Japanese perspective of driver information systems

A perspective view of of Japanese R&D activities of Driver Information Systems is given from the standpoint of developing joint projects by public and private sectors. First, a brief history of the R&D activities is illustrated from above mentioned standpoint. Then, two major projects, AMTICS and RACS, and the social backgrounds of these projects are explained. In order to give a clear idea of both projects some technical details are treated.Based on the history and the present status of the developments, the basic design conceptions of Japanese Driver Information Systems are explained and several factors or reasons which have influenced to the design conception are discussed. Then, the promoting systems of the R&D activities are generally explained and present status of the projects and future problems in developing more advanced systems are also handled. A brief discussion on the establishment of international standard is also mentioned as a final comment.     

电动汽车大数据处理活动中的伦理问题研究

为了解车辆零部件运行状态,分析整车故障原因,技术人员通过车载终端把电动汽车的整车控制系统各节点发送至大数据平台进行查阅、分析和下载使用.大数据的使用引发了信息工程领域的伦理问题.本文重点阐述了电动汽车产业运用大数据时可能产生的伦理问题及其产生原因,探索强化责任、加强职业道德教育等解决办法,并结合伦理学理论提出了建议.     

缸内直喷的丰田卡罗拉1.2T怠速不稳解析

一辆2020款丰田卡罗拉1.2T汽车,行驶里程为1000KM,出现怠速不稳甚至熄火,经过专用诊断仪测试,汽车的高压燃油压力最高为16MPA,小于正常的20PMA,而且高压压力值随着怠速时间延长逐渐下降,甚至达到0.2MPA,启动时不容易着车,反复启动着车正常,高压压力无规律波动,更换高压油泵故障仍出现,高压油泵属于日本...     

基于改进GAN的端到端自动驾驶图像生成方法

基于端到端数据系统的自动驾驶系统对驾驶图像存在巨大需求。为解决一般生成式对抗网络模型在扩充驾驶图像数据集时不稳定及生成图像特征缺乏多样性的问题,研究1种改进网络模型LS-InfoGAN。结合最小二乘对抗损失防止模型梯度消失,并缓解生成器优化矛盾,提升模型训练稳定性。通过最大化生成图像与真实图像间的互信息提升生成器特征学习能力,改善生成图像特征多样性。利用转置卷积层还原图像特征,提升生成图像特征清晰度。以自主构建的模拟驾驶场景中获取的带标签驾驶图像集对模型有效性及其数据集扩充应用效果进行验证。实验分析表明：相比改进前模型,LS-InfoGAN模型的图像生成过程稳定性平均提升35%;使用此模型扩充的数据集进行端到端自动驾驶系统中决策网络的训练能在不采集新图像的情况下将系统决策性能提升1%~2%;建议使用此模型扩充图像数据集时将生成图像数量设置为原始训练集图像数量的1~2倍。

     

基于数据驱动的电动汽车动力电池SOC预测

为准确预测电动汽车动力电池的能耗,缓解驾驶者的里程焦虑,本文中提出一种基于数据驱动的电动汽车动力电池SOC预测模型.首先分析电动汽车能耗构成并提取能耗影响因素,接着基于某款电动出租车CAN总线采集的汽车运行数据,采用机器学习算法,提出基于温度分层的能耗模型,通过宏观数据与微观数据的融合减小误差,最后使用该模型对车载BM...     

大数据时代天津智慧城市智能交通建设与道路交通发展展望

通过对天津市交通强国建设、京津冀交通一体化及第三届世界智能大会的介绍,从交通运输管理系统、出行服务系统、电子付费系统和交通监控指挥系统智能化升级等方面阐述了天津市逐步进入智能交通应用阶段的情况。在此基础上,论述了5G时代,在“创新、协调、绿色、开放、共享”五大发展理念的引领下,城市道路在道路横断面布置、自行车道设置、共享单车停车点布设、步行交通系统多样化等方面创新设计的思路,展望了在未来万物互联的智能世界中,低碳生态城市和以公共交通为主导的城市综合交通系统的发展前景。     

基于有向图卷积神经网络的交通预测与拥堵管控

为解决城市快速路正面临的日益严重的交通拥堵问题,提出了一种针对城市快速路的基于有向图卷积神经网络的交通预测与拥堵管控方法,该方法能够有效利用海量交通数据进行交通预测,实现拥堵的主动管控.首先,基于交通路网的空间有向性和交通流的时空特性,定义了有向的距离影响矩阵、修正欧式距离矩阵和自由流可达矩阵,构建出有向的图卷积算子,...     

The future of road safety: A worldwide perspective

Estimates by the World Health Organization suggest that, on a yearly basis, road crashes kill 1.25 million people—nearly 3400 road fatalities per day—and injure up to 50 million. Traffic injuries are not equally spread over the world, however; some countries are hit harder than others, and the chance of being killed in a road crash depends on where one lives. Almost 90% of all traffic casualties occur in low- and middle-income countries (LMIC). Globally, the number of fatalities per 100,000 population (mortality rate) ranges from less than 3 to almost 40. The rate is less than 9 in high-income countries (HIC) but averages around 20 in LMIC, with the African region demonstrating the highest rate (26.6). While road safety trends have been positive in HIC over the last few decades, trends in LMIC are not telling a positive story: road fatalities are expected to increase to almost 2 million road fatalities per year by 2020.The United Nations has adopted several resolutions on road safety and proposes actions to tackle the global road safety crisis. Considering the current level of road safety to be unacceptable, the UN has taken several initiatives. One effort, the Decade of Action for Road Safety 2011–2020, has generated substantial activity around the world over the last couple of years. Furthermore, it is very encouraging that the UN included road safety in the Sustainable Development Goals that it laid out in September 2015. Road safety is part of the public health agenda and the urban development agenda. Measured in “real actions,” however, the responses so far from the overall global community and individual countries do not suggest that we are already on the right track to bringing down the death toll on roads.The future of road safety is uncertain and definitely not the same for all regions of the world. Countries with a mature road safety approach and an ambition to make further progress are expected to move in the direction of a pro-active approach: a Safe System approach. It is reported that many LMIC, meanwhile, are on the brink of designing road safety strategies and implementing action plans. The international community is willing to support LMIC, but LMIC cannot simply copy successful HIC strategies because local circumstances differ. The principles of successful HIC strategies are applicable, but the priorities and action plans should take root in and align with local conditions.     

基于SHEL和三角模糊数理论的地铁钻爆法施工安全评价方法研究

为合理应用决策数据中不同专家的背景知识,找出地铁施工过程中安全测评体系的关注重点,根据地铁施工的特点和SHEL安全模型理论,建立以人-软件(L-S)、人-硬件(L-H)、人-环境(L-E)和人-人(L-L)4个环节为基础的钻爆法地铁施工安全指标体系。请专家给出指标语义评价等级和指标权重语义评价等级,并根据其与三角模糊数的对应关系分别得出三角模糊数评价值。针对不同专家的重要性程度差异,提出基于三角模糊数的群一致性评价方法,得出专家综合权重,集结评价数据,最终得到安全评价结果。最后,对青岛地铁1号线过海段进行实例分析。结果表明,该方法可以解决群评价过程中专家个体权威与群体共识难以兼顾的问题,能够充分利用群决策中的原始评价信息,提高地铁施工安全水平评价的准确性。     

船舶能效数据清洗方法研究

船舶能效数据清洗对于建立准确的船舶能效模型,提高船舶能效计算和分析的准确度,指导船舶节能航行具有重要意义.对船舶能效数据的故障特征进行了分析,运用阈值理论、船舶航行关联理论对故障数据进行了识别,运用插值法和灰色关联理论方法对水深、对水航速以及主机油耗数据中的异常数据进行修正,并且根据数据错误特征和数据清洗方法制定了船舶能效数据清洗流程,以内河邮轮"凯娅号"船舶上安装的船舶能效数据采集系统采集到的能效数据为研究对象,对数据清洗结果进行了分析.结果表明,水深数据修正值与实测值平均误差为1.58%,油耗数据、对水航速数据修正值与实测值平均误差为2.8%,1.5%,误差较小,能够满足后续建模以及数据挖掘等工作.