With the advancement of location acquisition technologies, a large amount of raw global positioning system (GPS) trajectory data is produced by many moving devices. Learning transportation modes from the GPS trajectory data is an important problem in the domain of trajectory data mining. Traditional supervised learning-based approaches rely heavily on data preprocessing and feature engineering, which require domain expertise and are time consuming. The authors propose a deep learning-based convolutional long short term memory (LSTM) model for transportation mode learning, in which the convolution neural network is first used to extract deep high-level features and then LSTM is used to learn the sequential patterns in the data that uses both GPS and weather features, thus making the full use of spatiotemporal operations. The authors have also analysed the impact of the geospatial region on human mobility. Experiments conducted on the Microsoft Geolife data set fused with the weather data set show that their model achieves the state-of-the-art results. The authors compare the performance of their model with the benchmark models, which shows the superiority of their model having 3% improvement in accuracy using only GPS features, and the accuracy is further improved by 4 and 7% on including the impact of geospatial region and weather attributes, respectively.

https://ietresearch.onlinelibrary.wiley.com/doi/epdf/10.1049/iet-its.2019.0017

Convolutional LSTM based transportation mode learning from raw GPS trajectories

This study addresses the tactical electric vehicle fleet size (EVFS) problem faced by carsharing service (CSS) providers while considering the operational vehicle assignment, vehicle relocation, and vehicle charging strategies (i.e., the charging duration at each station) in pursuit of profit maximization. To alleviate battery degradation and achieve cost-saving in the long term, we propose the on-demand charging strategy to determine fleet size. The novelty of this study lies in the incorporation of nonlinear battery wear cost incurred during the battery charging and discharging processes. A mixed-integer nonlinear programming (MINLP) model with concave and convex terms in the objective function is first developed for the EVFS problem. Piecewise linear approximation approach and outer-approximation method are employed to linearize the proposed model. Numerical experiments based on EVCARD, a one-way electric carsharing operator in China, are conducted to demonstrate the efficiency of the proposed model and solution method, as well as the necessity of incorporating the battery degradation into the fleet size determination of CSSs. The impacts of several key parameters, i.e., the daily fixed cost of EV and battery price, battery cycle efficiency, service charge, and relocation cost on the performance of one-way electric CSSs are also analyzed.

Electric vehicle fleet size for carsharing services considering on-demand charging strategy and battery degradation

Understanding the private car aggregation effect is conducive to a broad range of applications, from intelligent transportation management to urban planning. However, this work is challenging, especially on weekends, due to the inefficient representations of spatiotemporal features for such aggregation effect and the considerable randomness of private car mobility on weekends. In this article, we propose a deep learning framework for a spatiotemporal attention network (STANet) with a neural algorithm logic unit (NALU), the so-called STANet-NALU, to understand the dynamic aggregation effect of private cars on weekends. Specifically: 1) we design an improved kernel density estimator (KDE) by defining a log-cosh loss function to calculate the spatial distribution of the aggregation effect with guaranteed robustness and 2) we utilize the stay time of private cars as a temporal feature to represent the nonlinear temporal correlation of the aggregation effect. Next, we propose a spatiotemporal attention module that separately captures the dynamic spatial correlation and nonlinear temporal correlation of the private car aggregation effect, and then we design a gate control unit to fuse spatiotemporal features adaptively. Further, we establish the STANet-NALU structure, which provides the model with numerical extrapolation ability to generate promising prediction results of the private car aggregation effect on weekends. We conduct extensive experiments based on real-world private car trajectories data. The results reveal that the proposed STANet-NALU outperforms the well-known existing methods in terms of various metrics, including the mean absolute error (MAE), root mean square error (RMSE), Kullback–Leibler divergence (KL), and R2. 

Understanding Private Car Aggregation Effect via Spatio-Temporal Analysis of Trajectory Data

One-way carsharing is an alternative shared-use transportation mode that provides flexible travel accommodations for urban mobility. However, vehicle distributions can be mismatched with demand distributions because users are not required to return to their departure locations. Conventional operator-based vehicle relocation is limited by labor resources, but user-based and demand-controlled approaches can open new avenues for mitigating vehicle imbalance. This paper proposes a method for controlling demand patterns by applying measures of combinatorial monetary incentives and surcharges. A two-level nested logit model is adopted to analyze user decisions regarding the travel process in response to differentiated pricing combinations. A user choice model is aggregated and loaded into a time-space network that reveals the dynamics of the carsharing system. An optimization framework is proposed to determine the incentives and surcharges at different stations and times of day. This paper presents an algorithm for solving the proposed optimization model, as well as an example of parameter calibration and the solving process. Case analysis suggests that the proposed method can increase revenues by 22.5% compared to a scenario without demand control and vehicle relocation policies. Comparisons suggest that the proposed demand-based control policy can achieve higher revenues than operator-based relocation, whereas operator-based relocation could satisfy greater demand.

Demand control model with combinatorial incentives and surcharges for one-way carsharing operation

Metropolitan development has motivated car sharing into an attractive type of car leasing with the help of information technologies. In this paper, we propose a new approach based on deep learning techniques to assess the operation of a station-based car sharing system. First, we analyse the pick-up and drop-off operations of the station-based car sharing system, capturing the operational features of car sharing service and the behaviours of vehicle use from a temporal perspective. Then, we introduced an analytical system to detect the system operation concerning the spontaneous deviations derived from user demands from service provisions. We employed Long Short-Term Memory (LSTM) structure to forecast short-term future vehicle uses. An experimental case based on real-world data is reported to demonstrate the effectiveness of this approach. The results prove that the proposed structure generates high-quality predictions and the operation status derived from user demands.

/pdf/demand-management-of-station-based-car-sharing-system-based-10aarq792j.pdf

Demand Management of Station-Based Car Sharing System Based on Deep Learning Forecasting

On-demand one-way carsharing systems are increasingly gaining popularity nowadays and the market is growing unprecedentedly. However, operating techniques such as vehicle surveillance and fleet management fall behind the industrial development. Practice on EVCARD - an on-demand one-way electric vehicle carsharing system operating in Shanghai - raises an issue on dynamically predicting the user destinations, in order to support decisions on dynamic fleet management. This study presents a global positioning system (GPS)-data-driven method to solve the problem. The historical vehicle GPS data is enabled to match the user current trajectories and infer their possible destinations. Based on the GPS trajectory similarity measurement, the study presents a four-step procedure, including (i) similarity calculation, (ii) most-similar track detection, (iii) adjustment, and (iv) sorting. The method also takes the station correlations and the user historical destinations into account. A case study is given to demonstrate the dynamic prediction procedure of this method. Experiment on 96,821 valid test tracks shows that the positive prediction rate can be above 92% if the test trip has been completed over 70%. Factors that may influence the prediction result are additionally discussed, which include the existence of round-trips, the coverage of samples, and the alternatives of destinations.

https://ietresearch.onlinelibrary.wiley.com/doi/epdf/10.1049/iet-its.2018.5250

GPS-data-driven dynamic destination prediction for on-demand one-way carsharing system

The one-way carsharing system has been widely used in the carsharing field due to its flexibility. However, one of its main disadvantages is the imbalance between supply and pickup demand. At prese...

Yugao Zhong

Papers

GPS-data-driven dynamic destination prediction for on-demand one-way carsharing system

Multi-source data-driven prediction for the dynamic pickup demand of one-way carsharing systems