Object detection is an important task in intelligent transportation scenes, and the performance of small object detection is directly related to its practical application. Small objects typically have poor detection performance due to reasons such as inadequate feature information and localization challenges. This paper proposes object feedback and feature retention, which can effectively improve the performance of small object detection. Firstly, we provide an in-depth analysis of popular methods in small object detection. In order to address the issue that location loss functions are mainly designed for general object detection, this paper proposes the Small Object Intersection over Union (SOIoU) loss function based on object feedback. This function can be adaptively optimized for small objects according to their size, thereby making the model more focused on small objects. In addition, for the issue of insufficient detail information in the output layer, this paper proposes the Small Object Path Aggregation Network (SOPANet) based on feature information retention, which can effectively enhance the information conditions for detecting small objects. Based on the above, the Object Feedback and Feature Retention-You Only Look Once (OFFR-YOLO) model is obtained in this paper. The proposed method improved the small object detection performance by 2.5 % and 4.8 % on the COCO (Common Objects in Context) and BDD100K (Diverse Driving Dataset for Heterogeneous Multitask Learning) datasets, respectively. The experimental results show that the proposed method can effectively improve the detection performance of small objects.

CNN-based methods have made great progress in single-image rain removal. Most recent methods improve performance by increasing the depth of the network. To fully extract local and global features while reducing inference time, we propose a top-to-down attribute-insensitive multiscale hourglass network for rain streak and raindrop removal. For the rain removal task, we expect that the constructed network can accurately identify the various attributes of the rain information characteristics of the small target. Considering the difference in the size, shape, direction and density of rain streak and raindrop, inspired by the performance of hourglass architecture to capture multiscale features in human pose estimation, we introduce an attribute-insensitive hourglass module to recognize the attributes of rain streak and raindrop in a unified framework. This feature extraction module could capture the characteristics of rain streak and raindrop with different attributes. This stacked hourglass blocks down-sample features and then up-samples them back to the original resolution based on discrete wavelet transform and inverse discrete wavelet transform. We perform extensive experiments on five synthetic and real-world de-raining datasets to validate the effectiveness of our proposed network on rain streak and raindrop removal. The qualitative and quantitative results show that our method is suitable for removing rain streak and raindrop in a unified framework. We present the results of generalization and ablation study for key components, we also report the accuracy of semantic segmentation after preprocessing with all rain removal methods. Our source code will be available on the GitHub: https://github.com/Ruini94/AIMHNet.

In this paper, a robust integrated control scheme is developed for autonomous vehicles path following considering both yaw stability and roll stability. First, a controller with H infinity and optimal guaranteed cost performances is presented, and the poles of the closed-loop system are configured in a given disk region to improve the state response performance. The controller outputs front and rear wheel angles for path following, as well as an external yaw moment and an external roll moment for lateral stability control. Secondly, the braking forces acted on four wheels are optimally distributed by means of quadratic programing. Besides, the steering angles of the four wheels conform to the Ackermann geometry relation. Finally, the proposed control scheme is verified by CarSim/Simulink co-simulation, the results show that the scheme has better performance on path following accuracy and stability of yaw and roll. Meanwhile, constraining the poles in a given disk region can improve the state response performance of the vehicle during operation.

Determining Hopf bifurcation points is an important task in the study of nonlinear dynamic systems, but existing methods require a large amount of computation and have low ratios of success. In this study, we present a numerical method to determine Hopf bifurcation points: the Split Iteration Technique and prove its convergence. Compared with existing methods, this method results in much less computation and a much higher ratio of success.

With the ongoing proliferation of electric vehicle (EV) sharing systems, the need for dependable planning and operation of internet data centers and charging stations has become increasingly crucial. The article suggests a combined planning model that integrates a linked transportation system, power distribution system, and communication system (CS). The model combines the optimization of capacity and location for shared EV charging stations, Internet data centers, substations, and network expansion plans for both the power distribution and CSs. A new approach is presented for forecasting EV traffic flow by utilizing a multifactor based graph convolutional neural network and long short-term memory model that takes into account spatial-temporal correlations and external factors that may influence the flow. An enhanced weighting technique is employed to assess user preference factors, which are then integrated with projected EV traffic flow to anticipate shared EV demand. A practical experiment is conducted using simulations and data from an actual traffic scenario to assess the validity and efficacy of the proposed integrated planning model.

To further improve the feature representation ability of haze removal model on small-scale datasets, we propose a three-hierarchical encoder-decoder integrating learning architecture. Unlike increasing the depth of network to improve the performance, we use three sets of transfer learning encoders to aggregate features on multiple patches. This strategy makes full use of deep and shallow features from different spatial parts of original hazy images. These different spatial parts are used to jointly optimize non-homogeneous hazy images with different scene depths and degradation degrees. We propose an integrated learning decoder, in which the internal integrated learning module decodes by gradually infusing the deep and shallow features of the identical patch. We fuse all encoded features at each stage for original resolution decoder. We report the quantitative and visual results of eight advanced haze removal methods (2019 to 2021). We compare these methods on five real-world and synthetic hazy datasets, including homogeneous and non-homogeneous degraded datasets. We present the results of the ablation study for key components. The generalization of our method is significantly better than other methods on real hazy scenes. We report improvements in semantic segmentation performance. This result could be used for autonomous driving, automated guided vehicle scheduling in closed factory area, etc. (c) 2023SPIE and IS&T

为了提升共享电动汽车系统的经济性与运营效率,需要充分挖掘用户对共享使用价格的响应机制以及用户的充电等待行为.同时,用户流量需求的精准预测是提供合理共享定价方案的重要基础.基于此,首先建立以数据驱动的用户流量预测模型,通过结合图卷积神经网络与长短时记忆模型,捕捉用户需求的时空特征和动态相关性,并结合时间与天气特征因素实现多时间步长的准确流量预测.其次,考虑用户在共享系统下的等待自适应行为,建立基于用户等待成本的共享系统服务质量模型.而后以实现最大化共享系统利润收益与服务质量为目标,建立多 目标的共享定价模型以制定不同时刻与路径上的共享电动汽车使用价格.最后,以上海市虹口区EVCARD交通测试系统的算例分析,论证所提流量预测模型的准确性以及多 目标定价模型的有效性与经济性.研究结果表明:所提的SEV多因素融合需求预测模型较其现有的常用预测方法能获得更加准确的多时间步长流量预测;通过合理设置共享驾驶的空间和时间变化价格信号,可以提升共享电动汽车系统的运营盈利能力,并且实现了更好的系统利润收益与用户服务质量平衡的运行需求.

针对现有语义分割网络模型在道路语义分割方面检测精度低、计算量大等问题,基于BiSeNet V2 网络模型进行优化改进,引入一种高效的通道注意力(efficient channel attention,ECA)模块,在BiSeNet V2 的语义分支和细节分支的每个阶段末端分别加入ECA,得到ECA-Semantic-BiSeNet V2 网络.使用实车采集道路图像数据进行标注并构建自采数据集,在Cityscapes数据集、KITTI数据集及自采数据集上分别对改进前后的网络模型进行试验验证.试验结果表明,与BiSeNet V2 模型方法相比,本研究方法在Cityscapes数据集上MIoU提高 14.01%,在KITTI数据集上MIoU提高 1.86%,同时在BiSeNet V2 的语义分支加入ECA后运算量增加 0.02 GFlops的条件下,模型推理速度提高了 7.82 帧/s.

With the development of globalization and artificial intelligence, as well as the outbreak of COVID-19, unmanned vehicles have played an important role in cargo distribution. In order to better analyze the research directions of unmanned vehicle distribution, this paper summarizes the models and algorithms of unmanned vehicle distribution optimization. The research results show that most of the studies have established the goal of optimizing the total costs or travel time. Many researchers have begun to study multi-objective optimization problems, but there are certain limitations, so some studies convert these problems into single-objective optimization for solving, such as converting time and energy consumption into cost, waiting time into distance, and time delay into penalty cost. With the development of unmanned vehicle distribution technology, in future research, a multi-objective model with the lowest cost, the shortest distance and the best security should be established and solved. Most studies have proposed heuristic algorithms for solving the unmanned vehicle distribution problem, and improved optimization solutions have been obtained. In order to ensure the diversity of solution methods, and give consideration to solution time and solution quality, hybrid methods with other algorithms will be a future research direction, for example, the combination of heuristic algorithm and exact algorithm. With the gradual deepening of research, integrated distribution of multiple types of unmanned equipment will become the focus of future research.

This article studies the control problem of autonomous vehicle path following with coordination of active front steering and differential steering. A hierarchical control scheme including upper layer and lower layer is proposed. In the upper layer controller, a linear quadratic regulator based on extended state observer is proposed to generate the front-wheel steering angle and external yaw moment, where extended state observer is used to estimate and compensate for the system uncertainty and external disturbance which enhances the capability of the vehicle to suppress the disturbance. A brake force distribution scheme based on the theory of control allocation is proposed in the lower layer controller to optimize and coordinate each wheel brake force to achieve differential steering. Finally, the effectiveness of proposed control scheme is verified in a co-simulation platform based on CarSim/Simulink; it can be concluded that the linear quadratic regulator based on extended state observer scheme not only has a few parameters need to be tuned, but also has the capability of active disturbance rejection.

Object detection is one of the fundamental tasks in computer vision, holding immense significance in the realm of intelligent mobile scenes. This paper proposes a hybrid cross-feature interaction (HCFI) attention module for object detection in intelligent mobile scenes. Firstly, the paper introduces multiple kernel (MK) spatial pyramid pooling (SPP) based on SPP and improves the channel attention using its structure. This results in a hybrid cross-channel interaction (HCCI) attention module with better cross-channel interaction performance. Additionally, we bolster spatial attention by incorporating dilated convolutions, leading to the creation of the cross-spatial interaction (CSI) attention module with superior cross-spatial interaction performance. By seamlessly combining the above two modules, we achieve an improved HCFI attention module without resorting to computationally expensive operations. Through a series of experiments involving various detectors and datasets, our proposed method consistently demonstrates superior performance. This results in a performance improvement of 1.53% for YOLOX on COCO and a performance boost of 2.05% for YOLOv5 on BDD100K. Furthermore, we propose a solution that combines HCCI and HCFI to address the challenge of extremely small output feature layers in detectors, such as SSD. The experimental results indicate that the proposed method significantly improves the attention capability of object detection in intelligent mobile scenes.

The antidisturbance control problem of autonomous vehicle path tracking considering lateral stability is studied in this paper. This paper proposes an improved active disturbance rejection control (IADRC) control method including an improved extended state observer (IESO) and an error compensator based on LQR, where a new continuous nonlinear function is proposed in the IESO instead of the classical piecewise function. Based on the IADRC, an autonomous vehicle path-tracking controller considering lateral stability is designed. Using the output wheel steering angle and external yaw moment, the IESO estimates the disturbance value and compensates for the disturbance in the feedback to meet the goal of antidisturbance control. Based on the concept of control allocation (CA), the control distributor is designed to distribute the external yaw moment to the four wheels in a reasonable and optimal way to achieve differential braking. Finally, the control scheme is evaluated in the form of CarSim/Simulink cosimulation; the results show that the proposed autonomous vehicle path-tracking control scheme has better path-tracking effect and higher antidisturbance robustness.

With the development and progress of information technology, especially V2X technology, the research focus of intelligent vehicles gradually shifted from single-vehicle control to multi-vehicle control, and the cooperative control system of intelligent connected vehicles became an important topic of development. In order to track the research progress of intelligent connected vehicle cooperative driving systems in recent years, this paper discusses the current research of intelligent connected vehicle cooperative driving systems with vehicles, infrastructure, and test sites, and analyzes the current development status, development trend, and development limitations of each object. Based on the analysis results of relevant references of the cooperative control algorithm, this paper expounds on vehicle collaborative queue control, vehicle collaborative decision making, and vehicle collaborative positioning. In the case of taking the infrastructure as the object, this paper expounds the communication security, communication delay, and communication optimization algorithm of the vehicle terminal and the road terminal of intelligent connected vehicles. In the case of taking the test site as the object, this paper expounds the development process and research status of the real vehicle road test platform, virtual test platform, test method, and evaluation mechanism, and analyzes the problems existing in the intelligent connected vehicle test environment. Finally, the future development trend and limitations of intelligent networked vehicle collaborative control system are discussed. This paper summarizes the intelligent connected car collaborative control system, and puts forward the next problems to be solved and the direction of further exploration. The research results can provide a reference for the cooperative driving of intelligent vehicles.

Low-light image enhancement is a preprocessing work for many recognition and tracking tasks for autonomous driving at night. It needs to handle various factors simultaneously including uneven lighting, low contrast, and artifacts. We propose a novel end-to-end Retinex-based illumination attention low-light enhancement network. Specifically, our proposed method adopts multibranch architecture to extract rich features for different depth levels. Meanwhile, we consider the features from different scales in built-in illumination attention module. We encode reflectance features and illumination features into latent space based on Retinex in each submodule, which could cater for highly ill-posed image decomposition tasks. It aims to enhance the desired illumination features under different receptive fields. Subsequently, we propose a memory gate mechanism to learn adaptively long-term and short-term memory. Their weight could control how many high-level and low-level features should be reserved. This method could improve the image quality from both different feature scales and feature levels. Comprehensive experiments on BDD10K and cityscapes datasets demonstrate that our proposed method outperforms various types of methods in terms of visual quality and quantitative metrics. We also show that our proposed method has certain antinoise capability and generalizes well without fine-tuning when dealing with unseen images. Meanwhile, our restoration performance is comparable to that of advanced computationally intensive models.(1)

Cybersecurity is one of the most important challenges in the intelligent connected vehicle system. Interconnected vehicles are vulnerable to different network security attacks, which endanger the safety of passengers. This review paper firstly analyses the reasons why the current vehicle network is vulnerable to network attack and summarizes the three implementation methods of network security threats. The necessity of vehicle network security research and deployment is also analyzed. After giving a short introduction to the vehicular network security, this review paper identifies major security attacks on intelligent connected vehicles. Then the security enhancement technology of vehicle networks from three aspects are introduced, including vehicle network data encryption technology, vehicle network message authentication technology, and vehicle network anomaly intrusion detection technology. Then we analyze three common methods of abnormal intrusion detection in vehicle networks and explore the future research for preventing attacks on the network security of intelligent vehicle systems.

Pedestrian detection is a specific application of object detection. Compared with general object detection, it shows similarities and unique characteristics. In addition, it has important application value in the fields of intelligent driving and security monitoring. In recent years, with the rapid development of deep learning, pedestrian detection technology has also made great progress. However, there still exists a huge gap between it and human perception. Meanwhile, there are still a lot of problems, and there remains a lot of room for research. Regarding the application of pedestrian detection in intelligent driving technology, it is of necessity to ensure its real-time performance. Additionally, it is necessary to lighten the model while ensuring detection accuracy. This paper first briefly describes the development process of pedestrian detection and then concentrates on summarizing the research results of pedestrian detection technology in the deep learning stage. Subsequently, by summarizing the pedestrian detection dataset and evaluation criteria, the core issues of the current development of pedestrian detection are analyzed. Finally, the next possible development direction of pedestrian detection technology is explained at the end of the paper.

A solution to improve the FAST feature point detection algorithm susceptible to image noise interference is proposed. Eight directions are derived from the selected point to solve the gray level mutations in these eight directions, and based on this, it is judged whether the point is a feature point. And take the maximum value suppression for the dense area of feature points, and solve the maximum value of the sum of gray changes in all directions as the judgment condition to screen the feature points. A new feature point matching scheme is proposed, where the Hamming distance between feature points within the target image is derived by the BruteForce algorithm. If the Hamming distance between the first few target points and the original feature point is less than a certain threshold, it will be substituted into the original image to solve the problem. The Hamming distance between the feature points of the original image is selected as the correct matching point. Finally, the principle of region growth is introduced to the MC algorithm, which saves the amount of calculation and improves the generation accuracy.

With the development of additive manufacturing technology, the traditional STL model stratification method cannot gradually meet the requirements of increasing accuracy and printing speed. Aiming at the problems of large amount of data in STL model , slow stratification speed and contour distortion under complex surface , this paper proposes a partition adaptive stratification algorithm based on surface curvature and cross-sectional area of implicit surface , and further algorithm optimization is carried out through the marching square algorithm. At the same time, a linear time sorting algorithm and a two-way offset joint strategy are used to solve the problem of offset error of surface contour reconstruction after slicing, so as to achieve the purpose of multi-level precision adaptive stratification algorithm.

Vehicle-to-everything (V2X) communications can be applied in emergency material scheduling due to their performance in collecting and transmitting disaster-related data in real time. The urgency of disaster depots can be judged based on the disaster area video, and the scenario coefficient can be evaluated for building a fairness model. This paper presents a scenario-based approach for emergency material scheduling (SEMS) using V2X communications. We propose a SEMS model, with the objectives of minimum time and maximum fairness in the cases of multiple supply depots, disaster depots, commodities and transport modes for logistics management of relief commodities. We design the SEMS algorithm based on the artificial fish-swarm algorithm to obtain an optimized solution. The results demonstrate that the SEMS model can enhance the fairness of relief scheduling, especially for disaster depots with small demands compared to the Gini and enhanced Theil fairness models. Moreover, the acquired vehicle speed via V2X communications updates the SEMS model in real time, which approaches a solution closer to reality.

Urban traffic accidents are on the rise worldwide. In order to reduce the incidence rate of dangerous traffic events on road, it is important to train drivers using appropriate traffic scenarios. The objective of this paper is to develop the categories of hazardous traffic events that can be used to construct the simulated hazardous traffic scenarios. A survey questionnaire on drivers' perceptions of hazardous traffic events is distributed to 340 randomly selected drivers. It collects information on the drivers' perceptions of hazardous traffic events' frequency of happening degree of damage and drivers' stress levels triggered by such events. The survey responses are analyzed using fuzzy mathematics methods. Each hazardous traffic event includes the contributing factors of motor vehicles, non-motor vehicles, or pedestrians. The findings from this paper can be used to develop hazardous traffic scenarios to train drives and to provide theoretical basis for further developing and refining active risk avoidance systems.