1 Mobility-Aware Multipath Communication for Unmanned Aerial Surveillance Systems Shiva Raj Pokhrel, Jiong Jin and Hai Le Vu Abstract—Wireless communication platforms of unmanned aerial vehicles (UAVs), commonly known as drones, usually operate in dynamic conditions with highly fluctuating capacities and erratic wireless access to the network resources. Effective design for adequate, robust and consistent communications from the drones to the control center (aka server) is thus necessary to facilitate the implementation of the UAV system. Multipath TCP has the potential to exploit heterogeneous wireless paths and achieve robust bandwidth by controlling the dynamics of the convoy of drones. In this paper, we make use of fluid approach for proposing a generic system architecture towards a control mechanism for coordinating the convoy of drones. Moreover, we model multipath TCP and study their applications for the control scenario of a convoy of drones with three different communication interfaces, representing the head, tail and the middle of the convoy. The access is through a Satellite and several WiFi access points (APs) communicating to the server with time-variant wireless channel, subject to the dynamic mobility and distance of APs from the convoy’s head and tail. The traffic on different paths governed by the flow control based on the movement and position of the convoy enable us to investigate the performance and convergence of communication system dynamics. In particular, Lyapunov theory is adopted to derive the stability conditions of the dynamics over the mobility of the convoy. I. I NTRODUCTION With the increasing advancements in mobile computing, communication, and wireless networking as well as the minia- turisation of devices, unmanned aerial vehicles (UAVs) such as drones, quadcopters and gliders, and their networks have been receiving substantial consideration in the mobile communication research community. 1 Certainly, UAV networks become an integral component in numerous critical applications ranging from border surveillance, disaster response and recovery, traffic monitoring, to the shipping of goods, medicine, and first aid [1]–[3]. In contrast to the terrestrial wireless networks, such airborne communication networks (ACNs) have different characteristics, e.g., highly dynamic network topologies due to mobility and weak connection among communication links. Therefore, seminal standards, paradigms, and protocol design methodologies are not directly useful to ACNs. Moreover, the mobile platforms of UAV networks usually experience fluctuating wireless access, large and volatile delays, varying Shiva R. Pokhrel is with the School of Information Technology, Deakin University, Geelong; Jiong Jin is with the School of Software and Electrical Engineering, Swinburne University of Technology; and Hai L. Vu is with Institute of Transport Studies, Faculty of Engineering, Monash Univer- sity, Australia. Email: {shiva.pokhrel@deakin.edu.au, jiongjin@swin.edu.au, hai.vu@monash.edu}. 1 http://www.gartner.com/newsroom/id/3602317, 2017, press Release, Garter. Server Head Tail Satellite Access Point-n Access Point-2 Access Point-1 - - - - - - - Fig. 1: Network setting for a highly moving convoy of drones. The head and tail of the convoy have direct access to multiple WiFi access points while the center drone has access only to a Satellite. Packets are assumed to be transmitted instantaneously within the convoy, thus turning the drone convoy a single system in view of the transmission to and from the controlling server. asymmetric bandwidths and erratic connections [4]–[7]. As a result, well-designed techniques and algorithms for mobility- aware, robust and stable communications between such a dynamical system and their central servers are challenging but essential [7]–[10]. Authors in [8] investigate the deployment of multiple UAVs for on-demand coverage while at the same time maintaining the connectivity among UAVs. To solve this prob- lem, they proposed two algorithms: a centralized deployment algorithm and a distributed motion control algorithm. We consider a similar network scenario with a large number of small autonomous drones that form a cohesive group or “swarm” to accomplish complex missions as a whole is deployed. Swarm of drones offers numerous benefits over single UAV including a larger coverage, redundancy in numbers and reduced bandwidth requirements. In this context, the problem of multipath air-to-ground transmission has been partially investigated and explained in [11] and the references therein. More importantly, a number of challenges arise in order to achieve joint communication and movement control for the surveillance system, encompassing drones and a remote server as shown in Fig. 1. The key is not only setting the swarm behavior through robust communications, but also designing an effective mechanism of movement control so that the swarm can be directed to complete the desired surveillance mission. Specifically, the following two major challenges are identi- fied: ◦ Heterogeneous channel characteristics