Cooperative Data Muling from Ground Sensors to Base Stations Using UAVs Emmanuel Tuyishimire * , Antoine Bagula * , Slim Rekhis † , and Noureddine Boudriga † , * Computer Science, University of the Western Cape, Cape Town, South Africa † Cartage University, Tunis, Tunisia Abstract—Data muling using UAVs/drones is currently emerging as an alternative to the traditional traffic engineering techniques used in wireless sensor networks, when wireless communication is not an option or the least cost-efficient solution. This paper revisits the issue of traffic engineering in Internet-of-Things (IoT) settings, to assess the relevance of using UAVs for the persistent collection of sensor readings from the sensor nodes located into an environment and their delivery to base stations where further processing is performed. We consider a persistent path planning and UAV allocation model, where a team of UAVs coming from various base stations are used to collect data from ground sensors and deliver the collected information to their closest base stations. This problem is mathematically formalised and proven to be NP-hard. We propose a heuristic solution for the problem and evaluate its relative efficiency through simulation. Index Terms—Path planning, file carving, waiting time, path map, speed distribution I. I NTRODUCTION Unnamed Aerial Vehicles (UAVs) are emerging as a flexible and cheap alternative to traffic engineering techniques which have been traditionally used in IoT settings, to transport sensor readings from their points of collection to their processing places. However, the joint path finding and resource allocation for a team when tasked to achieve collaborative data muling, is still an issue that require further investigations. Furthermore, while accurate solutions to data muling problems are still scarce, especially when considering the limited flying autonomy of the battery-powered UAVs, issues related to the efficient task allocation to a team of UAVs under stringent data collection requirements have not yet been fully addressed. Potential applications of the proposed model include (i) city surveillance in order to evaluate risks and respond with appropriate actions by having a team of UAVs persistently visiting locations of interests in a smart city for public safety, parking spots localization [1] and pollution monitoring [4] ; (ii) drought mitigation to support small scale farming in rural areas [15, 16] by using a team of UAVs to collect farmland image collection and processing these images to achieve situation recognition for precision irrigation; (iii) periodic surveillance of buildings and cities’ infrastructures for structural health monitoring and maintenance; and (iv) extension of the reach of community mesh networks in rural settings for healthcare [14, 2] by using a team of UAVs (such as drones) as wireless access points. Sensors visitation under the fuel consumption constraints was addressed in [10], and the visitation under the revisit deadline constraint was proposed in [11]. Both works assume a single moving agent (UAV) which optimally visits various targets. [12] proposes a cooperative UAVs model where many targets are visited by a team of UAVs for persistent surveillance and pursuit. In this work, the UAVs do not communicate with each other but rather rely on the information from the static underground sensors, which are optimally placed as proposed in [13]. However, all these models do not consider the persistent data delivery and heterogeneity of UAVs which might have different fabrics and characteristics. Furthermore, neither the energy/battery consumption while the UAVs are waiting for the updated information from the terrestrial sensor network nor the penalty associated with stale information due to late visitation by the UAV to the sensor nodes have been accounted for. While models were proposed in [19, 3, 5, 7] for the periodic and persistent UAVs visitation of a single target from different positions, the models do not consider the path planing issues which are as necessary as the path planning especially for restricted environments. This paper proposes a persistent path planing and task allocation model where, a team of UAVs coming from various base stations are used to collect sensor readings from ground sensors and deliver the collected information to their closest base stations. The underlying data muling problem is i) mathematically formalised as a constrained optimisation problem, ii) proven to be intractable and iii) solved using a heuristic solution, whose relative efficiency is proven through simulation modelling. The rest of this paper is organised as follows. The cooperative data muling model is presented in Section II and its algorithmic solution provided in the same section. Simulated results are provided and discussed in Section III while the conclusion is drawn in Section IV. II. THE COOPERATIVE DATA MULING MODEL In this paper, we consider an “Internet-of-Things in Motion” model depicted by Figure 1. We assume that UAVs are assisted by special ground-based sensors which locally collect data from other sensors. That is, sensors are grouped into separated clusters, each with its own sink node (the cluster head), where the information is to be collected from other sensor nodes (cluster members) and relayed to UAVs which deliver the sensor readings to base stations. Note that only cluster heads can communicate with UAVs, and the optimal clustering scheme is not covered in this paper. Furthermore, the inner cluster communication technology is not covered here (it has been discussed in [20]). Fig. 1: Cooperative Data Muling The cooperative data muling model considered in this paper is illustrated by Figure 1, which reveals four base stations (Bs1, Bs2,