Methods to reduce dimensionality and identify candidate solutions in multi-objective signal timing problems

Adjusting traffic signal timings is a practical way for agencies to manage urban traffic without the need for significant infrastructure investments. Signal timings are generally selected to minimize the total control delay vehicles experience at an intersection, particularly when the intersection is isolated or undersaturated. However, in practice, there are many other potential objectives that might be considered in signal timing design, including: total passenger delay, pedestrian delays, delay inequity among competing movements, total number of stopping maneuvers, among others. These objectives do not tend to share the same relationships with signal timing plans and some of these objectives may be in direct conflict. The research proposes the use of a new multi-objective optimization (MOO) visualization technique—the mosaic plot—to easily quantify and identify significant tradeoffs between competing objectives using the set of Pareto optimal solutions that are normally provided by MOO algorithms. Using this tool, methods are also proposed to identify and remove potentially redundant or unnecessary objectives that do not have any significant tradeoffs with others in an effort to reduce problem dimensionality. Since MOO procedures will still be needed if more than one objective remains and MOO algorithms generally provide a set of candidate solutions instead of a single final solution, two methods are proposed to rank the set of Pareto optimal solutions based on how well they balance between the competing objectives to provide a final recommendation. These methods rely on converting the objectives to dimensionless values based on the optimal value for each specific objectives, which allows for direct comparison between and weighting of each. The proposed methods are demonstrated using a simple numerical example of an undersaturated intersection where all objectives can be analytically obtained. However, they can be readily applied to other signal timing problems where objectives can be obtained using simulation outputs to help identify the signal timing plan that provides the most reasonable tradeoff between competing objectives.     

Forecasting current and next trip purpose with social media data and Google Places

Trip purpose is crucial to travel behavior modeling and travel demand estimation for transportation planning and investment decisions. However, the spatial-temporal complexity of human activities makes the prediction of trip purpose a challenging problem. This research, an extension of work by Ermagun et al. (2017) and Meng et al. (2017), addresses the problem of predicting both current and next trip purposes with both Google Places and social media data. First, this paper implements a new approach to match points of interest (POIs) from the Google Places API with historical Twitter data. Therefore, the popularity of each POI can be obtained. Additionally, a Bayesian neural network (BNN) is employed to model the trip dependence on each individual’s daily trip chain and infer the trip purpose. Compared with traditional models, it is found that Google Places and Twitter information can greatly improve the overall accuracy of prediction for certain activities, including “EatOut”, “Personal”, “Recreation” and “Shopping”, but not for “Education” and “Transportation”. In addition, trip duration is found to be an important factor in inferring activity/trip purposes. Further, to address the computational challenge in the BNN, an elastic net is implemented for feature selection before the classification task. Our research can lead to three types of possible applications: activity-based travel demand modeling, survey labeling assistance, and online recommendations.     

Driving cycle prediction model based on bus route features

Bus fuel economy is deeply influenced by the driving cycles, which vary for different route conditions. Buses optimized for a standard driving cycle are not necessarily suitable for actual driving conditions, and, therefore, it is critical to predict the driving cycles based on the route conditions. To conveniently predict representative driving cycles of special bus routes, this paper proposed a prediction model based on bus route features, which supports bus optimization. The relations between 27 inter-station characteristics and bus fuel economy were analyzed. According to the analysis, five inter-station route characteristics were abstracted to represent the bus route features, and four inter-station driving characteristics were abstracted to represent the driving cycle features between bus stations. Inter-station driving characteristic equations were established based on the multiple linear regression, reflecting the linear relationships between the five inter-station route characteristics and the four inter-station driving characteristics. Using kinematic segment classification, a basic driving cycle database was established, including 4704 different transmission matrices. Based on the inter-station driving characteristic equations and the basic driving cycle database, the driving cycle prediction model was developed, generating drive cycles by the iterative Markov chain for the assigned bus lines. The model was finally validated by more than 2 years of acquired data. The experimental results show that the predicted driving cycle is consistent with the historical average velocity profile, and the prediction similarity is 78.69%. The proposed model can be an effective way for the driving cycle prediction of bus routes.     

A hybrid deep learning based traffic flow prediction method and its understanding

Deep neural networks (DNNs) have recently demonstrated the capability to predict traffic flow with big data. While existing DNN models can provide better performance than shallow models, it is still an open issue of making full use of spatial-temporal characteristics of the traffic flow to improve their performance. In addition, our understanding of them on traffic data remains limited. This paper proposes a DNN based traffic flow prediction model (DNN-BTF) to improve the prediction accuracy. The DNN-BTF model makes full use of weekly/daily periodicity and spatial-temporal characteristics of traffic flow. Inspired by recent work in machine learning, an attention based model was introduced that automatically learns to determine the importance of past traffic flow. The convolutional neural network was also used to mine the spatial features and the recurrent neural network to mine the temporal features of traffic flow. We also showed through visualization how DNN-BTF model understands traffic flow data and presents a challenge to conventional thinking about neural networks in the transportation field that neural networks is purely a “black-box” model. Data from open-access database PeMS was used to validate the proposed DNN-BTF model on a long-term horizon prediction task. Experimental results demonstrated that our method outperforms the state-of-the-art approaches.     

An optimal data-splitting algorithm for aircraft scheduling on a single runway to maximize throughput

In this research, we present a data-splitting algorithm to optimally solve the aircraft sequencing problem (ASP) on a single runway under both segregated and mixed-mode of operation. This problem is formulated as a 0–1 mixed-integer program (MIP), taking into account several realistic constraints, including safety separation standards, wide time-windows, and constrained position shifting, with the objective of maximizing the total throughput. Varied scenarios of large scale realistic instances of this problem, which is NP-hard in general, are computationally difficult to solve with the direct use of commercial solver as well as existing state-of-the-art dynamic programming method. The design of the algorithm is based on a recently introduced data-splitting algorithm which uses the divide-and-conquer paradigm, wherein the given set of flights is divided into several disjoint subsets, each of which is optimized using 0–1 MIP while ensuring the optimality of the entire set. Computational results show that the difficult instances can be solved in real-time and the solution is efficient in comparison to the commercial solver and dynamic programming, using both sequential, as well as parallel, implementation of this pleasingly parallel algorithm.     

客车捎带模式下高铁快捷货物输送方案研究

以提高高铁快运当日达产品的时效性、收益率为核心,对既有载客动车组捎带模式下的快捷货物输送方案进行优化。借助时空网络以列车运行成本与时间惩罚费用之和最小为目标,同时满足货主时限、列车容量以及列车停站方案等约束,建立输送方案优化模型,通过匈牙利算法,并借助Matlab的Yalmip工具箱求解模型。以兰州西站至天水南站、宝鸡南站及西安北站部分时间段的快捷货物运输需求为背景进行算例分析,验证模型的有效性。结果表明合理估算列车装载容量及货物的延迟时限对输送方案的选择起重要作用。     

技术站间货物列车协同配流方案研究

技术站间货物列车协同作业组织模式，可实现各站获益，整体加强，对提升铁路运输生产效率具有重要意义.本文建立技术站间货物列车协同配流模型.模型以最大化两技术站的正点出发列车数作为目标函数，采用启发式遗传算法进行寻优，得到货物列车解编顺序和配流方案.最后，通过对算例进行实验分析，验证协同配流模型的实用性.结果表明，技术站间协同配流作业明显压缩车辆在站总停留时间，增加了阶段计划内正点出发列车数，进而提高了技术站内线路使用能力.     

考虑收益管理的高铁平行车次动态差别定价

基于RP(Revealed Preference)和SP(Stated Preference)调查数据，利用潜在类别模型对高铁旅客进行细分，得到旅客对平行车次不同服务属性，如列车运行时间、发车时段和舒适度的偏好程度，并对其进行量化；引入收益管理，以多列车整体收益最大为目标，构建平行车次动态差别定价模型，并设计模拟退火算法进行求解；最后，通过京沪高铁进行实例验证.结果表明：与固定票价进行客票销售相比，所提方案能够适应高峰期和平峰期不同客流特点，提高铁路客票总收益，为高铁平行车次灵活定价提供参考.     

基于深度强化学习的智能船舶航迹跟踪控制

[目的]智能船舶的航迹跟踪控制问题往往面临着控制环境复杂、控制器稳定性不高以及大量的算法计算等问题。为实现对航迹跟踪的精准控制,提出一种引入深度强化学习技术的航向控制器。[方法]首先,结合视线(LOS)算法制导,以船舶的操纵特性和控制要求为基础,将航迹跟踪问题建模成马尔可夫决策过程,设计其状态空间、动作空间、奖励函数;然后,使用深度确定性策略梯度(DDPG)算法作为控制器的实现,采用离线学习方法对控制器进行训练;最后,将训练完成的控制器与BP-PID控制器进行对比研究,分析控制效果。[结果]仿真结果表明,设计的深度强化学习控制器可以从训练学习过程中快速收敛达到控制要求,训练后的网络与BP-PID控制器相比跟踪迅速,具有偏航误差小、舵角变化频率小等优点。[结论]研究成果可为智能船舶航迹跟踪控制提供参考。     

考虑空车调配协调的货物列车开行方案优化研究

为实现空车调配与货物列车开行方案协调优化，结合基本运行图架构与车流径路，构建货运时空服务拓展网络。考虑配空与装卸取送、集编发等环节的时间接续要求，节点与区段不对流空车要求，以重车流全程运送与空车配送等广义总费用最少为目标，建立整数规划弧路模型。针对既有算法设计局限性，结合重车或空车配空的时间接续要求，提出将不同的 k 短路重车流方案与空车配空方案相关联的改进可行解构造方法，设计混合差分进化求解算法。实例研究表明，考虑空车调配进行重车、空车流组织协调优化，能够减少空车走行费用，及时满足装车需求，有效保证作业车流配合中转车流集结编组及时挂线，提高方案可实施性。