Four-wheel independent steering (4WIS) vehicles have attracted increasing attention for their superior maneuverability. Human drivers typically choose to cross or drive over low-profile obstacles (e.g., plastic bags) to efficiently navigate through narrow spaces, while existing planners neglect obstacle attributes, leading to suboptimal efficiency or planning failures. To address this issue, we propose a novel multimodal trajectory planning framework that employs a neural network for scene perception, integrates 4WIS hybrid A* search to generate a warm start, and formulates an optimal control problem (OCP) for trajectory optimization. Specifically, a multimodal perception network fusing visual information and vehicle states is employed to capture semantic and contextual scene information, enabling the planner to adapt the strategy according to scene complexity (hard or easy planning task). For hard tasks, guided points are introduced to decompose complex tasks into local subtasks, improving search efficiency. The multiple steering modes of 4WIS vehicles—Ackermann, diagonal, and zero-turn—are also incorporated as kinematically feasible motion primitives. Moreover, a hierarchical obstacle handling strategy, which categorizes obstacles as “non-traversable”, “crossable”, and “drive-over”, is incorporated into the node expansion process, explicitly linking obstacle attributes to planning actions to enable efficient decisionmaking. Furthermore, to address dynamic obstacles with motion uncertainty, we introduce a probabilistic risk field model, constructing risk-aware driving corridors that serve as linear collision constraints in the OCP. Experimental results demonstrate the proposed framework’s effectiveness in generating safe, efficient, and smooth trajectories for 4WIS vehicles, especially in constrained environments.

The use of smartphones and in-vehicle infotainment systems has made distracted driving a major cause of traffic accidents. Image-based models are essential for advanced driver assistance systems (ADAS) to detect distractions and alert drivers. However, a major challenge is the poor generalization of image-based detection models, particularly in cross-domain scenarios. Given the variability of real-world driving environments, it is essential to develop a model that performs consistently and efficiently across diverse conditions. In real world, due to the differences between the training scenarios (i.e., source domain) and the application scenarios (i.e., target domain), relying solely on either local alignment or global alignment cannot effectively address the various domain shift scenarios. Therefore, we propose DAF (Dynamic Adaptation Framework), a novel domain adaptation approach that automatically handles cross-domain shifts through dynamic feature alignment. Unlike conventional methods, DAF jointly optimizes global context preservation and local feature discrimination in a unified framework, continuously adapting to domain discrepancies through integrated representation learning. Experiments are conducted on three cross-domain driver distraction datasets including State-farm, AUC-Real, and AUC-Laboratory. The results demonstrate that the proposed method outperforms the compared methods, offering a potential choice for enhancing distraction detection performance in diverse scenarios.

Trajectory planning is crucial for ensuring safety, efficiency, and comfort in autonomous driving, particularly in highway environments, which are critical components of intelligent transportation systems and require vehicles to navigate complex and uncertain traffic conditions. However, related methods are lack of adaptive decision-making capabilities and suffer from high algorithmic complexity. To this end, this paper proposes a novel trajectory planning framework based on deep reinforcement learning and spatio-temporal planning, aiming to improve the adaptability and overall performance of autonomous driving systems in complex traffic environments. First, a data-driven decision-making method based on deep reinforcement learning is developed to establish a stable and human-like decision system. At the planning level, a variable voxel structure is designed to handle different decisions and scenarios, with the spatio-temporal feasible region constructed based on vehicle dynamics model and transportation safety regulations. Trajectory planning is then carried out using piecewise Bézier curves, incorporating various constraints. Finally, a new decision detection module is developed to ensure decision feasibility, while enhancing the self-learning ability of the DRL agent and fully leveraging the performance of the planning layer. Our proposed trajectory planning method is established based on forward decision guidance and backward optimization feedback. Experiments conducted on the highway-env simulator and the CQSkyEyeX real-world dataset show that the proposed framework outperforms the compared reinforcement learning and spatio-temporal planning methods in terms of decision-making efficiency, safety, and human-like planning.

In real-world autonomous driving applications, reinforcement learning (RL) faces several persistent challenges including limited critical training data, low exploration efficiency, and poor policy generalization. These limitations often result in slow convergence or stagnation in local optima, undermining the safety and generalization of intelligent vehicles. In order to address these issues, we propose a generation-enhanced RL framework for challenging decision-making scenarios, which comprises three stages: data acquisition, data utilization, and data generation. Firstly, a risk-aware expert policy is developed to guide exploration on regions proximal to the boundary between safe and unsafe actions during the early phases of training. Secondly, we propose a hierarchical dynamic prioritized replay mechanism to enhance the utilization of collected experiences. The replay buffer dynamically ranks experience samples based on their relevance to current policy updates. By weighting essential transitions more heavily during replay, the agent gains enhanced capability to learn from critical scenarios. Thirdly, to mitigate the scarcity of high-risk transitions in the training data, we propose a conditional generation method based on diffusion model. This model synthesizes diverse and structurally relevant transitions with statistical realism, which is concentrated near the policy decision boundary. By employing the high-risk metric as conditional input, the generation model supplements the replay buffer by covering distributional gaps, leading to policy generalization in rare but safety-critical scenarios. Experimental results demonstrate that our method significantly accelerates policy learning and improves driving performance under complex scenarios. The learned policies also outperform existing baselines, indicating their advantages in terms of safety and generalization.

The integration of connected and automated vehicles (CAVs) into traffic systems holds potential to mitigate undesired disturbances. Nevertheless, coexisting human-driven vehicles (HDVs) introduce complex behavioral disturbances, which has imposed critical challenges for control robustness. This study develops a computational framework based on policy iteration to derive robust control policies with optimized attenuation performance for nonlinear mixed traffic flow. Specifically, robust$H_{\infty }$control problem is solved by applying the framework of zero-sum game, whose solution at the Nash equilibrium is transformed into a Hamilton–Jacobi (HJ) inequality with a Hamiltonian constraint. For achieving desired attenuation performance, the value function is updated by gradient descent based on counterexamples violating Hamiltonian and monotonicity constraints, where the positive definiteness of the value function is ensured by convex neural networks, facilitating the analysis of control stability via Lyapunov methods. By utilizing constraint gaps, the attenuation level is optimized through the analytical formulae derived from the HJ inequality. The stability and algorithm convergence are proved. Experimental results demonstrate the capability of the learned controller to effectively attenuate disturbance propagation and stabilize mixed traffic flow.

4D millimeter-wave radar, as a fundamental sensor for 3D object detection, has gained increasing attention in autonomous driving due to its robustness and additional elevation information. However, the sparsity and noisiness of 4D radar point clouds hinder its broader application. Fusing camera with 4D millimeter-wave radar provides an affordable and robust solution. In this paper, a fast and effective framework HP-C4D for camera and 4D radar fusion is proposed. Three key modules are proposed in HP-C4D. Firstly, we innovatively propose the Height BEVPool to predict the height of each BEV grid with negligible delay increase during the height compression. The predicted height information is incorporated into the final object height prediction. Secondly, Perceptive View Heatmap Guidance (PVHG) is proposed to suppress background noise's generation in image BEV and assist the Height BEVPool for better height prediction. Thirdly, Interactive Guidance Fusion (IGF) is proposed to efficiently fuse image BEV and 4D radar BEV. Extensive experiments on the View-of-Delft (VoD) and TJ4DRadset datasets demonstrate the effectiveness and advance of our proposed method. It is worth highlighting that the 3D mean average precision of our proposed method is 0.71% and 3.85% higher than the latest baseline method LXL (LiDAR-Excluded-Lean) on the VoD and TJ4DRadset datasets, respectively. To the best of our knowledge, HP-C4D is the fastest method for camera and 4D radar fusion in 3D object detection on the VoD dataset, achieving 21.4 FPS on single NVIDIA RTX 3090 GPU. Code will be released at https://github.com/c-yyyy/HP-C4D.

Deep learning-based Autonomous Driving (AD) perception models often exhibit poor generalization due to data heterogeneity in an ever domain-shifting environment. While Federated Learning (FL) could improve the generalization of an AD model (known as FedAD system), conventional models often struggle with under-fitting as the amount of accumulated training data progressively increases. To address this issue, instead of conventional small models, employing Large Vision Models (LVMs) in FedAD is a viable option for better learning of representations from a vast volume of data. However, implementing LVMs in FedAD introduces three challenges:(I)the extremely high communication overheads associated with transmitting LVMs between participating vehicles and a central server;(II)lack of computing resource to deploy LVMs on each vehicle;(III)the performance drop due to LVM focusing on shared features but overlooking local vehicle characteristics. To overcome these challenges, we propose pFedLVM, a LVM-Driven, Latent Feature-Based Personalized Federated Learning framework. In this approach, the LVM is deployed only on central server, which effectively alleviates the computational burden on individual vehicles. Furthermore, the exchange between central server and vehicles are the learned features rather than the LVM parameters, which significantly reduces communication overhead. In addition, we utilize both shared features from all participating vehicles and individual characteristics from each vehicle to establish a personalized learning mechanism. This enables each vehicle’s model to learn features from others while preserving its personalized characteristics, thereby outperforming globally shared models trained in general FL. As a demonstration of the proposed pFedLVM, this paper focuses on the semantic segmentation (SSeg) task. Extensive experiments demonstrate that pFedLVM outperforms the existing state-of-the-art approach by 18.47%, 25.60%, 51.03% and 14.19% in terms of mIoU, mF1, mPrecision and mRecall, respectively.

High-performance vision-based decision-making networks are often limited by hardware capabilities in practical applications. To address this challenge, this study proposes lightweight optimization strategies for decision-making models from the aspects of parameter size, training memory usage, and inference speed. Specifically, an innovative solution is proposed to achieve lightweight parameters. The Video Swin Transformer is employed to simultaneously extract temporal and spatial features, with the network trained using a Prioritized Replay Deep Q-Network (PRDQN) that incorporates risk assessment. To further reduce training memory usage, the Q-target network in PRDQN is removed, and the mellowmax operator is integrated to enhance the training process, resulting in the PRDeepMellow Swin Transformer. After analyzing the inference speed problems encountered by the algorithm in practical applications, the vanilla self-attention is replaced by a linear self-attention based on double softmax, namely Double Softmax Linear Video Swin Transformer (DSLVS Transformer) which improves the inference speed for long sequences. The proposed methods were evaluated across three high-speed lane change scenarios (a static scenario, a dynamic scenario, and a randomly changing scenario). Experimental results demonstrate that the proposed methods can still maintain excellent decision performance after the corresponding lightweight optimizations.

Designing brain-computer interface systems for electroencephalogram (EEG)-based driver drowsiness recognition remains a significant challenge due to the significant variation in EEG signals across subjects and recording sessions. To address this problem, this paper develops a novel two-stage information transfer strategy framework for domain generalization. The framework has two domain mappers to reduce the distribution differences of EEG features from different individuals, a mapper mix block for generating hybrid mapping features, and a domain adversarial neural network (DANN) for drowsiness recognition based on hybrid EEG features. In the process of DANN to capture common features, we additionally employ two models based on self-attention mechanism to capture domain-invariant attention relationships between electrode channels and between frequency bands. Experimental results show that the proposed framework achieves an average accuracy of 81.34% in the leave-one-out cross validation for driver drowsiness recognition, which is higher than the state-of-the-art model with the number of 79.37%. In addition, we explore the impact of EEG features from different frequency bands and brain regions on this cross-subject task. The results show that EEG features from delta, theta and alpha bands can achieve much better performance than the other two bands, and features from the frontal lobe region perform better than the other regions. These findings reveal domain-invariant features and their relationships with brain regions and frequency bands, enhancing our understanding of the underlying messages of EEG signals.

Autonomous vehicles are expected to navigate safely and efficiently in dynamic environments, which requires the seamless integration of efficient global route planning with stable and safe local motion planning. To enhance global route planning, we construct a state-value function to represent the global map and then iterate an inheritable state-transition matrix that can be used to rapidly search globally optimal routes. To improve the safety, stability and adaptability of local motion planning, we propose an adaptive dynamic safe topological structure that combines hierarchical computational layers to decompose decision-making processes and adaptively regulate decision outputs in diverse scenarios. These methods are integrated into a unified framework for cooperative decision-making, which is validated in various scenarios in CARLA. The experimental results demonstrate that the global route planning method efficiently searches optimal routes, while the local motion planning method ensures safe and robust decisions across various driving scenarios and styles. Additionally, the integrated framework achieves effective cooperation between efficient global navigation and effective local decision-making.

The ability to accurately predict the trajectory of surrounding vehicles is a critical hurdle to overcome on the journey to fully autonomous vehicles. To address this challenge, we pioneer a novel behavior-aware trajectory prediction model (BAT) that incorporates insights and findings from traffic psychology, human behavior, and decision-making. Our model consists of behavior-aware, interaction-aware, priority-aware, and position-aware modules that perceive and understand the underlying interactions and account for uncertainty and variability in prediction, enabling higher-level learning and flexibility without rigid categorization of driving behavior. Importantly, this approach eliminates the need for manual labeling in the training process and addresses the challenges of non-continuous behavior labeling and the selection of appropriate time windows. We evaluate BAT's performance across the Next Generation Simulation (NGSIM), Highway Drone (HighD), Roundabout Drone (RounD), and Macao Connected Autonomous Driving (MoCAD) datasets, showcasing its superiority over prevailing state-of-the-art (SOTA) benchmarks in terms of prediction accuracy and efficiency. Remarkably, even when trained on reduced portions of the training data (25%), our model outperforms most of the baselines, demonstrating its robustness and efficiency in predicting vehicle trajectories, and the potential to reduce the amount of data required to train autonomous vehicles, especially in corner cases. In conclusion, the behavior-aware model represents a significant advancement in the development of autonomous vehicles capable of predicting trajectories with the same level of proficiency as human drivers. The project page is available on our GitHub.

Accurately and safely predicting the trajectories of surrounding vehicles is essential for fully realizing autonomous driving (AD). This paper presents the Human-Like Trajectory Prediction model (HLTP++), which emulates human cognitive processes to improve trajectory prediction in AD. HLTP++ incorporates a novel teacher-student knowledge distillation framework. The “teacher” model, equipped with an adaptive visual sector, mimics the dynamic allocation of attention human drivers exhibit based on factors like spatial orientation, proximity, and driving speed. On the other hand, the “student” model focuses on real-time interaction and human decision-making, drawing parallels to the human memory storage mechanism. Furthermore, we improve the model’s efficiency by introducing a new Fourier Adaptive Spike Neural Network (FA-SNN), allowing for faster and more precise predictions with fewer parameters. Evaluated using the NGSIM, HighD, and MoCAD benchmarks, HLTP++ demonstrates superior performance compared to existing models, which reduces the predicted trajectory error with over 11% on the NGSIM dataset and 25% on the HighD datasets. Moreover, HLTP++ demonstrates strong adaptability in challenging environments with incomplete input data. This marks a significant stride in the journey towards fully AD systems.

Lithology identification is the research basis in oil and gas reservoir exploration and is critical for the formation characterization and reservoir development. Traditional lithofacies identification methods rely on the knowledge and experience of geologists and are usually done manually. With the development of deep learning technology and its application in the field of geophysics, lithofacies identification based on deep-learning approach has attracted great attention in recent years. Well logging data have obvious sequence characteristics. Therefore, we propose to use a bidirectional long and short-term memory (BiLSTM) neural network to learn long-term information for more effective lithology facies classification. In addition, we also perform correlation analysis on the input well logging curves and conduct median filter at different scales according to the correlation degree to extract the geological features within data itself and discard the interference of noise. The raw data-based lithofacies identification can reflect the noise resistance of the neural network model to some extent, while the filtered data are more beneficial for the model to extract the geological features correlated with lithofacies and provide accurate classification results. We validate our proposed framework by applying it to a case study from the Council Grove gas reservoir located in Kansas. Furthermore, we compare the effect of input data and network model on the identification results. The experimental results show that the proposed lithofacies identification method has higher classification accuracy.

Ground roll, a common coherent noise in field seismic data, exhibits distinct characteristics, including low frequency, low velocity, and high amplitude. This interference significantly degrades the signal-to-noise ratio (SNR) of recorded seismic data, impacting the reliability of the results of subsequent data analyses such as inversion and interpretation. To address this challenge, numerous techniques have been developed to effectively attenuate or separate ground roll from seismic data. In recent years, deep learning-based approaches for ground roll attenuation have emerged. These methods often incorporate the U-Net with mean squared error (MSE) to train the network in learning to estimate clean signals. In this article, we present a novel approach that utilizes a U-Net framework enhanced by attention mechanisms and residual modules to effectively extract ground roll features from training data. Additionally, we propose a hybrid loss function by integrating the mean squared error (MSE) with an adaptive frequency modulation (AFM) loss. By leveraging Fourier transform principles, the frequency modulation loss guides the network to integrate diverse frequency information within the time-space domain, particularly in scenarios with distinct frequency characteristics between noise and signal contents. Moreover, the incorporation of an adaptive mechanism within the frequency modulation loss enables the network to effectively handle complex and overlapped frequency components. Therefore, with the integration of this adaptive mechanism into the frequency modulation loss, the network significantly enhances its performance in attenuating ground roll. In comparison to both the traditional method and the deep learning (DL) approach, the proposed method excels in its capacity to retain low-frequency useful signals and eliminate ground roll.

In this paper, we propose a new RGB-thermal image segmentation method based on parameter sharing and attention fusion for safe autonomous driving. An encoder-decoder network structure is adopted. The encoder, which has shared convolution layer parameters and private batch normalization layer parameters (parameter sharing scheme), is used to extract features from RGB and thermal images. The extracted features are then fused by spatial and channel attention. The output of each residual block is fused, and the self-learning weight is used to integrate the fusion information of all residual blocks of the same levels. Subsequently, the fused features are integrated through a feature integration (FI) module in the decoder. Cross-entropy supervision of segmentation and edge is performed on the outputs of the decoders. Our proposed method is evaluated and compared with 17 state-of-the-art image segmentation methods, both qualitatively and quantitatively on the MFNet dataset which includes various objects in urban scenes. The results show that the proposed method outperforms previous methods by at least 0.3% and 1.8% in MRecall and MIoU, respectively, providing foundations for the development of autonomous driving technologies for safety enhancement.

The increased use of smartphones and in-vehicle infotainment systems leads to more distraction related accidents. Although numerous deep learning techniques have been developed to identify driver distraction based on images, they often perform poorly or even fail in cross-domain conditions. Retraining models on the target domain is a traditional solution, but it requires a significant number of manually annotated data, time, and computer resources. Therefore, this paper proposes a distance-based domain-adaptive approach for global feature matching. It lowers the$\boldmath{\mathcal{H}}$-divergence at the feature level for cross-domain classification. Specifically, a domain-adaptive algorithm is developed based on partial minimum classification confusion (PMCC) matching. The proposed method first predicts target image category weights using a classification network, and then regularizes them by minimizing the classification confusion. It subsequently employs the regularized category weights as pseudo-labels for target domain images, which are then aligned with identically labelled source domain image features. Three cross-domain distracted driving datasets are used to examine the proposed method, including State-farm, AUC-Real and AUC-Laboratory. The results show that our proposed strategy performs better than the state-of-the-art approaches, which provides a solution to further improve distraction detection performance in various situations.

Heavy rain significantly reduces image visibility, hindering tasks like autonomous driving and video surveillance. Many existing rain removal methods, while effective in light rain, falter under heavy rain due to their reliance on purely spatial features. Recognizing this challenge, we introduce the Wavelet-Spatial Dual Attention Transformer Framework (WSDformer). This innovative architecture adeptly captures both frequency and spatial characteristics, anchored by the wavelet-spatial dual attention (WSDA) mechanism. While the spatial attention zeroes in on intricate local details, the wavelet attention leverages wavelet decomposition to encompass diverse frequency information, augmenting the spatial representations. Furthermore, addressing the persistent issue of incomplete structural detail restoration, we integrate the PriorFormer Block (PFB). This unique module, underpinned by the Prior Fusion Attention (PFA), synergizes residual channel prior features with input features, thereby enhancing background structures and guiding precise rain feature extraction. To navigate the intrinsic constraints of U-shaped transformers, such as semantic discontinuities and subdued multi-scale interactions from skip connections, our Cross Interaction U-Shaped Transformer Network is introduced. This design empowers superior semantic layers to streamline the extraction of their lower-tier counterparts, optimizing network learning. Empirical analysis reveals our method's leading prowess across rainy image datasets and achieves state-of-the-art performance, with notable supremacy in heavy rainfall conditions. This superiority extends to diverse visual challenges and real-world rainy scenarios, affirming its broad applicability and robustness. The source code is available athttps://github.com/Jiongze-Yu/WSDformer.

Driver distraction is a significant factor leading to traffic accidents. Detecting driver distraction is crucial for the development of advanced driver assistance systems (ADAS). With the development of deep learning techniques, advanced computer vision technologies have been continuously applied for driver distraction detection. To date, most distraction detection approaches cannot well be adapted to the distraction behaviors that are not included in the training dataset. To address this problem, we propose a lightweight driver distraction detection method using semisupervised contrastive learning. Unlike other studies that rely on large-scale models, a lightweight vision transformer with convolutional neural network (CNN) obtained by knowledge distillation is adapted to extract features, and the design of the dual-stream backbone network increases the generalization ability without increasing the computational burden. Furthermore, the combination of three-level attention mechanisms (i.e., channel-level, spatial-level, and batch-level) enhances the representative power of the model. Both depth and RGB datasets are used to train and test our proposed method. The experimental results show that our method shows superior performance in comparison with other state-of-the-art methods. Its lightweight architecture is suitable for practical applications. This study contributes to the development of ADAS and provides a new perspective on driver distraction detection.

Traffic congestion is a common phenomenon in city traffic, which may cause drivers' negative emotion to degrade driving safety. It has been reported that music can regulate human emotion and the influence of negative emotion continuously challenges driving safety in post-congestion traffic. Therefore, this study aims to examine the effect of different music intervention strategies on mitigating drivers' negative emotion in post-congestion driving. Three experiments (i.e., driving with soft music, driving with disco jockey (DJ) music, and driving without music) are designed to collect drivers' driving performance measures, eye movement and electroencephalogram (EEG) responses in post-congestion driving. The results show that the music intervention strategies influence drivers' eye movement and EEG responses to varying degrees but do not have distinct effect in driving performance measures. These obtained results indicate that designing personalized music intervention strategies might help mitigate drivers' negative emotion to increase driving safety and comfort.

In this article, a new dataset, the driver emotion facial expression (DEFE) dataset for drivers’ spontaneous emotions analysis is introduced. The dataset includes facial expression recordings from 60 participants during driving. After watching a selected video-audio clip to elicit a specific emotion, each participant completed the driving tasks in the same driving scenario and rated his/her emotional responses during the driving processes from the aspects of dimensional emotion method and discrete emotion method. The study also conducted classification experiments to recognize the scales of arousal, valence, dominance, as well as the emotion category and intensity to establish baseline results for the proposed dataset. Furthermore, this paper compared emotion recognition results difference through facial expressions between dynamic driving and static life scenarios. The results showed that dynamic driving and static life datasets were different in emotion recognition results. To further explore the reasons for the difference in emotion recognition results, the analysis from the AU (action unit) presence perspective was studied. The results showed significant differences in the AUs presence of facial expressions between dynamic driving and static life scenarios, indicating that drivers’ facial expressions may be affected by the driving task to influence the recognition of drivers’ emotions through facial expressions. Therefore, to accurately recognize the drivers’ emotions to establish a reliable emotion-aware human-machine interaction system, thereby improving driving safety and comfort, publishing a human emotion dataset specifically for the driver is necessary. The proposed dataset will be publicly available so that researchers worldwide can use it to develop and examine their driver emotion analysis methods. To the best of our knowledge, this is currently the only public driver facial expression dataset.

The resolution of seismic data determines the ability to characterize individual geological structures in a seismic image. Sparse spike inversion (SSI) is an effective approach for improving the resolution of seismic data. However, the basic assumption of SSI is that the strong reflectivity of the formation is sparse, which may not be a reasonable fit for weak thin-layer reflections. In this study, we propose a deep learning-based method to reconstruct high-resolution seismic data by combining information from the longitudinal reflectivity distribution and lateral geological structure features in the field data. A U-shaped network that fuses residual block and attention mechanism is used to learn the relationship between low resolution and high resolution. In addition, we use a hybrid loss function that combines${\ell _{1}}$loss and structural similarity (SSIM) loss to optimize the network parameters for better distinguishing the geometrical features characterized by structural amplitude changes. To train the network, we adopt a workflow to automatically generate numerous 2-D low-resolution data and their corresponding high-resolution data. In this workflow, the prior information, such as statistical reflectivity distribution of well logs and structural features of the data are considered. The synthetic data and field data tests show that our method can work well compared to the traditional method even though only a few well logs are available. Especially in the field data example, our method recovers thin layers better and yields laterally more consistent high-resolution results.

Seismic wave suffers from amplitude attenuation and phase distortion when propagating in the attenuating media, thus reverse time migration (RTM) for viscous media should take the attenuation effects into consideration. Compensating for the attenuation effects in RTM may occur numerical instability because of the exponential amplification of the extrapolated wavefields. To obtain stable imaging results, we have developed a stabilized Q-compensated reverse time migration in the frequency domain. This algorithm is implemented by the following steps: first, we use the Kolsky-Futterman model to derive a frequency-domain viscoacoustic wave equation, which can simulate the amplitude loss and phase dispersion effects separately. Then, we calculate the source wavefields in the viscoacoustic media. Next, treating the recorded (viscoacoustic) data as the receiver sources, we can obtain the phase-dispersion-only and viscoacoustic receiver wavefields, which can be used to construct the stabilized Q-compensated receiver wavefields. Finally, we apply the deconvolution imaging condition for obtaining a Q-compensated image. A simple anticline model and gas chimney model are used to verify the effectiveness of the proposed approach. The Q-compensated images for the noise-free data indicate the algorithmic stability and compensation accuracy of the proposed scheme. The noisy data tests for the gas chimney model demonstrate the good anti-noise property of our method. The field data applications further prove its feasibility and practicability.

High-quality fusion images with infrared and visible information contribute to intelligent and safe driving. In the infrared and visible fusion images, the useless noise information in infrared image makes the fusion image unclear, which leads to the loss of texture information from visible image. In order to solve this problem, we propose a novel two-stage network (i.e., SOSMaskFuse) that can effectively reduce the noise and extract the important thermal information in infrared images as well as display sufficient texture details of visible images. In the first stage of this network, a salient object segmentation (SOS) network is proposed. Infrared images are fed into the SOS network to obtain the corresponding binarization mask of the region of interest. In the second stage, in each layer of features extracted by the encoder network, the newly proposed IMV-F (infrared mask visible fusion) fusion strategy uses the mask to decompose both infrared and visible images into infrared-foreground, visible-foreground, infrared-background and visible-background, and then fuses the foreground and background parts separately into fused-foreground and fused-background. Finally, the decoder network reconstructs the fused features into the final fused image. Compared with eighteen competitive algorithms on three public datasets, the experimental results indicate that our proposed network can produce high quality fused images with clear background texture information while highlighting infrared thermal information. Our proposed SOSMaskFuse generally outperforms the eighteen compared methods from both the quantitative and qualitative perspectives.

Although latent hazard notification for highly automated driving is expected to enhance traffic safety, its practical effects have yet to be verified. This study systemically investigated the expected safety benefits and driver behavioral adaptation based on structural equation modeling. First, we developed a notification system to inform drivers of latent hazards with auditory alerts and conducted a driving simulation experiment involving eyes-off-road situations. To test the system, we adopted two types of events (i.e., the collision avoidance function working or failure) in which latent hazards transform into immediate risks. Then, a measurement model was developed to evaluate driver trust, driver attention, and traffic safety. Subsequently, we examined the corresponding causal relationships. On the one hand, latent hazard notification significantly improves driver attention (i.e., more fixations on latent hazards, less engagement in non-driving-related tasks, and faster notice of immediate risks), which significantly enhances traffic safety. On the other hand, latent hazard notification significantly increases driver trust, which lowers driver attention and consequently impairs traffic safety. This causality reveals driver behavioral adaptation, although driver trust does not directly affect traffic safety. Overall, we find that latent hazard notification for highly automated driving can improve traffic safety, but the consequent driver behavioral adaptation impairs 15.12% of the expected safety benefits.

For highly automated vehicles, effective take-over performance measures are essential for establishing quantitative take-over models and exploring approaches to improve take-over performance. However, there is a lack of comprehensive take-over performance measures that suitably combine multiple objective metrics based on an average evaluation from human drivers. In this study, we proposed a human-centered comprehensive measure of take-over performance (HCMTP). There are four main building blocks for the HCMTP. First, we adopted sparse principal component analysis to identify the main aspects of take-over performance based on multiple original objective take-over performance metrics. Second, we developed a scale of take-over performance assessment to obtain drivers’ original subjective self-assessments of take-over performance. Third, we established nonlinear individual mapping functions to acquire different drivers’ evaluation criteria for take-over performance. Fourth, we proposed a relabeling algorithm to obtain drivers’ average evaluation of take-over performance. To verify the effectiveness of the HCMTP, we conducted a verification experiment involving 68 participants. The results indicate that the HCMTP is effective and able to reduce the interference of individual differences, stochasticity, and data imbalance. This study contributes to identifying the main aspects of take-over performance, systematically understanding how human drivers subjectively evaluate take-over performance, and evaluating drivers’ take-over performance comprehensively.

Reliability allocation is a significant link in product design. To solve the problems of poor rationality of data use, considerable difficulty of calculation using existing methods, low accuracy of results, and weak ability of experts to express and process fuzzy information, this article proposes a reliability allocation method for the initial stage of product design based on the 2-tuple linguistic weighted Muirhead mean (2TLWMM) operator and 2-tuple linguistic best-worst method (2TLBWM). The 2TLBWM introduces 2-tuple linguistic to enhance the experts’ ability to express fuzzy information. The 2TLWMM operator is used to judge the reliability index ranking of each subsystem, thereby improving the rationality of scoring data, providing a basis for experts to establish the comparison vector table, and reducing the influence of experts’ subjective factors. Experts use 2TLBWM, refer to the ranking results of subsystem reliability index, only need to establish a contrast vector table, and can calculate the weight of the subsystem, thereby reducing the number of comparisons between elements and computational complexity. The advantages of the proposed method are illustrated by a specific case.

Improving the vertical resolution of seismic data to satisfy the demands for detailed characterization of reservoirs is an important part in seismic data processing. The sparse spike inversion(SSI) technique greatly enhances the seismic resolution by assuming that the reflection coefficients follow the sparse distributions. However, its characterization for the spatial structure of thin and thin interbed reservoirs is still limited. In this letter, we propose a novel approach for enhancing seismic resolution using a bidirectional long short-term memory(BiLSTM) neural network by extracting reflection coefficients directly from seismic data. By designing the network architectures, multiple seismic samples map to the specific reflection coefficient. Compared with the SSI method, the BiLSTM neural network provides higher resolution inversion results on model data, which demonstrates the effectiveness of the proposed approach for thin structure characterizations. The convergence speed of the proposed method is fast and the training process is stable. In addition, we conduct the high-resolution seismic data processing on the field data based on the transfer learning technology. High-resolution processing results illustrate the generalization ability and adaptability of the BiLSTM neural network.

Distracted driving has been recognized as a major challenge to traffic safety improvement. This article presents a novel driving distraction detection method that is based on a new deep network. Unlike traditional methods, the proposed method uses both temporal information and spatial information of electroencephalography (EEG) signals as model inputs. Convolutional techniques and gated recurrent units were adopted to map the relationship between drivers’ distraction status and EEG signals in the time domain. A driving simulation experiment was conducted to examine the effectiveness of the proposed method. Twenty-four healthy volunteers participated and three types of secondary tasks (i.e., cellphone operation task, clock task, and 2-back task) were used to induce distraction during driving. Drivers’ EEG responses were measured using a 32-channel electrode cap, and the EEG signals were preprocessed to remove artifacts and then split into short EEG sequences. The proposed deep-network-based distraction detection method was trained and tested on the collected EEG data. To evaluate its effectiveness, it was also compared with the networks using temporal or spatial information alone. The results showed that our proposed distraction detection method achieved an overall binary (distraction versus nondistraction) classification accuracy of 0.92. In terms of task-specific distraction detection, its accuracy was 0.88. Further analysis on the individual difference in detection performance showed that drivers’ EEG performance differed across individuals, which suggests that adaptive learning for each individual driver would be needed when developing in-vehicle distraction detection applications. Note to Practitioners—Driver distraction detection is crucial for safety enhancement to avoid crashes caused by nondriving-related activities, such as calling and texting while driving. Related previous studies mainly focus on detection by monitoring head and eye movement using computer vision technologies or by extracting indicators from driving performance measures for driver state inference. However, complex traffic environments (e.g., dynamically changing light distribution on driver’s face and nighttime driving with low illumination) strongly limit the effectiveness of computer vision technologies, and the driving performance characteristics may also be caused by factors other than distraction (e.g., fatigue). To solve these problems, this article seeks to develop a deep learning-based approach to map the unique relationship between driver distraction and the bioelectric electroencephalography (EEG) signals that are not affected by traffic environments. The proposed method can be integrated into the driver assistance systems and autonomous vehicles to deal with emergency situations that need drivers to handle. The timely detection of distraction by our method will significantly facilitate its practical applications in collision avoidance or danger mitigation in the handover process.

Seismic high-resolution (HR) reconstruction is a crucial process for identifying increasingly thin layers from observed seismic data. Nowadays, machine learning (ML) has been adopted in seismic resolution improvement; however, most of the ML-based methods directly use 1-D neural networks and ignore the spatial information along seismic traces. Thus, these methods cause poor stability and accuracy issues in improving the resolution of multidimensional seismic data. In this article, we propose to incorporate a structural constraint into a neural network framework to perform seismic HR reconstruction. The loss function of our network consists of two parts. One part is used to extract useful HR seismic trace features from the training set generated by the well-log data so that the network can facilitate the solution from low resolution (LR) to HR. More importantly, the other part is used to preserve the reflection structure features of seismic data, which guarantees the stability and accuracy of HR reconstruction results. Tests on synthetic and field examples verify that the proposed method can provide a better reconstruction result in terms of resisting noise, retrieving thin layers, and preserving lateral structure than the traditional method and the ML-based method with the same network framework.

As human-machine collaborative driving systems, highly automated driving vehicles require human drivers to take over when take-over requests are triggered. Extensive studies have shown that drivers’ take-over performance is affected by their fatigue state, traffic conditions, and the take-over time budget (TB). However, there is still a paucity of a systematic understanding of how these factors affect take-over performance, which prevents the implementation of adaptive take-over systems. This study establishes a highly accurate take-over performance prediction model to systematically explore the effects of these factors on take-over performance and to propose an adaptive TB adjustment strategy for highly automated driving vehicles. First, we propose metrics to evaluate drivers’ fatigue states and the relative positions of surrounding traffic. Second, a generalized additive model is established to predict take-over performance and accurately evaluate the influence of the aforementioned factors on take-over performance. Based on the model, we propose an adaptive adjustment strategy of the TB for take-over systems and demonstrate its effectiveness by a verification experiment. This study contributes to understanding the influence of drivers’ passive fatigue states, the relative positions of surrounding traffic, and the TB on drivers’ take-over performance as well as to the development of adaptive take-over systems for highly automated vehicles.

Due to the growing interest in automated driving, a deep understanding on the characteristics of human driving behavior is critical for human-like autonomous vehicles. Among various driving behaviors, lane change is the most important one for vehicle lateral driving safety. This study proposes an unsupervised method to extract and discover the behavioral patterns of lane change maneuvers for the purpose of exploring the composed behavioral patterns during lane change. This method involves two phases: Firstly, the lane change sequences will be segmented into blocks using time-series segmentation algorithms. Three segmentation algorithms were utilized in this study. In the second phase, the segments will be clustered to find the corresponding behavioral pattern of each segment. Two extended latent Dirichlet allocation (LDA) models were adopted to cluster the segments. The combination of different segmentation and clustering algorithms were evaluated and compared by employing entropy and perplexity as the evaluation criteria. Collected lane change data from naturalistic driving were applied to examine its effectiveness. The results show that this method could effectively mine descriptive behavioral patterns from lane change data. This study provides a promising data mining solution to facilitating deep and comprehensive understanding on driver lane change behaviors, which will promote the development of human-like autonomous vehicles.

This paper proposes a novel driver-automation shared control based on a potential-field-driven model predictive controller (PF-MPC) and a two-layer fuzzy strategy (TLFS) to address driver-automation conflicts and control authority allocation issues. The PF-MPC approach based on the driver-vehicle model is introduced to deal with obstacles avoidance and driver-automation conflicts. The potential field is constructed to evaluate the driving risk by considering the driving environment and vehicle states, meanwhile, it is also involved in the optimized objective in the PF-MPC controller for obstacle avoidance. The tuning weight is designed to adjust the trade-off between the motion planning-related cost and driver-related cost to reduce driver-automation conflicts. To further alleviate the conflict and control authority allocation between the human driver and PF-MPC controller, the TLFS for shared control is designed based on the evaluation of the driving risk level and conflict situation, and the values of the tuning weight and cooperative coefficient are determined using the fuzzy control method. Moreover, comparative studies are conducted to verify the driving safety and conflict management performance of the proposed shared control method on a straight road and a curvy road. The results show that the proposed shared control method can help drivers avoid obstacles safely and alleviate the driver-automation conflicts in different driving conditions.

In recent years, deep learning technologies for object detection have made great progress and have powered the emergence of state-of-the-art models to address object detection problems. Since the domain shift can make detectors unstable or even crash, the detection of cross-domain becomes very important for the design of object detectors. However, traditional deep learning technologies for object detection always rely on a large amount of reliable ground-truth labelling that is laborious, costly, and time-consuming. Although an advanced approach CycleGAN has been proposed for cross-domain object detection tasks, the ability of CycleGAN to reduce the divergence across domains at the feature level is limited. In this paper, a stepwise domain adaptation (SDA) detection method is proposed to further improve the performance of CycleGAN by minimizing the divergence in cross-domain object detection tasks. Specifically, the domain shift is addressed in two steps. In the first step, to bridge the domain gap, an unpaired image-to-image translator is trained to construct a fake target domain by translating the source images to the similar ones in the target domain. In the second step, to further minimize divergence across domains, an adaptive CenterNet is designed to align distributions at the feature level in an adversarial learning manner. Our proposed method is evaluated in domain shift scenarios based on the driving datasets including Cityscapes, Foggy Cityscapes, SIM10k, and BDD100K. The results show that our method is superior to the state-of-the-art methods and is effective for object detection in domain shift scenarios.

Truck drivers is a key group to promote road safety. For them, proper priority ranking scheme of content adaptation design benefits the high system effectiveness of in-vehicle driver decision support. Taking Chinese truck drivers as an example, the present study revealed the context-adaptive support information from the perspective of truck drivers; their perceptions of in-vehicle information contents priority in 6 typical driving contexts. Data of 19 participants from 7 logistics companies were collected using a simulation interview method. Based on qualitative summary and statistical analysis, the results are summarized in two aspects; contextual information priority and impacts of driving experience on it. From the perspective of truck driver requirement, these results provide references for the design of context-adaptive driver decision support.

Detection of driver cognitive distraction is critical for active safety systems of road vehicles. Compared with visual distraction, cognitive distraction is more challenging for detection due to the lack of apparent exterior features. This paper presents a novel real-time detection algorithm for driver cognitive distraction by using support vector machine (SVM). Data are collected from 26 subjects, driving in typical urban and highway scenarios in a simulator. The chosen urban scenario is the stop-controlled intersection and the highway scenario is the speed-limited highway. Driver cognitive distraction while driving is induced by clock tasks which compete with the main driving tasks for visuospatial short working memory. For each subject, distracted driving instances and the equal number of non-distracted driving instances were collected (24 for urban scenario and 20 for highway scenario in total). Features concerning both driving performance and eye movement are used for training and validation. The proposed algorithm have correct rate of 93.0% and 98.5% for highway and urban scenarios respectively. Results also show that driver distraction can be recognized 6.5 s to 9.0 s after its happening, indicating good performance of the detection algorithm.

Driver distraction has been identified as one major cause of unsafe driving. The existing studies on cognitive distraction detection mainly focused on high-speed driving situations, but less on low-speed traffic in urban driving. This paper presents a method for the detection of driver cognitive distraction at stop-controlled intersections and compares its feature subsets and classification accuracy with that on a speed-limited highway. In the simulator study, 27 subjects were recruited to participate. Driver cognitive distraction is induced by the clock task that taxes visuospatial working memory. The support vector machine (SVM) recursive feature elimination algorithm is used to extract an optimal feature subset out of features constructed from driving performance and eye movement. After feature extraction, the SVM classifier is trained and cross-validated within subjects. On average, the classifier based on the fusion of driving performance and eye movement yields the best correct rate and F-measure (correctrate = 95.8 ± 4.4%; for stop-controlled intersections and correct rate = 93.7 ± 5.0%; for a speed-limited highway) among four types of the SVM model based on different candidate features. The comparisons of extracted optimal feature subsets and the SVM performance between two typical driving scenarios are presented.

Lane change maneuver recognition is critical in driver characteristics analysis and driver behavior modeling for active safety systems. This paper presents an enhanced classification method to recognize lane change maneuver by using optimized features exclusively extracted from vehicle state and driver operation signals. The sequential forward floating selection (SFFS) algorithm was adopted to select the optimized feature set to maximize the k-nearest-neighbor classifier performance. The hidden Markov models (HMMs), based on the optimized feature set, were developed to classify driver lane change and lane keeping maneuvers. Fifteen drivers participated in the road test for validation with an accumulation of 2,200 km naturalistic driving data, from which 372 lane changes were extracted. Results show that the recognition rate of lane change maneuver achieves 88.2%. The numbers are 87.6% and 88.8% for left and right lane change maneuvers, respectively, superior to the results from conventional classifiers.

Driver distraction has been identified as an important driving safety issue. However, existed studies focused less on low-speed condition, especially at intersections. This paper aims to find the impact of driver cognitive distraction on vehicle performance at stop-controlled intersections. Eight subjects (young adult: 4, older adult: 4) participated in this study and each of them drove through 40 stop-controlled intersections. The intersections were presented randomly at two levels of FOV (field of view). Driver cognitive distraction was induced by a one-back task and a clock task. Results showed that the cognitive tasks led to more abrupt steering in both age groups while significant influence on lane-keeping capability was only observed in the young group. Steering smoothness was mainly influenced by the cognitive tasks at brake on-restart phase in the young group while at after-restart phase in the older group. Impaired longitudinal control (stop for watching) was observed in the older adult group. These findings can be applied to automatically recognize driver distraction at stop-controlled intersections in future.

Eco-driving technologies are able to largely reduce the fuel consumption of ground vehicles. This paper presents how to determine the fuel-optimized operating strategies of passenger cars under cruising process. The design naturally casts into an optimal control problem with the S-shaped engine fueling rate as the integrand of cost function. The solutions are numerically solved by the Legendre pseudospectral method, of which many are found to demonstrate periodic behaviors. In the periodic operation, the engine switches between the minimum brake specific fuel consumption (BSFC) point and the idling point, while the vehicle speed oscillates between its upper and lower bounds. The formation of periodic operation are analyzed and explained by the π-test theory and steady state analysis method.

This paper discusses the reconstruction of the subsurface from radar data using a time-domain reconstruction method. Before the reconstruction is conducted, data preprocessing is carried out. The preprocessing transforms measured data transmitted and received by 3-D dipole antennas into equivalent data excited and received by 3-D ideal electric dipoles and, then, into equivalent data excited and received by 2-D ideal electric dipoles. Distributions of permittivity, conductivity, and permeability are reconstructed from the preprocessed data by a 2-D forward-backward time-stepping waveform inverse scattering technique. Two 2-D examples of permittivity and conductivity reconstruction from synthetic noisy and field borehole radar data are demonstrated to show the effectiveness of the proposed method