Bimodal Extended Kalman Filter-Based Pedestrian Trajectory Prediction.

We propose a pedestrian trajectory prediction algorithm based on the bimodal extended Kalman filter. With this filter, we are able to make full use of the dual-state nature of the pedestrian movement, i.e., the pedestrian is either moving or remains stationary. We apply the dual-mode probability model to describe the state of the pedestrian. Based on this model, we construct the proposed bimodal extended Kalman filter to estimate pedestrian state distribution. The filter obtains the state distribution for each pedestrian in the scene, respectively, and use that state distribution to predict the future trajectories of all the people in the scene. This prediction method estimates the prior probability of each parameter of the model through the dataset and updates the individual posterior probability of the pedestrian state through the bimodal extended Kalman filter. Our model can predict the trajectory of every individual, by taking the social interaction of pedestrians as well as the surrounding physical obstacles into account, with less than fifty model parameters being used, while with the limited parameter, our model could be nearly accurate as other deep learning models and still be comprehensible for model users.

Comparison of Eye and Face Features on Drowsiness Analysis.

Drowsiness is one of the leading causes of traffic accidents. For those who operate large machinery or motor vehicles, incidents due to lack of sleep can cause property damage and sometimes lead to grave consequences of injuries and fatality. This study aims to design learning models to recognize drowsiness through human facial features. In addition, this work analyzes the attentions of individual neurons in the learning model to understand how neural networks interpret drowsiness. For this analysis, gradient-weighted class activation mapping (Grad-CAM) is implemented in the neural networks to display the attention of neurons. The eye and face images are processed separately to the model for the training process. The results initially show that better results can be obtained by delivering eye images alone. The effect of Grad-CAM is also more reasonable using eye images alone. Furthermore, this work proposed a feature analysis method, K-nearest neighbors Sigma (KNN-Sigma), to estimate the homogeneous concentration and heterogeneous separation of the extracted features. In the end, we found that the fusion of face and eye signals gave the best results for recognition accuracy and KNN-sigma. The area under the curve (AUC) of using face, eye, and fusion images are 0.814, 0.897, and 0.935, respectively.

An efficient framework of developing video-based driving simulation for traffic sign evaluation.

INTRODUCTION: The driving simulator is a widely adopted experimental platform for investigating human-factors questions related to traffic signs and other traffic control devices in a safe environment. This paper presents a methodological framework for developing a video-based simulation program for traffic-sign evaluation. METHOD: We firstly collected video data and vehicle movement data from on-road driving. Secondly, the signs on the collected video footage were detected and tracked automatically using image processing techniques. Images of newly designed signs were integrated onto the video footage and placed onto the real-world sign locations. The inserted image properties were fused to fit into the video background to yield a natural visual effect. Thirdly, the vehicle-movement data collected during the drive-through were incorporated into the video sequence as well as the motion of the driving simulator. Using throttle and brake pedals of the driving simulator, participants drove through the video sequence with control over the video's playback speed and the simulator's movement to achieve a comparable visualization and motion experience as real-world driving. Results Conclusions: This framework was used to investigate drivers' visual attention and understanding of various newly proposed changeable message signs (CMSs). The results prove that this framework effectively engaged drivers in the driving task in the realistic traffic scene and successfully evaluated drivers' perception and understanding of the traffic signs. PRACTICAL APPLICATIONS: With this methodological framework, a driving simulation program based on real-world video data from specified road environment and vehicle-movement information can be quickly established and used for testing a variety of traffic control devices, especially traffic signs, in the study of human-machine interaction.

Prediction of Pedestrian Crossing Behavior Based on Surveillance Video.

Prediction of pedestrian crossing behavior is an important issue faced by the realization of autonomous driving. The current research on pedestrian crossing behavior prediction is mainly based on vehicle camera. However, the sight line of vehicle camera may be blocked by other vehicles or the road environment, making it difficult to obtain key information in the scene. Pedestrian crossing behavior prediction based on surveillance video can be used in key road sections or accident-prone areas to provide supplementary information for vehicle decision-making, thereby reducing the risk of accidents. To this end, we propose a pedestrian crossing behavior prediction network for surveillance video. The network integrates pedestrian posture, local context and global context features through a new cross-stacked gated recurrence unit (GRU) structure to achieve accurate prediction of pedestrian crossing behavior. Applied onto the surveillance video dataset from the University of California, Berkeley to predict the pedestrian crossing behavior, our model achieves the best results regarding accuracy, F1 parameter, etc. In addition, we conducted experiments to study the effects of time to prediction and pedestrian speed on the prediction accuracy. This paper proves the feasibility of pedestrian crossing behavior prediction based on surveillance video. It provides a reference for the application of edge computing in the safety guarantee of automatic driving.

A Novel Graph-Based Trajectory Predictor With Pseudo-Oracle.

Pedestrian trajectory prediction in dynamic scenes remains a challenging and critical problem in numerous applications, such as self-driving cars and socially aware robots. Challenges concentrate on capturing pedestrians' motion patterns and social interactions, as well as handling the future uncertainties. Recent studies focus on modeling pedestrians' motion patterns with recurrent neural networks, capturing social interactions with pooling- or graph-based methods, and handling future uncertainties by using the random Gaussian noise as the latent variable. However, they do not integrate specific obstacle avoidance experiences (OAEs) that may improve prediction performance. For example, pedestrians' future trajectories are always influenced by others in front. Here, we propose the Graph-based Trajectory Predictor with Pseudo-Oracle (GTPPO), an encoder-decoder-based method conditioned on pedestrians' future behaviors. Pedestrians' motion patterns are encoded with a long short-term memory unit, which introduces temporal attention to highlight specific time steps. Their interactions are captured by a graph-based attention mechanism, which draws OAE into the data-driven learning process of graph attention. Future uncertainties are handled by generating multimodal outputs with an informative latent variable. Such a variable is generated by a novel pseudo-oracle predictor, which minimizes the knowledge gap between historical and ground-truth trajectories. Finally, the GTPPO is evaluated on ETH, UCY, and Stanford Drone datasets, and the results demonstrate state-of-the-art performance. Besides, the qualitative evaluations show successful cases of handling sudden motion changes in the future. Such findings indicate that GTPPO can peek into the future.

Drivers' Visual Attention Characteristics under Different Cognitive Workloads: An On-Road Driving Behavior Study.

In this study, an on-road driving experiment was designed to investigate the visual attention fixation and transition characteristics of drivers when they are under different cognitive workloads. First, visual attention was macroscopically analyzed through the entropy method. Second, the Markov glance one- and two-step transition probability matrices were constructed, which can study the visual transition characteristics under different conditions from a microscopic perspective. Results indicate that the fixation entropy value of male drivers is 23.08% higher than that of female drivers. Under the normal driving state, drivers' fixation on in-vehicle systems is not continuous and usually shifts to the front and left areas quickly after such fixation. When under cognitive workload, drivers' vision transition is concentrated only in the front and right areas. In mild cognitive workload, drivers' sight trajectory is mainly focused on the distant front area. As the workload level increases, the transition trajectory shifts to the junction near the front and far sides. The current study finds that the difference between an on-road test and a driving simulation is that during the on-road driving process, drivers are twice as attentive to the front area than to the driving simulator. The research provides practical guidance for the improvement of traffic safety.

Acceptance of Full Driving Automation: Personally Owned and Shared-Use Concepts.

OBJECTIVE: This study aims to develop user acceptance models for two concepts of full driving automation: personally owned and shared use. BACKGROUND: Many manufacturers have been investing considerably in and actively developing full driving automation. However, factors influencing user acceptance of full driving automation are not yet fully understood. METHOD: This study consisted of two parts: focus group discussions and online surveys. A total of 30 potential users participated in focus groups to discuss their perception of full driving automation acceptance. Based on the findings from focus group discussions, theoretical foundations, and empirical evidence, we hypothesized the acceptance models for both personally owned and shared-use concepts. We tested the models with 310 and 250 participants, respectively, online. RESULTS: The results of focus groups indicated that users' concerns are centered around safety, usefulness, compatibility, trust, and ease of use. The survey results revealed the important roles of perceived usefulness and perceived safety in both models, whereas the direct impact of perceived ease of use was found to be insignificant. The indirect impact of perceived ease of use was less significant in the personally owned than in the shared-use model, whereas usefulness, trust, and compatibility played more important roles in the personally owned when compared with the shared-use model. CONCLUSION: The findings uncovered a chain of constructs that affect behavioral intention to use for both full driving automation concepts. APPLICATION: The framework and outcome of this study provide valuable guidelines that allow better understanding for government agencies, manufacturers, and automation designers regarding users' acceptance of full driving automation.

Identifying the crash characteristics on freeway segments based on different ramp influence areas.

Objective: This study aimed to explore the relationship between crash types and different freeway segments and identify the factors contributing to crashes on different freeway segments. Unlike most of the previous studies on freeway segments, this study separately investigates basic freeway segments, single ramp influence segments, and multiple ramp influence segments. Methods: Nonlinear canonical correlation analysis (NLCCA) and proportionality test were used to identify the relationship between crash types and different freeway segments. The data sets for the different freeway segments accumulated for this study consist of 9,867 crash samples with complete information on all 22 chosen variables. A multinomial logit model (MNL) was used to estimate the influence of crash factors on different freeway segments. Results: The results show that weaving and diverge overlap influence segments (WD) are more likely to have injury or fatal crashes; diverge and diverge overlap influence segments (DD) are more likely to have property damage-only (PDO) crashes; merge and merge overlap influence segments (MM) are more likely to have sideswipe crashes; and WD have non-sideswipe crashes; WD and weaving overlap influence segments (MW) are more likely to have rear end crashes; and MM segments are less likely to have hit object crashes. The contributing factors are identified by MNL and the results show that different traffic variables, environmental variables, vehicle variables, driver variables, and geometric variables significantly affected the likelihood of crashes on different freeway segments. Conclusions: Investigation of crash types and factors contributing to crashes on different freeway segments is based on multiple ramp influence segments, which can promote a better understanding of the safety performance of various freeway segments.

A Cost-Effective Vehicle Localization Solution Using an Interacting Multiple Model-Unscented Kalman Filters (IMM-UKF) Algorithm and Grey Neural Network.

In this paper, we propose a cost-effective localization solution for land vehicles, which can simultaneously adapt to the uncertain noise of inertial sensors and bridge Global Positioning System (GPS) outages. First, three Unscented Kalman filters (UKFs) with different noise covariances are introduced into the framework of Interacting Multiple Model (IMM) algorithm to form the proposed IMM-based UKF, termed as IMM-UKF. The IMM algorithm can provide a soft switching among the three UKFs and therefore adapt to different noise characteristics. Further, two IMM-UKFs are executed in parallel when GPS is available. One fuses the information of low-cost GPS, in-vehicle sensors, and micro electromechanical system (MEMS)-based reduced inertial sensor systems (RISS), while the other fuses only in-vehicle sensors and MEMS-RISS. The differences between the state vectors of the two IMM-UKFs are considered as training data of a Grey Neural Network (GNN) module, which is known for its high prediction accuracy with a limited amount of samples. The GNN module can predict and compensate position errors when GPS signals are blocked. To verify the feasibility and effectiveness of the proposed solution, road-test experiments with various driving scenarios were performed. The experimental results indicate that the proposed solution outperforms all the compared methods.

Identifying factors affecting the safety of mid-block bicycle lanes considering mixed 2-wheeled traffic flow.

OBJECTIVE: Electric bikes (e-bikes) have been one of the fastest growing trip modes in Southeast Asia over the past 2 decades. The increasing popularity of e-bikes raised some safety concerns regarding urban transport systems. The primary objective of this study was to identify whether and how the generalized linear regression model (GLM) could be used to relate cyclists' safety with various contributing factors when riding in a mid-block bike lane. The types of 2-wheeled vehicles in the study included bicycle-style electric bicycles (BSEBs), scooter-style electric bicycles (SSEBs), and regular bicycles (RBs). METHODS: Traffic conflict technology was applied as a surrogate measure to evaluate the safety of 2-wheeled vehicles. The safety performance model was developed by adopting a generalized linear regression model for relating the frequency of rear-end conflicts between e-bikes and regular bikes to the operating speeds of BSEBs, SSEBs, and RBs in mid-block bike lanes. RESULTS: The frequency of rear-end conflicts between e-bikes and bikes increased with an increase in the operating speeds of e-bikes and the volume of e-bikes and bikes and decreased with an increase in the width of bike lanes. The large speed difference between e-bikes and bikes increased the frequency of rear-end conflicts between e-bikes and bikes in mid-block bike lanes. A 1% increase in the average operating speed of e-bikes would increase the expected number of rear-end conflicts between e-bikes and bikes by 1.48%. A 1% increase in the speed difference between e-bikes and bikes would increase the expected number of rear-end conflicts between e-bikes/bikes by 0.16%. CONCLUSIONS: The conflict frequency in mid-block bike lanes can be modeled using generalized linear regression models. The factors that significantly affected the frequency of rear-end conflicts included the operating speeds of e-bikes, the speed difference between e-bikes and regular bikes, the volume of e-bikes, the volume of bikes, and the width of bike lanes. The safety performance model can help better understand the causes of crash occurrences in mid-block bike lanes.

A cellular automaton simulation model for pedestrian and vehicle interaction behaviors at unsignalized mid-block crosswalks.

At unsignalized crosswalks, interactions between pedestrians and vehicles often lead to traffic safety hazards due to absence of traffic control and unclear right-of-ways. To address this safety problem, there is a need to understand the interaction behaviors of pedestrians and vehicles that are complicated by a variety of traffic and roadway attributes. The prime objective of this study is to establish a reliable simulation model to represent the vehicle yielding and pedestrian crossing behaviors at unsignalized crosswalks in a realistic way. The model is calibrated with detailed behavioral data collected and extracted from field observations. The capability of the calibrated model in predicting the pedestrian-interaction events as well as estimating the driver yielding rate and pedestrian delay are also tested and demonstrated. Meanwhile, the traffic dynamics in the vicinity of the crosswalk can be meaningfully represented with simulation results based on the model. Moreover, with the definitions of the vehicle-pedestrian conflicts, the proposed model is capable to evaluate the pedestrian safety. Thereby, the simulation model has the potential to serve as a useful tool for assessing safety performance and traffic operations at existing facilities. Furthermore, the model can enable the evaluation of policy effectiveness and the selection of engineering treatments at unsignalized crosswalks to improve safety and efficiency of pedestrian crossing.