A New Speed Limit Recognition Methodology Based on Ensemble Learning:Hardware Validation

Advanced DriverAssistance Systems(ADAS)technologies can assist drivers or be part of automatic driving systems to support the driving process and improve the level of safety and comfort on the road.Traffic Sign Recognition System(TSRS)is one of themost important components ofADAS.Among the challengeswith TSRS is being able to recognize road signs with the highest accuracy and the shortest processing time.Accordingly,this paper introduces a new real time methodology recognizing Speed Limit Signs based on a trio of developed modules.Firstly,the Speed Limit Detection(SLD)module uses the Haar Cascade technique to generate a new SL detector in order to localize SL signs within captured frames.Secondly,the Speed Limit Classification(SLC)module,featuring machine learning classifiers alongside a newly developed model called DeepSL,harnesses the power of a CNN architecture to extract intricate features from speed limit sign images,ensuring efficient and precise recognition.In addition,a new Speed Limit Classifiers Fusion(SLCF)module has been developed by combining trained ML classifiers and the DeepSL model by using the Dempster-Shafer theory of belief functions and ensemble learning’s voting technique.Through rigorous software and hardware validation processes,the proposedmethodology has achieved highly significant F1 scores of 99.98%and 99.96%for DS theory and the votingmethod,respectively.Furthermore,a prototype encompassing all components demonstrates outstanding reliability and efficacy,with processing times of 150 ms for the Raspberry Pi board and 81.5 ms for the Nano Jetson board,marking a significant advancement in TSRS technology.

Highway Foreign Body Intrusion Detection System Based on Deep Learning

This paper introduces the expressway intrusion detection system based on deep learning to improve traffic safety.The system adopts deep learning,image recognition,and foreign body detection technology to monitor the road condition in real-time through lidar and binocular camera groups to detect and distance the foreign body on the road.The system visualizes the detection results on the onboard screen to assist the driver to avoid and improve the safety of highway driving.In addition,the system also includes emergency braking,blind spot monitoring,lane departure warning,and other functions.The system has wide application prospects and development potential and is expected to be widely used in the future,providing a strong guarantee for the safe operation of expressways in China.

Estimation of Vehicle Speed Based on Wheel Speeds from ASR System in Four-Wheel Drive Vehicles

Three major methods currently in the use of determining vehicle speed based on wheel speeds, the minimum wheel speed, minimum wheel speed corrected by slope method and the Kalman filter method, are analyzed, with merits and defects of each approach stated. Through simulations, the Kalman filter method based on minimum wheel speed shows improved accuracy, in addition to better adaptivity to vehicle reference speed. It also can be used to acceleration ship regulation （ASR） in part-time four-wheel drive vehicles.

Track Tension Analysis of Four-Wheel Drive Tracked Vehicles

The distribution of track tension on track link is complex when the tracked vehicles run at a high speed.A multi-drive track link structure,which changes the traditional induction wheel into the driving wheel was proposed.The mathematical model of the system was established and the distribution of track tension was studied.The combined simulation model of RecurDyn and Simulink of the structure with multi-drive track was established.The simulation results show that our proposed structure has more uniform tension distribution than traditional structures,especially under the high speed condition.The maximum tension can be reduced by 28 kN-36 kN and the transmission efficiency can be improved by10%-16% under high speed condition with this new structure.

Deep Learning Control for Autonomous Robot

Several applications of machine learning and artificial intelligence,have acquired importance and come to the fore as a result of recent advances and improvements in these approaches.Autonomous cars are one such application.This is expected to have a significant and revolutionary influence on society.Integration with smart cities,new infrastructure and urban planning with sophisticated cyber-security are some of the current ramifications of self-driving automobiles.The autonomous automobile,often known as selfdriving systems or driverless vehicles,is a vehicle that can perceive its surroundings and navigate predetermined routes without human involvement.Cars are on the verge of evolving into autonomous robots,thanks to significant breakthroughs in artificial intelligence and related technologies,and this will have a wide range of socio-economic implications.However,in order for these automobiles to become a reality,they must be endowed with the perception and cognition necessary to deal with high-pressure real-life events and make proper judgments and take appropriate action.The majority of self-driving car technologies are based on computer systems that automate vehicle control parts.From forward-collision warning and antilock brakes to lane-keeping and adaptive drive control,to fully automated driving,these technological components have a wide range of capabilities.A self-driving car combines a wide range of sensors,actuators,and cameras.Recent researches on computer vision and deep learning are used to control autonomous driving systems.For self-driving automobiles,lane-keeping is crucial.This study presents a deep learning approach to obtain the proper steering angle to maintain the robot in the lane.We propose an advanced control for a selfdriving robot by using two controllers simultaneously.Convolutional neural networks(CNNs)are employed,to predict the car’and a proportionalintegral-derivative(PID)controller is designed for speed and steering control.This study uses a Raspberry PI based camera to control the robot car.

A comprehensive framework for assessing the spatial drivers of flood disasters using an Optimal Parameter-based Geographical Detector-machine learning coupled model

Flood disasters pose serious threats to human life and property worldwide.Exploring the spatial drivers of flood disasters on a macroscopic scale is of great significance for mitigating their impacts.This study proposes a comprehensive framework for integrating driving-factor optimization and interpretability,while considering spatial heterogeneity.In this framework,the Optimal Parameter-based Geographic Detector(OPGD),Recursive Feature Estimation(RFE),and Light Gradient Boosting Machine(LGBM)models were utilized to construct the OPGD–RFE–LGBM coupled model to identify the essential driving factors and simulate the spatial distribution of flood disasters.The SHapley Additive ExPlanation(SHAP)interpreter was employed to quantitatively explain the driving mechanisms behind the spatial distribution of flood disasters.Yunnan Province,a typical mountainous and plateau area in Southwest China,was selected to implement the proposed framework and conduct a case study.For this purpose,a flood disaster inventory of 7332 historical events was prepared,and 22 potential driving factors related to precipitation,surface environment,and human activity were initially selected.Results revealed that flood disasters in Yunnan Province exhibit high spatial heterogeneity,with geomorphic zoning accounting for 66.1%of the spatial variation in historical flood disasters.The OPGD–RFE–LGBM coupled model offers clear advantages over a single LGBM in identifying essential driving factors and quantitatively analyzing their impacts.Moreover,the simulation performance shows a slight improvement(a 6%average decrease in RMSE and an average increase of 1%in R2)even with reduced factor data.Factor explanatory analysis indicated that the combination of the essential driving factor sets varied across different subregions;nevertheless,precipitation-related factors,such as precipitation intensity index(SDII),wet days(R10MM),and 5-day maximum precipitation(RX5day),were the main driving factors controlling flood disasters.This study provides a quantitative analytical framework for the spatial drivers of flood disasters at large scales with significant heterogeneity,offering a reference for disaster management authorities in developing macro-strategies for disaster prevention.

Fault Detection for Motor Drive Control System of Industrial Robots Using CNN-LSTM-based Observers

The complex working conditions and nonlinear characteristics of the motor drive control system of industrial robots make it difficult to detect faults.In this paper,a deep learning-based observer,which combines the convolutional neural network(CNN)and the long short-term memory network(LSTM),is employed to approximate the nonlinear driving control system.CNN layers are introduced to extract dynamic features of the data,whereas LSTM layers perform time-sequential prediction of the target system.In terms of application,normal samples are fed into the observer to build an offline prediction model for the target system.The trained CNN-LSTM-based observer is then deployed along with the target system to estimate the system outputs.Online fault detection can be realized by analyzing the residuals.Finally,an application of the proposed fault detection method to a brushless DC motor drive system is given to verify the effectiveness of the proposed scheme.Simulation results indicate the impressive fault detection capability of the presented method for driving control systems of industrial robots.

The Effectiveness of Driver Education and Information Programs in the State of Nevada

According to National Highway Traffic Safety Administration, pedestrian and driver crashes are increasing at an alarming rate due to technological advancements and human errors. There is a need to improve existing driver education programs to mitigate the chances of crashes. The objectives of this research were 1) to examine the quality of Nevada’s driver education by evaluating the effectiveness of its programs, and 2) to provide recommendations to improve driving education in Nevada based on the results from this study. Two different surveys were conducted in Clark County, Southern Nevada. The first survey focused on assessing the strengths and limitations of the current Driver Education Programs in Nevada by capturing the opinions and attitudes of those who went through the process as teenagers. The second survey focused on driver safety through the involvement of pedestrians on the road. These surveys and the corresponding statistical analysis as well as the exiting literature have provided insights to improve driving education. The corresponding recommendations were organized into seven major categories: 1) lack of rigor of online driver education, 2) interactive learning and technology, 3) follow-up exams, 4) practice/training at home, 5) collecting information about crashes, 6) pedestrians, and 7) additional emphasis. Finally, due to the dangers of driving distractions (texting and calling on the cell phone) and impairments (driving under the influence of alcohol or drugs), more emphasis on these topics—as well as more public announcements through billboards, television commercials, and magazines— can help to constantly remind drivers about having good driving habits.

Research Based on Innovation and Entrepreneurship Driven by "Four Wheels" and Combination of Professional Education and Teaching-- Taking the " Specialty of Computer Application Technology in Hunan Ap-plied Technology University" as an Example

Implementing innovation and entrepreneurship education by combining with professional education in universities and colleges is an important measure to promote higher-quality employment and entrepreneurship of the graduates. The problems existing in the fusing teaching of computer application technology and innovation and entrepreneurship education are analyzed in this paper. By taking Hunan Applied Technology University as an example and in view of the existing problems, the mode of reform driven by "four wheels","professional talent training scheme by integrating optimization, innovation and entrepreneurship","implementing the specific teaching by integrating imovation,entrepreneurship and professional education","building many forms and university-enterprise cooperation platforms for innovation and entrepreneurship" and "setting up reasonable management and incentive mechanism for teachers and students" are proposed, to realize the dynamic integration of professional education and innovation and entrepreneurship education for the specialty of computer application technology.

A Coaching Program for Recently Licensed Young Drivers in the Netherlands： Which Drivers Are Attracted？

In line with European developments, a Dutch second phase coaching program, referred to as the DX- （Driver Xperience） program, was developed for young novice drivers to counteract their high accident risk. More specifically, the aim of the DX-program was to enable young drivers to make responsible decisions and develop positive attitudes regarding four levels of the driving task： combining life style and driving, planning and navigation, participating in different traffic situations and handling the vehicle. In this paper, the design principles of the program are described. The empirical study focused on the entry characteristics of the participating young drivers （n = 3,117） as compared to a reference group of young drivers （n = 345）. Results show that the DX-program attracted young drivers that, in some respects, showed a more risky profile than average young drivers in terms of speed violations, anger and the number of fines. In addition, four groups of participants with sharply differing driving styles could be distinguished. Implications for educational design and follow-up research are discussed within the theoretical framework of self-regulated learning.

改进行为克隆与DDPG的无人驾驶决策模型

无人驾驶技术的关键是决策层根据感知环节输入信息做出准确指令。强化学习和模仿学习比传统规则更适用于复杂场景。但以行为克隆为代表的模仿学习存在复合误差问题,使用优先经验回放算法对行为克隆进行改进,提升模型对演示数据集的拟合能力;原DDPG(deep deterministic policy gradient)算法存在探索效率低下问题,使用经验池分离以及随机网络蒸馏技术(random network distillation,RND)对DDPG算法进行改进,提升DDPG算法训练效率。使用改进后的算法进行联合训练,减少DDPG训练前期的无用探索。通过TORCS(the open racing car simulator)仿真平台验证,实验结果表明该方法在相同的训练次数内,能够探索出更稳定的道路保持、速度保持和避障能力。

以学习操作系统构建数字化转型的数智动能

伴随着数据要素价值的凸显,如何以教育数据的有效治理,充分发挥数智驱动教育变革的强大潜力,成为教育数字化转型的关键所在。在智能教育大脑的隐喻下,学习技术系统依托于充当“数智大脑”角色的核心构件,搭建“数据组织—数据建模—数据分析”的教育数据治理通路,深度挖掘并最大化释放教育数据价值。基于此,文章以数据为主线重塑新一代学习技术系统框架,并以学习操作系统作为核心构件的隐喻,从“为何”“是何”及“如何”三个方面深度阐释其来源、内涵与体系架构,聚焦于数智动能的系统实现。同时,基于团队研发的“数智大脑”平台,文章以案例故事的形式描绘其在学校教育中的多元化使用场景,展现出以学习操作系统构建的数智动能的强大应用潜力,以期为数智驱动教育数字化转型提供全新的视角和思路。

基于交叉注意力的多任务交通场景检测模型

感知是自动驾驶的基础和关键,但大多数单个模型无法同时完成交通目标、可行驶区域和车道线等多项检测任务。提出一种基于交叉注意力的多任务交通场景检测模型,可以同时检测交通目标、可行驶区域和车道线。使用编解码网络提取初始特征,利用混合空洞卷积对初始特征进行强化,并通过交叉注意力模块得到分割和检测特征图。在分割特征图上进行语义分割,在检测特征图上进行目标检测。实验结果表明:在具有挑战性的BDD100K数据集中,所提模型在任务精度和总体计算效率方面优于其他多任务模型。

数字孪生辅助联邦学习中的边缘选择和资源分配联合优化

在基于联邦学习的智能驾驶中,智能网联汽车(ICV)的资源限制和可能出现的设备故障会导致联邦学习训练精度下降、时延和能耗增加等问题。为此该文提出数字孪生辅助联邦学习中的边缘选择和资源分配优化方案。该方案首先提出数字孪生辅助联邦学习机制,使得ICV能够选择在本地或利用其数字孪生体参与联邦学习。其次,通过构建数字孪生辅助联邦学习的计算和通信模型,建立以最小化累积训练时延和能耗为目标的边缘选择和资源分配联合优化问题,并将其转化为部分可观测的马尔可夫决策过程。最后,提出基于多智能体参数化Q网络(MPDQN)的边缘选择和资源分配算法,用于学习近似最优的边缘选择和资源分配策略,以实现联邦学习累积时延和能耗最小化。仿真结果表明,所提算法在保证模型精度的同时,有效降低联邦学习累积训练时延和能耗。

基于数据驱动的范数最优迭代学习控制

在系统模型确定的前提下,传统的范数最优迭代学习控制(NOILC)可以有效提高伺服系统的跟踪精度。但是在实际控制过程中,系统模型参数往往是变化的,从而导致控制器性能的下降。为此,提出了一种基于数据驱动的范数最优迭代学习控制方法。以系统的输入输出为依据,建立系统估计模型的代价函数,对代价函数进行偏微分处理,得到一种基于数据驱动的非参数模型辨识方法,最后将此模型辨识方法和NOILC相结合。实验结果表明:针对时变系统,此控制方法的跟踪误差为NOILC(Norm optimal iterative learning control,NOILC)的57.1%,并且相比NOILC提前5次收敛。因此,提出的方法能有效改善时变系统的跟踪性能。

基于迁移学习和深度学习的驾驶员分心行为识别研究

为了解决传统驾驶员分心行为识别模型准确率过度依赖大样本数据集、耗时较长等问题,提出了一种结合迁移学习策略和卷积神经网络模型的方法来对驾驶员分心行为进行识别。首先在模型中引入ImageNet数据集上训练好的网络权重,冻结网络的卷积层;然后去掉原网络中的全连接层,重新添加输出维度为10的FC层;最后在验证集上对比基于迁移学习策略模型与原网络模型的识别精度。结果表明,基于迁移学习策略的分心行为识别模型比原网络模型的平均准确率提升了约4%,显著提高了分心行为的识别率。本研究结果可为驾驶员分心行为识别提供理论与技术支持。

PBL联合DOL教学模式在皮肤科临床实习教学中的应用

目的:该研究对PBL与DOL联合教学运用于皮肤科临床实习带教方面展开探究。方法:将50名临床实习生通过随机方式平均分为观察组和对照组,观察组展开PBL联合DOL教学,对照组展开传统教学。借助学生满意度调查、mini-CEX以及认知考试成绩等对两组教学效果进行评估。结果:观察组实习生的理论知识考试得分高于对照组(P<0.05)。根据mini-CEX评价结果,观察组实习生的临床综合、组织效率、临床诊断、沟通咨询、面谈技能能力优于对照组(P<0.05)。学生问卷调查显示,观察组实习生在增加团队合作能力、临床技能提升、主动学习刺激、咨询技能提高、总体满意度方面优于对照组(P<0.05)。结论:PBL联合DOL教学有利于提高实习生临床实践能力和综合能力,可为皮肤科临床实习教学模式的改进提供参考。

基于CNN-LSTM混合驱动的焊接成形质量监测

焊接成形质量监测对于现代制造业至关重要,现有的质量识别方法大多基于单一传感器,识别精度难以进一步提升,面对复杂条件下的抗干扰能力较弱.针对单一传感器识别技术存在的不足,多源信息融合技术能够发挥不同类型传感器的自身优势,实现对焊接过程更为全面且准确的监测.在进行多信息融合过程中,深度学习模型的特征挖掘机制仍然欠缺解释,不同信息的互补性仍未明晰,为此,提出一种基于多源信息混合驱动的CNN-LSTM焊接成形质量监测模型.结果表明,通过融合图像和电压信号实现了99.72%的平均识别准确率,可视化结果还展示了不同信息之间的互补优势.

考虑驾驶风格的混合动力汽车强化学习能量管理策略

为了提升混合动力汽车能量管理策略对不同风格驾驶员的适应性,基于深度强化学习和等效燃油消耗最小策略(equivalent consumption minimization strategy,ECMS),提出一种考虑驾驶风格的混合动力汽车能量管理策略。通过实车试验采集驾驶员行驶数据,基于采集数据进行驾驶员驾驶风格的聚类分析,建立驾驶风格识别模型;构建基于强化学习和ECMS的能量管理策略,将驾驶风格系数作为强化学习状态变量,利用多种驾驶风格的组合工况训练深度确定性策略梯度智能体,获取不同工况和驾驶风格下ECMS等效因子,采用ECMS求解最优发动机、电机转矩分配以及变速箱挡位;搭建硬件在环测试平台,并基于实际采集的不同驾驶员驾驶数据构建测试工况,验证所提出控制策略的有效性。研究结果表明,相较于基于规则策略、基于等效因子比例修正的自适应ECMS以及DRL-SAC策略,提出的考虑驾驶风格的强化学习能量管理策略使整车能量消耗分别降低16.35%、11.11%和7.56%,所提控制策略的有效性得到了验证。

自动驾驶目标检测不确定性估计方法综述

随着自动驾驶技术的发展,目标检测的准确性和可靠性变得至关重要。深度学习作为自动驾驶系统中的核心组成部分,其预测结果的不确定性估计对于系统的安全性和稳定性具有显著影响。总结了深度学习不确定性估计理论在自动驾驶目标检测中的应用,并探讨了有效的不确定性评价体系的重要性。介绍了深度学习不确定性估计的基本理论,包括贝叶斯神经网络、蒙特卡洛方法以及集成学习方法等。这些方法通过不同的途径量化模型预测的不确定性,为自动驾驶系统提供了更丰富的信息。深入探讨了自动驾驶目标检测中不确定性估计的应用。通过案例分析,展示了如何利用不确定性信息来提高目标检测的准确性,特别是在面对复杂环境和极端条件时,不确定性估计可以作为决策支持,帮助系统避免潜在的风险。总结了自动驾驶目标检测不确定性估计评价指标,同时,考虑了模型的预测性能、不确定性估计的准确性。