^{1}

^{2}

^{1}

^{2}

Internet of Things devices are popular in civilian and military applications, including smart device cities, smart grids, smart pipelines, and medical Internet of Things. Among them, carsharing supported by the Internet of Things is developing rapidly due to their advantages in environmental protection and reducing traffic congestion. The optimization of the carsharing system needs to consider the uncertainty of demand and the coupling relationship of multiple decision variables, which brings difficulties to the establishment of mathematical models and the design of efficient algorithms. Existing studies about carsharing optimization are mainly divided into four subproblems: the operation mode selection, vehicle type selection, demand analysis, or decision-making, rather than comprehensive consideration. This paper summarizes the four subproblems from the perspective of mathematical models, solving algorithms, and statistical methods and provides references for more comprehensive research in the future.

In 1999, the Massachusetts Institute of Technology defined the Internet of Things: connecting all items to the Internet through information sensing devices such as radio frequency identification. The Internet of Things was widely used in smart device cities, smart grids, smart pipelines, and medical Internet of Things. The Internet of Things used the Internet as a cornerstone of further expansion and development. With the help of GPS, infrared sensor, and other sensing devices, it transmitted and exchanged information between different mobile digital devices, namely, different entities. It had three characteristics: (1) intelligent sensing, (2) two-way transmission, and (3) intelligent control. There had been several proposals for unique object identifiers that uniquely identified objects and locations in the real world. Information could be associated with objects and places, and decoding could be used to retrieve relevant information. Karakostas [

Structure of carsharing system.

Carsharing emerged informally as a consequence of gasoline prices rising in the 1940s [

The research on optimal design and operation of carsharing was divided into four subproblems, as shown in Figure

Classification of the research on optimal design and operation of carsharing.

The operation modes of the carsharing were mainly divided into three types according to the terminal location: round trip, one-way trip, and free floating. Round trip and one-way trip were based on stations that differed from free floating [

The driving routes of operation modes.

Brendel et al. [

There were two main types of vehicles according to the power system: green energy vehicles (GEVs) and gasoline vehicles (GVs). GEVs could better reflect the main characteristic of carsharing, which was environmental protection. The most studied and introduced type of GEVs among the studies was the group of electric vehicles (EVs). Further discussion about the GEVs would mainly focus on EVs. The charging time of EVs was long, the travel distance was limited, and the investment could be enormous (the cost of charging station and charging facility). Normally, GEVs were more suitable for a round trip and a one-way trip. GVs were just the opposite, so they were more suitable for free-floating mode. It was the main research direction for operators to choose which type of vehicles was more economical or whether operators that had adopted GVs in the early stage should introduce GEVs.

The social costs of electric vehicles and conventional vehicles could be a standard to determine which way was more profitable, and we should also consider the air pollution costs and the noise costs of conventional vehicles [

The demand in a certain area was partly affected by many territorial factors, such as population density, education levels, age, and private car park rate [

This subsection was to state how to predict the demand for a new region to provide advice for operators. When we make the demand prediction, various scenarios needed to be considered, including the travel distance, the number of travelers, the ratio of public transportation users, and the ratio of households without cars [

In the early years, scholars studied key indicators of whether or not the carsharing system could be successfully introduced [

This subsection was to state how to consider the uncertain demand in space and time.

The uncertain demand was a hot spot. Scholars before generally studied the linear elastic demand function [

Due to so many factors affecting the demand, the model might not be realistic considering the limited factors. Therefore, some scholars used the neural network and support vector machine based on historical data to predict demand. The support vector machine could accurately predict the demand by selecting the appropriate kernel function. Cheu et al. [

The business strategy could be classified into the following three decision level: strategic decision, tactical decision, and operational decision. The strategic decision determined the station location and capacity (the number of parking spaces). The tactical decision determined the vehicle supply and the number of operators, while operational decision determined the relocation scheme and how to price by time slots or distance. Actually, the established model should consider the three levels simultaneously for the strong interaction among them, but the model would be too large and cannot be used in cities with large demand. To address this problem, Boyacı et al. [

The places with more parking spaces, longer business hours, and higher population density had more booking demands and higher turnover rates, and they provided a basis for station selection [

The carsharing system could be expressed as a hybrid queuing network model which took the road congestion into account in the optimization model to solve parking capacities and fleet size [

Relocation of vehicles which belonged to the operational stage could be operated in two ways, including operator-based location and user-based location [

For the operator-based location problems, scholars might adopt two-stage optimization or three-stage optimization to reduce the size of the model which could reduce the scheduling cost and shorten the time of solving simultaneously [

In addition, the regions were divided into blocks according to the demand or peak and nonpeak periods according to time, and different pricing strategies were adopted for different blocks to minimize the relocation operation [

In the past fifteen years, research on carsharing has become more and more plentiful. Scholars have optimized and simulated the carsharing system through mathematical models and advanced algorithms, which provide a theoretical basis for future research. Future research can be based on the following aspects:

In the modeling solution, literature introduces the assumptions for the simplified model, and the solution results deviate from the actual situation. For example, (1) there are few documents that consider the situation of delaying or canceling travel after the customer has made a reservation and the flexible choice of destination choice for customers; (2) charging time and level are ignored mostly.

The model size of relocation in the carsharing system increases greatly with the increase of the number of stations and the demand for carsharing services. Therefore, it is urgent to develop an efficient algorithm or learn from mature fields such as refined oil scheduling and shared bikes scheduling.

The model should be closer to the real-world instance and integrate the three decisions under uncertain demands, which is rarely considered in the existing research.

Most researches lead providers to choose large cities with high demand. However, from a social perspective, the underdeveloped areas with lower demand should also enjoy the same convenience of sharing cars. Future research can focus on the operation of low-demand areas under the support of the government.

Yuxuan Wang and Huixia Feng are co-first authors of the paper.

The authors declare that they do not have any commercial or associative interest that represents a conflict of interest in connection with the work submitted.

Yuxuan Wang wrote the initial paper. Huixia Feng provided the overall idea of the paper and revised it. All authors contributed to the final paper.