of traffic flow. The improved model can be used as active safety technology to prevent collision accidents or as a car-following strategy in driverless algorithms.

A Novel Lane-Changing Recognition Method Using Frequency Analysis

Lane-changing recognition is an important task for advanced driver assistance systems, but is heavily challenged by poor driving habits, such as turning without using turn signals. To address this problem in this study, a lane-changing recognition method using frequency analysis was proposed. First, highest-frequency–based and frequency-bands–based methods were employed to evaluate the three behaviors of left lane changing, lane keeping, and right lane changing. To improve the recognition accuracy, the two methods were fused according to their classification advantages for different behaviors. The fused method was verified by lateral position data incorporating lane features that were manually extracted and annotated from the Next-Generation Simulation dataset. The frequency analysis framework achieved recognition accuracy of 91.8%, 97.4%, and 99.1% in 2, 1, and 0 s, respectively, before the vehicle crossed the lane line, which were significant improvements over the time-domain analysis methods. The proposed method was also validated by real-world road data with promising results.

Robust Data Sampling in Machine Learning: A Game-Theoretic Framework for Training and Validation Data Selection

How to sample training/validation data is an important question for machine learning models, especially when the dataset is heterogeneous and skewed. In this paper, we propose a data sampling method that robustly selects training/validation data. We formulate the training/validation data sampling process as a two-player game: a trainer aims to sample training data so as to minimize the test error, while a validator adversarially samples validation data that can increase the test error. Robust sampling is achieved at the game equilibrium. To accelerate the searching process, we adopt reinforcement learning aided Monte Carlo trees search (MCTS). We apply our method to a car-following modeling problem, a complicated scenario with heterogeneous and random human driving behavior. Real-world data, the Next Generation SIMulation (NGSIM), is used to validate this method, and experiment results demonstrate the sampling robustness and thereby the model out-of-sample performance.

Iterative learning control for lane-changing trajectories upstream off-ramp bottlenecks and safety evaluation

This paper proposes an iterative learning control framework for lane changing to improve traffic operation and safety at a diverging area nearby a highway off-ramp in an environment with connected and automated vehicles (CAVs). This framework controls CAVs in the off-ramp bottlenecks by imitating the trajectories optimized by machine learning algorithms. Next Generation Simulation (NGSIM) dataset is utilized as the raw data and filtered by cost function. The traffic models, including lane-changing decision (LCD) models and lane-changing execution (LCE) models, are completed by Random Forest (RF) and Back Propagation Neural Network (BPNN) algorithms. Based on simulation results, simulation data satisfying the predetermined criterion will be added to dataset in the next iteration. Various metrics are considered to evaluate the proposed framework systematically from both lateral and longitudinal aspects, including time exposed time-to collision (TET), time integrated time-to-collision (TIT), rear-end collision risk indexes (RCRI) and lane-changing risk index (LCRI). The results present that the iterative framework can decrease the longitudinal risk of the system by a factor of two times, and can reduce the lateral risk by 28.7%. When the CAVs market penetration rate (MPR) reaches 100%, the longitudinal and lateral risk values of the off-ramp system can be reduced by 90% and 35%, respectively. However, it is worth noting that only when the CAVs MPR reaches 50% does the system's value at risk change significantly.

Sparse least squares support vector machine based methods for vehicle driving behavior recognition

Intelligent vehicles are expected to accurately recognize the driving intentions of surrounding vehicles so as to precisely identify the hazards to automatic driving and accomplish reasonable motion planning. This paper introduces a model for vehicles driving behavior recognition (VDBR) based on Sparse Least Squares Support Vector Machine (S-LSSVM) by means of machine learning methods, with the subject vehicle and its surrounding vehicles as the research subject. First, the relative lateral displacement and relative lateral speed between vehicles are captured as the eigenvectors after calculation of trajectory curvature and change time window. Then, the pruning algorithm is used to make Least Squares Support Vector Machine (LSSVM) training samples sparse and the Particle Swarm Optimization algorithm (PSO) is employed to accomplish parameter tuning of S-LSSVM. Thus, a modified S-LSSVM model is constructed to grasp the interaction behavior between vehicles. The experiment results demonstrate that the S-LSSVM based model obtain better accuracy and timeliness compared with SVM and LSSVM on the Next Generation Simulation (NGSIM) dataset and the data from autonomous driving experimental platform.

An Improved Cellular Automata Traffic Flow Model Considering Driving Styles

An improved cellular automata model (CA model) considering driving styles is proposed to analyze traffic flow characteristics and study traffic congestion’s dissipation mechanism. The data were taken from a particular case in the Next Generation Simulation (NGSIM) program, which selected US-101 as the survey location from 7:50 a.m.–8:05 a.m. to investigate vehicle trajectory information. Different driving styles and the differences in vehicle parameters (speed, acceleration, deceleration, etc.) were obtained using principal component analysis and the k-means clustering method. The selected model was proposed for improvement based on analyzing the existing CA models and combining them with the actual road conditions. Considerations of driving styles and two operation mechanisms (over-acceleration and speed adaptation) were introduced in the improved model. The result obtained after the traffic simulation shows that the improved CA model is effective, and the mutual transformation of different traffic flow phases can be simulated. In the improved CA model, dissipating traffic congestion effectively and balancing the overall flow of the road are realized to improve the traffic capacity up to around 115% compared to the NaSch model and meet the demand of all kinds of drivers expecting to drive at the safest distance, which provides a theoretical basis for relieving traffic congestion. The various driving styles in terms of safety, comfort, and effectiveness are performed differently in the improved CA model. An aggressive driving style contributes to increasing traffic capacity up to around 181% compared to a calm driving style, while the calm style contributes to maintaining traffic flow stability.