Transport

Twice clustering based hybrid model for short-term passenger flow forecasting

Sheng Wang, Xinfeng Yang — Thu, 21 Nov 2024 15:59:15 +0200

Short-term metro passenger flow prediction plays a great role in traffic planning and management, and it is an important prerequisite for achieving intelligent transportation. So, a novel hybrid Support Vector Regression (SVR) model based on Twice Clustering (TC) is proposed for short-term metro passenger flow prediction. The training sets and test sets are generated by TC with respect to values of passenger flow in different time periods to improve the prediction accuracy. Furthermore, each obtained cluster is decomposed by using the Complete Ensemble Empirical Mode Decomposition with Adaptive Noise (CEEMDAN) algorithm and the Ensemble Empirical Mode Decomposition (EEMD) algorithm, respectively. The volatility of each component obtained after decomposition is further reduced. Then, the SVR model optimized by the Grey Wolf Optimization (GWO) algorithm is used to predict the decomposed components. Moreover, forecast based on one-month data from Xi’an Metro Line 2 Library Station (China). By comparing the prediction results of the TC condition, the Once Clustering (OC) condition and the non-clustering condition, it shows that the TC approach can adequately model the volatility and effectively improve the prediction accuracy. At the same time, experimental results show that the novel hybrid TC–CEEMDAN–GWO–SVR model has superior performance than Genetic Algorithm (GA) optimized SVR (SVR–GA) model and hybrid Back Propagation Neural Network (BPNN) model.

Starting driving style recognition of electric city bus based on deep learning and CAN data

Dengfeng Zhao, Zhijun Fu, Chaohui Liu, Junjian Hou, Shesen Dong, Yudong Zhong — Tue, 10 Dec 2024 09:18:26 +0200

Drivers with aggressive driving style driving electric city buses with rapid response and high acceleration performance characteristics are more prone to have traffic accidents in the starting stage. It is of great importance to accurately identify the drivers with aggressive driving style for preventing traffic accidents of city buses. In this article, a starting driving style recognition method of electric city bus is firstly proposed based on deep learning with in-vehicle Controller Area Network (CAN) bus data. The proposed model can automatically extract the deep spatiotemporal features of multi-channel time series data and achieve end-to-end data processing with higher accuracy and generalization ability. The sample data set of driving style is established by pre-processing the collected in-vehicle CAN bus data including the status of driving and vehicle motion, the data pre-processing method includes data cleaning, normalization and sample segmentation. Data set is labelled with subjective evaluation method. The starting driving style recognition method based on Convolutional Neural Network (CNN) model is constructed. Multiple sets of convolutional layers and pooling layers are used to automatically extract the spatiotemporal characteristics of starting driving style hidden in the data such as velocity and pedal position etc. The fully connected neural network and incentive function Softmax are applied to establish the relationship mapping between driving data characteristics and the starting driving styles, which are categorized as cautious, normal and aggressive. The results show that the proposed model can accurately recognize the starting driving style of electric city bus drivers with an accuracy of 98.3%. In addition, the impact of different model structures on model performance such as accuracy and F1 scores was discussed, and the performance of the proposed model was also compared with Support Vector Machine (SVM) and random forest model. The method can be used to accurately identify drivers with aggressive starting driving style and provide references for driver’s safety education, so as to prevent accidents at the starting stage of electric city bus and reduce crash accidents.

Optimal integrated location and dispatching decisions for feeder bus route design problem

Bo Sun, Ming Wei, Chunfeng Yang — Tue, 17 Dec 2024 08:21:41 +0200

Dispatch centres are an important part of the feeder bus network, and their location affects the design process of the feeder route. In some remote areas with weak transport infrastructure, it is very important to find an effective tool to simultaneously select the optimal location of the dispatch centre as well as transit routing process, which could improve the performance of the feeder bus system. The purpose of this article is to present an integrated optimization model for joint location and dispatching decisions for Feeder Bus Route Design (FBRD). The proposed methodology can select a number of best dispatch centres in alternative sets and calculate the order of the demand points visited by the feeder route. The objective of the model is to simultaneously minimize the total construction cost of selected dispatch centres and the total operational cost of the designed feeder bus system. The methodology facilitates obtaining solutions using the design of an improved double population Bacterial Foraging Optimization (BFO) algorithm. For example, it redefines the solution coding and the heuristic used to randomly initialize the initial population. When applied to the design of a feeder bus system for a station at Nanjing (China), the results reveal that a reduced budget may lead to change in the location of the dispatch centre; a more distant centre is required, which may increase the total mileage cost of all feeder routes. A detailed comparison of the improved and standard BFO and CPLEX shows that the difference between solutions is acceptable. However, the calculation time is greatly reduced, thus proving the effectiveness of the proposed algorithm.