Ran Wang, Hui Wang, Kun Zhu, Changyan Yi, Ping Wang, Dusit Niyato
©SHUTTERSTOCK/BLUE PLANET STUDIO
Electric vehicles (EVs) are one important enabler for the evolution of sustainable transportation sector. However, one major barrier to the widespread adoption of EVs is their limited battery capacity and the excessive time required to fully charge depleted batteries. Although techniques such as charging with a rapid charger and battery swapping may partially alleviate such concerns, the corresponding high capital investment may not be profitable for areas with low EV adoption rates. Moreover, the diversified charging demands of randomly distributed EVs, both spatially and temporally, restrict high-quality services from current fixed charging stations (FCSs). To resolve such challenges, mobile charging services (MCSs) have been investigated as a supplemental charging method for EVs, wherein energy replenishment is provided by mobile charging vehicles (MCVs). Under the framework of the Internet of EVs (IoEV), this article describes the concept of MCSs and investigates their charging merits under some distinct charging scenarios, e.g., real-time response service and emergency rescue service. A case study concerning real-time response services is investigated to illustrate the influence of MCSs on the current charging system. Simulation results demonstrate the ability of MCSs to improve the quality of charging service by reducing the average service time and service acceptance ratio. Finally, technical challenges, future research issues, and possible methodologies regarding the MCSs for EVs are discussed.
EVs are deemed an eco-friendly and efficient alternative to traditional internal combustion engine-based vehicles. However, the key barrier to the wide adoption of EVs is their limited battery capacity, which requires the frequent adoption of charging services [1]. Furthermore, depleted batteries typically require several hours to be fully charged, while the tanks in gasoline-driven vehicles can be filled up in just a few minutes. For example, the 90-kWh Tesla Model-S, with a range of 294 mi, takes approximately 10 h to charge from a NEMA 14–50 power outlet. Recent techniques, such as the adoption of fast chargers (e.g., the Tesla V3 supercharger) and battery swapping (e.g., the NIO battery-swapping station), may partially overcome these concerns [2]. However, the huge capital investments of FCSs and battery-swapping stations may not be profitable for areas with low EV adoption rates as a battery-swapping station typically costs roughly $500,000 [3]. In addition, due to mobility issues, randomly distributed EVs may have diversified charging demands in both space and time, reducing the quality of services of current FCSs. For example, EVs with charging demands may be located in districts far from FCSs, and hence have to travel long distances to obtain charging services, inhibiting travel efficiency. The abbreviations used in this article are listed in Table 1.
Table 1 The abbreviations used in this article.
Moreover, with the widespread adoption of EVs, the uncoordinated charging of EVs may overload the grid, exerting negative impacts on the security and stability of their operations, in turn causing voltage deviations, transformer and line saturations, excessive electrical losses, and so on. In addition, FCSs with a limited number of charging piles are unable to cope with sudden high charging demands. Elevated charging requirements may consequently result in the charging requests from EVs being denied or a marked increase in EV waiting times at the FCS [4].
MCVs, as a supplement charging infrastructure of FCSs, can provide a promising solution to the aforementioned challenges. MCVs have a self-contained energy storage system that is employed to replenish the energy of a certain number of EVs. Its removable, inexpensive, and easy-to-deploy properties facilitate charging services by proactively and instantly responding to charging requirements, particularly at the moment of potentially high grid pressure. Moreover, MCVs can be exploited to temporarily enlarge the capacities of FCSs and reduce EV waiting times during peak hours. Note that when an MCV is employed to charge EVs, the charging requests do not have a direct impact on the power grid. Thus, during peak hours, the pressure of the electrical power grid can be largely alleviated by dispatching a proportion of the EV charging requests to MCVs, where they can be charged during off-peak hours [5].
In this article, we present the concept of MCSs via charging EVs partially or totally by MCVs instead of only FCSs, and investigate several charging scenarios. The charging of EVs by MCVs can be categorized into four types: real-time response services, reservation services, emergency rescue services, and fixed-site services. We describe the specific scenarios of these services by investigating the interactive processes between the system entities and their benefits. A case study is presented on the real-time response service, focusing on factors that influence EV charging time and the quality of services under these scenarios. The simulation reveals that the average service time and miss (denied-request) ratio of services can be greatly reduced by the MCS under two baseline charging scheduling methods, indicating that the quality of charging services can be greatly improved via MCSs. The technical challenges and possible methodologies that can be investigated to address future research issues regarding the MCSs for EVs are subsequently summarized.
Much of the recent MCS-related literature concentrates on how to provide convenient charging services to EVs. Wang et al. [6] investigate a multiobjective mobile charging scheduling method to promote the quality of MCSs. The MCV’s scheduling problem is solved through the proposed deep reinforcement learning (DRL)-based framework. In Huang et al. [3], an MCS is proposed for an urban scenario, where a queuing-based framework is applied to employ distinct service disciplines to cope with several charging requests. Liu et al. [7] employ a federated learning-based placement decision method to help idle MCVs predict future charging positions. Then, idle MCVs decide their placements according to the predicted information. Further research applies the Lyapunov optimization theory to maximize MCV profits based on the randomness of EV users’ arrival and the dynamic power supply [8]. Zhang et al. [2] aim to present a novel mobile charging control mechanism by promoting charging reservations. Instead of focusing on the service discipline of EV selection, this article pays more attention to the scheduling of MCVs.
An additional branch of the literature investigates the structures of MCVs, which affect not only charging service quality but also the types of services that can be provided. Atmaja and Mirdanies [9] classify MCVs into two types according to their inclusion of energy storage, focusing on the distinct working modes of MCV types. MCVs with no energy storage are able to support FCSs, acting as mobile charging plugs, while MCVs with energy storage can be deployed as an emergency charging alternative for EVs. The authors in [10] design an MCV with a 2-MWh energy storage capacity and the ability to simultaneously charge five EVs with a maximum 75-kW charging rate for each vehicle. The article investigates the system’s performance under rapid charging requirements.
Further topics of discussion include routing or designing of depots in MCSs. Cui et al. [11] investigate the MCV’s location-routing problem with time windows as a mixed-integer linear program considering the running range of MCVs. Cui et al. [12] consider a similar routing problem, with a focus on two distinct service-efficiency MCVs. The authors in [13] divide the MCS optimization problem into two levels, i.e., planning and operation. At the operational level, a scenario-sampling-based online policy is adopted to operate MCVs. At the planning level, a simulation-based optimization framework is constructed to decide the number and locations of depots as well as the battery size of MCVs.
To provide reliable charging services to EVs and enhance energy efficiency of the power grid, the characteristics of key MCS components should be investigated. In this section, we introduce the components involved in the system’s operation. Our goal is to draw upon existing concepts and integrate them with MCSs to provide system-level concepts and architectures. Figure 1 depicts the relationships among different components in the MCS of the IoEV.
Figure 1 The relationship between each component in the MCS for the IoEV.
MCVs, typically vans with one or more plug-in chargers on board, can be divided into two categories, namely, those with and without energy storage [5]. MCVs without batteries are not capable of individually providing charging services; instead, they must be connected to the power grid while charging the EVs, acting as mobile charging plugs to support the FCS during peak times. In contrast, MCVs equipped with batteries can operate in three modes: 1) grid-connected, in which the MCVs charge their own batteries or provide ancillary services (which may include demand-side management, spinning reserve, peak shaving, valley filling, frequency regulation, and so on) to the power system by connecting to the upstream electrical grid; 2) service, whereby the MCV’s batteries are used as the power supply source to charge the EVs; and 3) idle, in which the MCVs are unoccupied and in-waiting for charging requests from the EVs. Here, we focus on MCVs with energy storage throughout the article, unless otherwise stated.
During the charging process of an EV by the MCV, the energy flows from the battery of MCV into that of the EV through a dc–dc converter. The converter is able to control the charging rate by regulating the charging process. To further satisfy the charging requirements under different scenarios, MCVs are divided into two types according to the number of charging modules equipped, namely, single- and multiple-port MCVs.
The depot is the public infrastructure that provides parking, charging, and battery-swapping services for MCVs. In addition, with appropriate regulations on the charging process of MCV batteries, the depot is capable of providing ancillary services (e.g., frequency regulation, valley filling, and peak shaving) for the power grid, with real-time responses to the demand of the grid or electricity prices in the market. MCV batteries can be replenished via two processes, namely, battery recharging and battery swapping, neither of which exhibit advantages over the other (battery swapping is generally less time consuming than recharging but needs more investment expenditure [3]).
The electricity consumer (EC) refers to the vehicle expecting to be charged. The key indicators for ECs include electricity price, waiting time, and charging rate. Furthermore, EVs can be deemed an electricity prosumer (EPs) as well as ECs if they are also interested in selling their redundant electricity to earn extra income. The relatively larger quantity and wider distribution of EPs allow for more flexible and efficient charging services to be provided in such IoEV networks.
The system controller is responsible for scheduling the entire processes of the charging services, including receiving the charging requests from EVs, dispatching the MCVs or the EPs, selecting appropriate charging locations, coordinating the information exchange between the MCV and depot, forecasting the charging demand, and so on [4]. The computation and communication resources that the system controller possesses should be powerful enough to support various requests from EVs, MCVs, depots, and a great diversity of prediction tasks. The system controller also helps to protect the safety and privacy of the exchanged information among the EVs, MCVs, depots, and so on. The entities involved in the charging service are only exposed to the information required to guarantee the quality of service. For example, the MCVs only need information such as the prenegotiated location, charging demand, and plate number of the EVs.
In this section, we introduce various types of charging service scenarios involving MCVs. The service modes can generally be divided into two groups: the moving and parking modes. MCVs in the moving mode have to migrate more frequently to track the serving targets (i.e., EVs) due to their frequent change in location. The services provided by MCVs under this mode include real-time response services, reservation services, and emergency rescue services. In the parking mode, MCVs are able to provide fixed-site services, whereby EVs are driven to the location of the MCVs to access the charging services. In the following, we provide detailed descriptions of the service scenarios.
Figure 2 presents a possible scenario of this service. In particular, an EC realizes that it does not have enough electric power to finish its trip. As there are no FCSs nearby to offer the charging service, the EC requests the real-time response service and sends the system controller the real-time charging request containing the information of its location and required amount of electricity. Based on this information, the controller will choose an SPCV to provide this service. If there are no free SPCVs, the controller may send the request to an EP, who then decides whether to accept it. Upon acceptance of the request by the SPCV or EP, the trading price (which may be different between the SPCV and EP) will be decided and sent to the EC. Once the EC accepts this price, the controller will compute the optimal charging location (e.g., a parking lot), and the charging location and registration identifiers of both sides will be sent to each party. Following the completion of the charging service, the EC pays for the service and the transaction ends.
Figure 2 The real-time response of the charging service. BS: base station.
Real-time response service requires the MCV or EC to head for the location selected by the system controller to complete the charging service. The EC cannot leave during the entire charging process. Some consumers would expect a type of reservation service in which they only need to park vehicles at a fixed place, and then the SPCV will head for its position to complete the charging at the appointed time. During the charging processes, consumers can do other processing such as handling computation tasks. Reserving a charge in advance and meeting the charging demands of some ECs ahead of time can partially relieve the real-time response pressure for the whole charging system.
The service targets of emergency rescue are generally EVs that cannot move normally. In such a situation, the EV waits for an SPCV in its current location after sending the charging request. Therefore, SPCVs of this service are generally smaller models with more flexible mobility. Moreover, emergency rescue services are generally required to be completed as soon as possible, and thus, the priority of these services is higher than those of the real-time response and reservation services. The high priority and high charging demand of emergency rescue services result in a relatively higher service difficulty than that of real-time response and reservation services, and hence, the price of this service is generally higher.
The service mode of fixed-site services is different from those of the previous services, whereby the MCV goes to EC’s location, or both move to the agreed-upon location. For the fixed-site service, the MCV will not move when the serving entity changes, but the controller will predict the future charging demand and subsequently dispatch the MCV to regions with the highest charging demand (e.g., FCSs during peak time, or parking lots with no FCSs nearby and many ECs). The MCV will move to the next location only when the charging demand has changed. As the MCV in charge of the fixed-site service will not move frequently and will deal with relatively more charging requests compared to the other services, the MPCVs are more likely to be selected to undertake this service. The system controller will announce the new location of the MPCV as well as its remaining free charging modules and the corresponding charging price. ECs can send requests with information about their required power to the controller, who then chooses whether to accept this request or not. Here, the request may be denied if the required power is too large or if too much time is required to complete the task. Once the request is accepted, the EC will move to the MPCV’s location to complete the charging service.
To analyze the impacts of MCSs on the current IoEV, we evaluate the real-time response service as an example. The pattern of the charging requests is the same as that in [14], whereby the authors create a charging demand forecasting model via data mining and data fusion. In this article, we focus on the results of the spatial-temporal distribution characteristics. A square service area of 60 km2 is defined [6] and separated into four equal parts: the residential area in the bottom left, commercial area in the bottom right, industrial area in the upper left, and public service area in the upper right. Different areas have different spatial-temporal distribution characteristics. A total of 500 charging requests are generated by EVs randomly, and the ratio of charging requests from different regions is set as 7:12:14:8. Each of these charging requests requires 20 kWh of electricity and will randomly arrive within 24 h. We set 10 min as the length of one time period, with 144 time periods per day. The software platform to implement the experiment is Matlab 9.11. Figure 3 presents the temporal distribution of the charging requests across the different areas.
Figure 3 The temporal distribution of charging requests in different areas.
The charging demand under the real-time response service can be satisfied via FCSs, SPCVs, and EPs. Each of these components will be added into the available charging facility set step by step. We locate two FCSs as the center of the commercial and public service areas. Each FCS has five dc chargers with a charging speed of 60 kW. Five SPCVs are deployed to offer the real-time response service. The battery charge capacity of the SPCVs is assumed to be 1 MWh with a charging speed of 60 kW. The SPCVs are initially located in the depot at the center of the service area. After the SPCV accepts a charging request, it will move to the planned location to complete the charging. The SPCV will not move until it accepts the next charging request unless it is powerless. The EP will also participate in this service and accept requests that cannot be completed by the FCS or SPCV. We assume that 100 EPs are willing to offer extra electricity to gain benefits. They are randomly distributed in the area with a charging speed of 10 kW. Following matching, the EP and EC will go to the planned area to complete the charging. The driving speed of all vehicles is assumed to be at the constant rate of 30 km/h. A time limit is also set, whereby the sum of the moving and waiting times cannot exceed 1 h or the request will be considered as unable to finish (EPs do not have such a limit due to their low charging speed).
To assess the impact of the MCS on different charging strategies, we introduce two metrics that measure how well the charging demand is satisfied. The first metric is the average service time, where the service time starts from the EC sending the charging request and ends when the request is met. The second metric is the miss ratio of service. The evaluation of average service time considers only the accepted requests, while the miss ratio represents the proportion of the unaccepted requests from all requests. In addition, we design two matching strategies for the SPCV: current-free matching (CFM), whereby we only match the closest free SPCV with the EC, and least-time matching (LTM), where we match the current busy SPCV to an EC as it may be closer to the EC.
Figure 4 presents the average charging service time by adopting different charging strategies. Initially, only the FCS is added to accept charging requests, with an average service time of 53.31 min. Following the inclusion of the SPCVs, the two strategies exhibit distinct characteristics. For CFM, the average service time declines to 50.71 min; however, for LTM, the average service time increases to 56.37 min. This is attributed to the dispatching of currently busy SPCVs with the LTM strategy, thus increasing the waiting time. When the EP is added, the average service time increases for both CFM and LTM, reaching 61.22 and 61.71 min, respectively. This is linked to the 10-kW charging speed of the EP, which significantly increases the average service time.
Figure 4 The average service time and miss ratio with different components.
With the initial inclusion of the FCS, the miss ratio is 0.45. The denied requests mainly originate from the industrial and residential areas where the FCSs are not set up. With the inclusion of the SPCVs, the miss ratio declines to 0.23 for CFM and 0.1560 for LTM. The average LTM service time is higher than that of CFM, hence, LTM can accept more requests. When the EPs are added, the miss ratio is reduced to 0.144 and 0.106.
The 53 denied requests for LTM are observed to occur during peak time, while all off-peak requests are satisfied. An analysis of the working conditions of the FCS chargers and SPCVs reveals the corresponding free ratios, which indicates the proportion of free time in the whole day, to be 0.618 and 0.2611, respectively. During peak time, the FCSs and SPCVs do not have any breaks, and thus, the number of SPCVs must increase to further reduce the miss ratio. As shown in Figure 5, if we increase the number of SPCVs to eight, the miss ratio for LTM will decline to 0.05. At this time, the SPCVs will serve 191 requests, while the number of requests accepted by the EPs will decrease to nine. These nine requests occur during peak time. Furthermore, the SPCV-free ratio increases to 0.3784. If we increase the number of SPCVs to 10, the miss ratio will reduce to 0.036, from which the SPCVs will serve 207 requests while the EP only serves four requests. In addition, the SPCV-free ratio increases to 0.5472. Thus, increasing the number of SPCVs can reduce the miss ratio. However, this will also eliminate off-peak requests for the EPs and considerably increase the SPCV-free ratio.
Figure 5 Miss and free ratios under different SPCV numbers.
The simulation results reveal that introducing SPCVs can result in the acceptance of more EC charging requests. This is particularly true for areas without FCSs, whereby the SPCVs are observed to almost replace the FCSs during off-peak hours. However, as with the FCSs, the SPCVs experience a lack of work during off-peak hours and cannot satisfy all peak-time requests. The introduction of EPs allows more requests to be satisfied during peak time.
In this section, we introduce open research issues, future research directions, and possible methodologies to handle the technical challenges concerning the MCSs in IoEV, as illustrated in Figure 6.
Figure 6 Open research issues and future research directions regarding the deployment of MCVs. UAV: unmanned aerial vehicle.
Service prices are important control signals that will impact electricity demand indirectly. EVs may make their individual decisions based on the posed prices of MCSs. Additionally, it is necessary to determine different pricing mechanisms for the aforementioned different types of services. Dynamic pricing schemes are desired to adjust prices according to the real-time conditions of MCSs. The pricing scheme usually includes two components, i.e., designing the pricing models and modeling the customers’ actions to the pricing schemes. To properly capture the interactions between EVs and MCVs, mathematic models including game theory, auctions, and matching theory are suitable and promising method frameworks.
Optimizing the dispatching of MCVs in the IoEV is usually a complicated task. A large variety of factors including the number of MCVs deployed, routing scheduling of MCVs, types of MCSs, and allocation of resources (energy, time, depot, and so on) should be comprehensively investigated for MCV dispatching optimization. In addition, different entities in the system (e.g., the service operator, MCVs, and ECs) may pursue different or even conflicting objectives, such as improving the quality of service, reducing the operation cost, and increasing the facilities’ utilization. Under such a case, a win-win MCV dispatching strategy and the associated multiobjective optimization scheme are desired to jointly balance the profits of different sides. Meanwhile, due to the variation of charging service demands as well as the dynamic system conditions, appropriate models are needed to capture the system’s state-transition uncertainties, and adaptive mechanisms have to be designed to achieve automatic and intelligent MCS management.
To address the aforementioned challenges, an adaptive and multiobjective MCV dispatching framework should be properly constructed. Based on a big data platform, the framework should be able to analyze and predict charging service performance, customer behaviors, and MCV dispatching strategies. Based on this, it should support dynamic routing optimizations, efficient resource allocations, service-level agreement guarantees, and other functions, providing a basis for the unified policy management module and gradually realizing intelligent and automatic MCV management. For example, DRL-based methods can be employed to optimize the possible scheduling of MCVs, where we can define states as current locations of ECs and MCVs, and actions as the next locations of SPCVs given the possibility to match with any EC that yields the maximum charging amount (or revenue) or profit (revenue−cost), and the reward can be the charging amount/revenue/profit. Multiobjective optimization, learning-based decision-making methods, learning-based prediction methods and online optimization methods, and so on are potential methodologies used to model and optimize MCV dispatching.
The depot plays an important role in the charging of MCVs, hence, the settings and management of the depot affect service efficiency of the MCS. Planning and deploying the depot constitute some challenges, requiring particular evaluations in terms of capacity design, power grid loading, capital investment, operation efficiency, and uncertainties in MCVs’ behaviors. Compared with residential charging, charging scheduling in the depot is more challenging as it is usually performed in the daytime, when electricity prices in the market changes more dynamically and MCVs’ charging demand pattern is more uncertain. Queuing theory, learning-based prediction methods, dynamic programming, stochastic programming, robust optimization methods, and so on are possible methodologies for the optimization of depots’ planning, deployment, and management.
Although the security and privacy issues in vehicular networks have been well investigated, the unique features of the IoEV system that support MCSs pose new technical challenges. For instance, when scheduling MCVs, security- and privacy-preserving payments are required to establish mutual authentication in a fast, efficient, and traceable manner. Moreover, EVs’ and MCVs’ identity, location, and state-of-charge information need to be preserved properly while enough information has to be provided to the system controller to efficiently manage the MCS. In addition, attack detection and risk evaluations are also important for operation security of MCSs. Under such a case, distributed security protection methods such as blockchain are more preferable than centralized methods as the number of entities involved in the MCS is huge. In addition, deep learning-based methods are desired to improve the accuracy of anomaly detection/prediction for MCSs.
In the MCS, EVs are considered the customers. In reality, more types of devices or entities may require MCSs, e.g., unmanned aerial vehicles and electric bicycles. Under such a case, new types of MCVs have to be designed according to the customer type served. Meanwhile, to overcome the issues associated with the widespread use of EVs, vehicle-to-vehicle wireless power transfer has received increasing attention [15]. Under this service mode, EVs’ charging no longer relies on the FCSs, nor requires the MCV to provide charging services for EVs at a fixed location. Such a platform can also outsource energy from other EVs without any network enhancement and is of significance in solving EV power supply problems. However, many limitations remain to be overcome, such as providing more stable power, protecting privacy, and optimizing the battery structure design. With the widespread employment of EVs, the importance of mobile charging will become more prominent and worth further studies.
The effort toward convenient and efficient charging systems for EVs has been the focus of increasing attention. In this article, MCSs were investigated and discussed, aiming to minimize the charging time of EVs and facilitate service diversifications. Several concepts and scenarios were explored to demonstrate the benefits and characteristics of MCSs. We presented a case study on the real-time response service to illustrate the impact of MCSs on current charging systems. Simulation results demonstrated the ability of MCSs to improve the quality of charging service, such as reducing the average charging time and improving the request-acceptance ratio. Finally, technical challenges and open research issues were highlighted to facilitate the development of MCSs.
This work was supported by National Natural Science Foundation of China under Grant 62171218; the National Research Foundation, Singapore, and Infocomm Media Development Authority under its Future Communications Research & Development Programme; DSO National Laboratories under the AI Singapore Programme (Award Number AISG2-RP-2020-019); Energy Research Test-Bed and Industry Partnership Funding Initiative; Energy Grid 2.0 programme; DesCartes and the Campus for Research Excellence and Technological Enterprise programme; and MOE Tier 1 (RG87/22). Kun Zhu is the corresponding author of this article.
Ran Wang (wang0686@e.ntu.edu.sg) is an associate professor in the College of Computer Science and Technology, Nanjing University of Aeronautics and Astronautics, Nanjing 211106, China. He received his B.E. degree in electronic and information engineering from Honors School, Harbin Institute of Technology, China, and his Ph.D. degree in computer science and engineering from Nanyang Technological University, Singapore. His current research interests include network performance analysis and the Internet of Electric Vehicles. He is a Member of IEEE.
Hui Wang (wanghui9727@nuaa.edu.cn) is currently pursuing his master’s degree with the College of Computer Science and Technology, Nanjing University of Aeronautics and Astronautics, Nanjing 211106, China. He received his B.E. degree from the College of Computer Science and Technology, Jinling Institute of Technology, Nanjing, China. His research interests include the Internet of Electric Vehicles and machine learning in smart grids.
Kun Zhu (zhukun@nuaa.edu.cn) is a professor in the College of Computer Science and Technology, Nanjing University of Aeronautics and Astronautics, Nanjing 211106, China. He received his Ph.D. degree from School of Computer Engineering, Nanyang Technological University, Singapore. He was a research fellow with the Wireless Communications Networks and Services Research Group at the University of Manitoba, Canada, from 2012 to 2015. He is also a Jiangsu specially appointed professor. His research interests include resource allocation in 5G and self-organizing networks.
Changyan Yi (changyan.yi@nuaa.edu.cn) is a professor in the College of Computer Science and Technology, Nanjing University of Aeronautics and Astronautics, Nanjing 211106, China. He received his Ph.D. degree from the Department of Electrical and Computer Engineering, University of Manitoba, Canada. From September 2018 to August 2019, he was a research associate at the University of Manitoba, Canada. His research interests include game theory and machine learning and their applications in various wireless networks. He is a Member of IEEE.
Ping Wang (ping.wang@lassonde.yorku.ca) is an associate professor in the Department of Electrical Engineering and Computer Science, York University, Toronto M3J 1P3, Canada and a Tier 2 York Research Chair. Prior to that, she was with Nanyang Technological University, Singapore. Her research interests are mainly in radio resource allocation, network design, performance analysis and optimization for wireless communication networks, mobile cloud computing, and the Internet of Things. She is a Fellow of IEEE and a Distinguished Lecturer of the IEEE Vehicular Technology Society.
Dusit Niyato (dniyato@ntu.edu.sg) is currently a professor in the School of Computer Science and Engineering at the Nanyang Technological University, 639798, Singapore. He received his Ph.D. degree in electrical and computer engineering from the University of Manitoba, Canada. He has published more than 400 technical papers in the area of wireless and mobile computing. He was a Distinguished Lecturer of the IEEE Communications Society from 2016 to 2017. He is a Fellow of IEEE.
[1] W. Ejaz et al., “IoV-based deployment and scheduling of charging infrastructure in intelligent transportation systems,” IEEE Sensors J., vol. 21, no. 14, pp. 15,504–15,514, Jul. 2021, doi: 10.1109/JSEN.2020.3006706.
[2] X. Zhang, Y. Cao, L. Peng, J. Li, N. Ahmad, and S. Yu, “Mobile charging as a service: A reservation-based approach,” IEEE Trans. Autom. Sci. Eng., vol. 17, no. 4, pp. 1976–1988, Oct. 2020, doi: 10.1109/TASE.2020.2983819.
[3] S. Huang, L. He, Y. Gu, K. Wood, and S. Benjaafar, “Design of a mobile charging service for electric vehicles in an urban environment,” IEEE Trans. Intell. Transp. Syst., vol. 16, no. 2, pp. 787–798, Apr. 2014, doi: 10.1109/TITS.2014.2341695.
[4] N. Chen, M. Wang, N. Zhang, and X. Shen, “Energy and information management of electric vehicular network: A survey,” IEEE Commun. Surveys Tuts., vol. 22, no. 2, pp. 967–997, 2nd Quart. 2020, doi: 10.1109/COMST.2020.2982118.
[5] U. Qureshi, A. Ghosh, and B. K. Panigrahi, “Scheduling and routing of mobile charging stations with stochastic travel times to service heterogeneous spatiotemporal electric vehicle charging requests with time windows,” IEEE Trans. Ind. Appl., vol. 58, no. 5, pp. 6546–6556, Sep./Oct. 2022, doi: 10.1109/TIA.2022.3182323.
[6] H. Wang, R. Wang, H. Xu, Z. Kun, C. Yi, and D. Niyato, “Multi-objective mobile charging scheduling on the Internet of Electric Vehicles: A DRL approach,” in Proc. IEEE Global Commun. Conf. (GLOBECOM), Madrid, Spain, 2021, pp. 1–6, doi: 10.1109/GLOBECOM46510.2021.9685354.
[7] L. Liu, Z. Xi, K. Zhu, R. Wang, and E. Hossain, “Mobile charging station placements in Internet of Electric Vehicles: A federated learning approach,” IEEE Trans. Intell. Transp. Syst., vol. 23, no. 12, pp. 24,561–24,577, Dec. 2022, doi: 10.1109/TITS.2022.3205596.
[8] H. Chen, Z. Su, Y. Hui, and H. Hui, “Dynamic charging optimization for mobile charging stations in Internet of Things,” IEEE Access, vol. 6, pp. 53,509–53,520, Sep. 2018, doi: 10.1109/ACCESS.2018.2868937.
[9] T. D. Atmaja and M. Mirdanies, “Electric vehicle mobile charging station dispatch algorithm,” Energy Procedia, vol. 68, pp. 326–335, Apr. 2015, doi: 10.1016/j.egypro.2015.03.263.
[10] M. O. Badawy et al., “Design and implementation of a 75-kW mobile charging system for electric vehicles,” IEEE Trans. Ind. Appl., vol. 52, no. 1, pp. 369–377, Jan./Feb. 2016, doi: 10.1109/TIA.2015.2469775.
[11] S. Cui, H. Zhao, and C. Zhang, “Multiple types of plug-in charging facilities’ location-routing problem with time windows for mobile charging vehicles,” Sustainability, vol. 10, no. 8, Aug. 2018, Art. no. 2855, doi: 10.3390/su10082855.
[12] S. Cui et al., “The multi-mode mobile charging service based on electric vehicle spatiotemporal distribution,” Energy, vol. 198, no. C, May 2020, Art. no. 117302, doi: 10.1016/j.energy.2020.117302.
[13] P. Tang, F. He, X. Lin, and M. Li, “Online-to-offline mobile charging system for electric vehicles: Strategic planning and online operation,” Transp. Res. D, Transp. Environ., vol. 87, Oct. 2020, Art. no. 102522, doi: 10.1016/j.trd.2020.102522.
[14] Q. Xing et al., “Charging demand forecasting model for electric vehicles based on online ride-hailing trip data,” IEEE Access, vol. 7, pp. 137,390–137,409, Sep. 2019, doi: 10.1109/ACCESS.2019.2940597.
[15] M. Abdolmaleki, N. Masoud, and Y. Yin, “Vehicle-to-vehicle wireless power transfer: Paving the way toward an electrified transportation system,” Transp. Res. C, Emerg. Technol., vol. 103, pp. 261–280, Jun. 2019, doi: 10.1016/j.trc.2019.04.008.
Digital Object Identifier 10.1109/MVT.2023.3289302