基于梯度关联规则的老年行人交通事故风险识别

为构建老年行人交通事故严重程度风险关联因素识别方法体系，本文应用极限梯度提升关联规则挖掘算法(Extreme Gradient Boost-Apriori，XGB-Apriori)识别城市道路老年行人交通事故风险因子。运用机器学习优化关联规则算法结构，通过机器学习库 scikit-learn 中 XGBoost (Extreme Gradient Boost)算法与SFM(Select From Model)特征选择类功能实现变量特征值的选择。进而，对Apriori算法设置有序定向约束，得到适用于交通事故致因分析的数据挖掘技术。通过逐层迭代识别关联项，选取频繁项集，总结高置信度、高提升度的关联规则。关联因素模型评估结果表明，本文采用的SFM功能准确度可达78.31%，关联规则XGB-Apriori算法较传统算法精度提升了91%。挖掘结果显示，驾驶员与行人的自身特征、车辆特征、碰撞状态以及道路特征均对老年行人交通事故的严重程度具有重要影响。其中，男性驾驶员造成的行人死亡事故频次较高，女性驾驶员造成的受伤事故频次较高；大型、重型车辆(SUV、卡车、施工车)发生死亡事故频次相对小轿车更高；位于匝道等道路线型弯曲的坡道中，老年行人发生致死交通事故的频次相对线型缓和路段更高。本文对老年行人交通事故耦合因素全面识别并针对性提出风险防控精准预判方法，为有效保护道路弱势群体提供必要的理论支持。     

基于分段学习模型的自动驾驶行为决策算法研究

在具有车道线的特定自动驾驶场景中,针对目前端到端的行为决策算法直接输入原始图像进行决策导致的网络模型迁移性差、预测精度欠佳、泛化能力不足等问题,提出一种基于分段学习模型的车辆自动驾驶行为决策算法。首先,基于GoogLeNet建立一种端到端的车道线检测网络模型,并引入车道中心线作为决策的重要线索提高算法的迁移能力,同时利用YOLOv3目标检测模型对本车道内前方最近障碍物进行位置检测;而后,经几何测量模型将两者检测结果转换成环境状态信息向量为决策做支撑;最后,构建基于长短期记忆(LSTM)网络的驾驶行为决策模型,根据编码的历史状态信息刻画出动态环境中车辆的运动模式,并结合当前时刻的状态推理得到驾驶行为参量。使用建立的真实驾驶场景数据集对模型分别进行训练、验证与测试,离线测试结果显示车道线检测模型的检测位置误差小于1.3%,车道内前方障碍物检测模型的检测精度达98%以上,驾驶行为决策网络模型表征预测优度的决定系数 大于0.7。为进一步验证算法的有效性,搭建了Simulink/PreScan联合仿真平台,多种工况下的仿真验证试验中多个评价指标均达到工程精度要求,实车测试的试验结果也表明该算法可实现复杂驾驶场景下平稳、准确无偏航的预测效果并满足实时性要求,且与传统端到端模式的算法相比,具有更好的迁移性和泛化能力。     

A hybrid heuristic strategy for liquefied natural gas inventory routing

As liquefied natural gas (LNG) steadily grows to be a common mode for commercializing natural gas, LNG supply chain optimization is becoming a key technology for gas companies to maintain competitiveness. This paper develops methods for improving the solutions for a previously stated form of an LNG inventory routing problem (LNG-IRP). Motivated by the poor performance of a Dantzig-Wolfe-based decomposition approach for exact solutions, we develop a suite of advanced heuristic techniques and propose a hybrid heuristic strategy aiming to achieve improved solutions in shorter computational time. The heuristics include two phases: the advanced construction phase is based on a rolling time algorithm and a greedy randomized adaptive search procedure (GRASP); and the solution improvement phase is a series of novel MIP-based neighborhood search techniques. The proposed algorithms are evaluated based on a set of realistic large-scale instances seen in recent literature. Extensive computational results indicate that the hybrid heuristic strategy is able to obtain optimal or near optimal feasible solutions substantially faster than commercial optimization software and also the previously proposed heuristic methods.     

Dynamic airspace sectorisation for flight-centric operations

Today’s air traffic operations follow the paradigm of ‘flow follows structure’, which already limits the operational efficiency and punctuality of current air traffic movements. Therefore, we introduce the dynamic airspace sectorisation and consequently change this paradigm to the more appropriate approach of ‘structure follows flow’. The dynamic airspace sectorisation allows an efficient allocation of scarce resources considering operational, economic and ecological constraints in both nominal and variable air traffic conditions. Our approach clusters traffic patterns and uses evolutionary algorithms for optimisation of the airspace, focusing on high capacity utilisation through flexible use of airspace, appropriate distribution of task load for air traffic controllers and fast adaptation to changed operational constraints. We thereby offer a solution for handling non-convex airspace boundaries and provide a proof of concept using current operational airspace structures and enabling a flight-centric air traffic management. We are confident that our developed dynamic airspace sectorisation significantly contributes to the challenges of future airspace by providing appropriate structures for future 4D aircraft trajectories taking into account various operational aspects of air traffic such as temporally restricted areas, limited capacities, zones of convective weather or urban air mobility. Dynamic sectorisation is a key enabling technology in the achievement of the ambitious goals of Single European Sky and Flightpath 2050 through a reduction in coordination efforts, efficient resource allocation, reduced aircraft emissions, fewer detours, and minimisation of air traffic delays.     

A context aware system for driving style evaluation by an ensemble learning on smartphone sensors data

There are many systems to evaluate driving style based on smartphone sensors without enough awareness from the context. To cover this gap, we propose a new system namely CADSE system to consider the effects of traffic levels and car types on driving evaluation. CADSE system includes three subsystems to calibrate smartphone, to classify the maneuvers, and to evaluate driving styles. For each maneuver, the smartphone sensors data are gathered in three successive time intervals referred as pre-maneuver, in-maneuver, and post-maneuver times. Then, we extract some important mathematical and experimental features from these data. Afterwards, we propose an ensemble learning method on these features to classify the maneuvers. This ensemble method includes decision tree, support vector machine, multi-layer perceptron, and k-nearest neighbors. Finally, we develop a rule-based fuzzy inference system to integrate the outputs of these algorithms and to recognize dangerous and safe maneuvers. CADSE saves this result in driver’s profile to consider more for dangerous driving recognition. The experimental results show that accuracy, precision, recall, and F-measure of CADSE system are greater than 94%, 92%, 92%, and 93%, respectively that prove the system efficiency.     

基于分层次教学法的学习动力激发 ——以汽配专业《汽车电工电子技术基础》教学为例

在“互联网+”时代的到来,以及全民教育、终身学习的思想在全社会的推广,高职院校课堂教学重点应适应时代发展,向“激发学生内在学习动力,实现终身学习”的方向转变。根据教学经验,制定出针对不同层次学生的分层次教学方法,运用在日常教学活动中,最终通过该方法完成教学任务,实现教学目标。     

“工程力学”课程教学中加强高职学生工程能力和兴趣培养的探索

概述了工程力学性质,对学生现状进行了分析,提出了从引入工程案例、制作力学模型、苦练教师内功等三个方面来加强高职学生工程能力和兴趣的培养。     

人机混驾环境下混行车辆雾模型研究

目前搭载高级驾驶辅助系统和车联网（Vehicular Ad Hoc Network，VANET）技术的智能网联车（Intelligent Connected Vehicles，ICV）正大量涌入人工驾驶车（Manual Vehicle，MV）流之中，ICV与MV共存的异构车辆混行交通态势逐步形成，异构车辆之间的交互产生壁垒。混行之下单个ICV虽可依托单车硬件传感与单车计算单元实现与MV的交互意图识别，但其受有限算力与有限传感的影响，资源负载增大，时效性与安全性方面存在一定的误差与风险，而混行之下的VANET技术也不能够提供全局性车路资源用以高度匹配ICV与MV的交互场景，而且越来越多的ICV计算需求也在激增VANET的负载压力。对此，结合边缘计算概念中的雾计算理论，提出混行车辆雾模型（Mixed Vehicle Fog，MVF），充分发挥车联网络边缘节点能力，通过合理整合调度ICV资源的方法，解决对MV正常交互意图计算的时效性与安全性问题。该模型首先通过各感知单元响应混行交通环境下ICV与MV的正常交互事件，然后利用基于容错节点分簇的资源调度算法（Fault-tolerant Node Clustering Resource Scheduling Algorithm，FNC-RSA），动态划分局部路段内对交互事件具有相关意图感知与计算需求的ICV为一组"协同雾群"，再评估雾内ICV节点自身资源与路由代价，定向定量调度资源，最终实现雾群内部MV交互信息共享与驾驶意图协同计算。试验借助Prescan和MATLAB搭建联合仿真平台，与低能耗自适应分簇型路由算法（Low Energy Adaptive Clustering Hierarchy，LEACH）模型对比，验证MVF模型的运行效率与模型鲁棒性。研究结果表明：MVF模型通过交互事件细分协同雾群，保证了计算负载均衡，提高了ICV定向资源计算与传输效率，比LEACH模型降低了55.17%的平均跳数，缩短了45.40%的平均任务完成时间，抗时延干扰能力强，鲁棒性能优异。该模型对于打破混行环境异构车辆交互壁垒，提高混行道路交通行车安全，创造车联网络良性发展空间具有积极作用。     

基于交通大数据的移动模式分析综述

交通大数据可为揭示交通主体显性出行行为背后的深层规律（即移动模式）提供重要基础。精确掌握大数据驱动下交通主体的移动模式，可为需求预测、客流组织、土地利用、事件管理等应用提供理论依据。交通大数据主体繁多、时空多态、关联性复杂的特性迫使小数据时代下的移动模式分析方法转型和升级，但仍可能遇到移动模式一致性表达难、异常类型检测难、复杂关联性表达难、时空多态性建模难、一体可视化分析难等普遍问题。针对这些问题，利用科学知识图谱，对2010~2020年期间3 747篇文献的热点关键词分布、发表趋势分布、出版刊物分布等特点进行归纳总结，结合常见的移动模式分析数据集，系统综述现有研究在移动模式常态分析、非常态分析、关联分析、预测分析和可视化分析方向上的阶段性进展。其中，移动模式常态分析综述了个体活动特性分析、出行类别划分、特定群体分析等应用的研究进展。移动模式非常态分析综述了基于模板匹配和数据驱动的非常态事件检测方法的发展脉络。移动模式关联分析综述了面向不同数据源的关联特性检测方法的发展近况。移动模式预测分析综述了基于数理统计和数据驱动的交通属性预测方法的发展状况。移动模式可视化分析综述了用户交互设计、移动模式宏观可视化、微观可视化和整体可视化的发展近况。最后，系统总结各个分支方向面临的主要问题与挑战，并从数据整合、模型创新、机制变革等角度提炼移动模式分析研究的未来发展趋势，期望为后续研究采用新理论技术开展移动模式分析提供参考。     

Solving a discrete multimodal transportation network design problem

This paper investigates the multimodal network design problem (MMNDP) that optimizes the auto network expansion scheme and bus network design scheme in an integrated manner. The problem is formulated as a single-level mathematical program with complementarity constraints (MPCC). The decision variables, including the expanded capacity of auto links, the layout of bus routes, the fare levels and the route frequencies, are transformed into multiple sets of binary variables. The layout of transit routes is explicitly modeled using an alternative approach by introducing a set of complementarity constraints. The congestion interaction among different travel modes is captured by an asymmetric multimodal user equilibrium problem (MUE). An active-set algorithm is employed to deal with the MPCC, by sequentially solving a relaxed MMNDP and a scheme updating problem. Numerical tests on nine-node and Sioux Falls networks are performed to demonstrate the proposed model and algorithm.