A spatio‐temporal ensemble method for large‐scale traffic state prediction

The estimation of present passenger flow state plays a vital role in the urban rail transit (URT) operation process and it is the basis of passenger flow control and train dispatching. Most of the existing researches used the Automatic Fare Collection (AFC) System data as the single data source for estimating present passenger flow state. In consideration of the delay in the data uploading process, the obtained estimation results based on the single AFC data source are not accurate enough, which could lead to the improper handling and decision failure and cause high risk and low efficiency for the operation and management of URT. To address problems mentioned above, this paper adopted the automatic differentiation method as the framework for the fusion of multi-source heterogeneous data (including uploaded AFC data, the mobile phone signaling data provided by mobile phones, and the historical passenger flow data, etc.). Then, this paper proposed the online estimation model and the error layered optimization algorithm to implement the estimation of present passenger flow state. The calculation results indicate that the proposed model and algorithm can obtain reasonable and reliable URT passenger flow state. Finally, the study developed a present passenger flow state estimation system using the proposed model and algorithm. The system has been deployed in Chengdu Metro that serves about 3 million passengers per day from October 2018 to April 2020, with the overall error within 3% as of now.

Section: Introductionmentioning

confidence: 99%

Online estimation model for passenger flow state in urban rail transit using multi‐source data

Tao¹,

Tang

Hou

2021

“…In addition, the aforementioned methods are proposed to address the data missing problem for stationary detectors, which typically have low deployment rates. In this study, the global positioning system (GPS)‐based vehicle trajectories are used to extract network‐wide traffic states, due to their increasing availability (Bie, Xiong, Yan, Qu, 2020; He, Zheng, Chen, & Guan, 2017; He, Qi, Lu, & Chen, 2019; Liu, Liu, Vu, & Lyu, 2019). However, the vehicle trajectories are typically opportunistic and sparse because of the low penetration rate of GPS vehicles in the road network.…”

Section: Introductionmentioning

confidence: 99%

A generative adversarial network for travel times imputation using trajectory data

Zhang

Zheng

et al. 2020

Knowledge of travel times serves an important role in traffic control and management. As an increasingly popular data source, vehicle trajectories can provide large‐scale travel time information. However, real‐world travel time information extracted from sparse or low‐resolution trajectory data often contains missing data that need to be imputed for further traffic analysis. Thus, this study proposes a travel times imputation generative adversarial network (TTI‐GAN) for travel times imputation. Considering the network‐wide spatiotemporal correlations, the TTI‐GAN can generate travel times for links without sufficient observations by modeling travel time distributions (TTDs) for links with rich data. Then, numerical experiments are carried out with trajectory data from Didi Chuxing. The results show that the TTI‐GAN can well estimate link TTDs and performs better than other counterparts in imputing mean travel times under various data missing rates.

“…NNs are a collection of models that consist of multiple connected layers, and they can automatically learn hidden patterns from given samples and make further predictions. NN‐based models have been widely applied in intelligent transportation systems (Chen, Leng, & Labi, 2019; Dharia & Adeli, 2003; Hashemi & Abdelghany, 2018; Jiang & Adeli, 2005; Liu, Liu, Vu, & Lyu, 2020). In the field of OD flow estimation, several shallow NNs (NNs with few layers) have been adopted, including an artificial NN (Padinjarapat & Mathew, 2013), Hopfield NN (Gong, 1998), and time‐delayed NN (Zhao et al., 2017).…”

Section: Introduction and Reviewmentioning

confidence: 99%

Dynamic origin‐destination flow estimation using automatic vehicle identification data: A 3D convolutional neural network approach

Tang

Cao

Chen

et al. 2020

Dynamic origin-destination (OD) flow estimation is one of the most fundamental problems in traffic engineering. Despite numerous existing studies, the OD flow estimation problem remains challenging, as there is large dimensional difference between the unknown values to be estimated and the known traffic observations. To meet the needs of active traffic management and control, accurate time-dependent OD flows are required to understand time-of-day traffic flow patterns. In this work, we propose a three-dimensional (3D) convolution-based deep neural network, "Res3D," to learn the high-dimensional correlations between local traffic patterns presented by automatic vehicle identification observations and OD flows. In this paper, a practical framework combining simulation-based model training and few-shot transfer learning is introduced to enhance the applicability of the proposed model, as continuously observing OD flows could be expensive. The proposed model is extensively tested based on a realistic road network, and the results show that for significant OD flows, the relative errors are stable around 5%, outperforming several other models, including prevalent neural networks as well as existing estimation models. Meanwhile, corrupted and out-of-distribution samples are generated as real-world samples to validate Res3D's transferability, and the results indicated a 60% improvement with few-shot transfer learning. Therefore, this proposed framework could help to bridge the gaps between traffic simulations and empirical cases.