1 UAV Virtualization for Enabling Heterogeneous and Persistent UAV-as-a-Service Nidhi Pathak, Student Member, IEEE, Sudip Misra, Senior Member, IEEE, Anandarup Mukherjee, Student Member, IEEE, Arijit Roy, Student Member, IEEE, and Albert Zomaya, Fellow, IEEE Abstract—In this paper, we propose an architecture for UAV virtualization with the primary aim to provide virtualized UAV services to multiple users by envisioning the concept of UAV-as- a-Service. In contrast to traditional UAVs, which are resource- constraint in nature and exhibit shorter flight times, our proposed UAV virtualization overcomes the limitations of short flight time of traditional UAVs, in turn allowing them to provide persistent and ubiquitous services. We achieve the virtualization of a UAV through multiple collaborating real-life UAVs performing various tasks in tandem. In this work, we focus on the placement and se- lection of UAVs to achieve virtualization. We use a social welfare- based approach to select suitable UAVs, from the available ones, and map the UAV to a virtual one. This work enables the provision of different UAV services to multiple end-users, without actual procurement of the UAVs by the end-users. We compare the results for optimal placement, normal maximum energy- based UAV selection, and Atkinson-based selection method. We also compare the virtual model and simple UAV-to-task model to physical UAV usage, task completion ratio, and residual energy of the system. Our proposed model outperforms the traditional model with a task completion efficiency of 94.26%. The residual energy of the system also increases with an increase in the number of tasks. Index Terms—Unmanned Aerial Vehicle, persistent service, virtualization, scheduling, task allocation, social welfare. I. I NTRODUCTION T HE rapid development of technology in the domain of UAVs, has led to the emergence of crucial application do- mains such as aerial monitoring, disaster management, search- and-rescue, surveillance, cargo deliveries, and aerial imagery and mapping. However, the limited resources of the UAVs, es- pecially its energy, severely restricts its utility. The low energy capacity and high energy consumption of these UAVs make it challenging to sustain a long UAV-flight duration to complete a task. Typically, a UAV mission is split into smaller sub- tasks and assigned to individual UAVs for completion of the mission, collaboratively [1]. The implementation of automated UAV-battery recharging stations ensures the completion of long duration missions by periodically recharging its batteries [2]. However, such solutions require prior setup of ground control stations, flight operators, and support teams. Due to the significant dependence of these UAVs on the ground control stations, the systems are mostly operated locally within the N. Pathak and A. Roy are with the Advanced Technology Development Center, Indian Institute of Technology Kharagpur, India S. Misra and A. Mukherjee are with the Department of Computer Science and Engineering, Indian Institute of Technology, Kharagpur, West Bengal, India A. Zomaya is with the School of Information Technologies, The University of Sydney, Australia UAV’s communication range. Additionally, these setups are not mobile and dynamic enough to support long-range UAV missions, which restricts the services to only the number of UAVs within the range of the ground setup. In this paper, we propose an architecture to provide persistent and ubiquitous UAV services to the end-users, unlike the traditional UAV services, which are intermittent and short-lived. To enable this proposed architecture, we introduce the concept of UAV virtualization. Fig. 1 depicts the overall system architecture of the proposed UAV virtualization, along with the involved actors. The architecture not only facilitates the services but also provides monetary benefits to the actors involved in this system. Virtualization allows multiple similar UAVs, that may or may not belong to the same owners, to take up a requested task. The proposed architecture consists of three actors– 1) the UAV owners, 2) the service provider, and 3) the end- users. We consider the UAVs are heterogeneous in terms of the number of connected sensors, types of sensor, and battery capacity. The proposed architecture uses the cloud as the backend infrastructure for its implementation. The available Fig. 1: The proposed system architecture for UAV virtualiza- tion. UAVs which are in the permissible range to take up the tasks form a group, which we term as the local UAV society (soc i local ), specific to that task. These groups are eventually used to select physical UAV and map it to the virtual UAV. A social welfare-based selection scheme is applied to maximize the overall residual energy (E res ) of the society. For simplicity, we consider only the energy consumption for UAV traversal and task performance by the sensors. for personal use only