A Planner for Autonomous Risk-Sensitive Coverage (PARC OV) by a Team of Unmanned Aerial Vehicles Alex Wallar Erion Plaku Donald A. Sofge Abstract— This paper proposes a path-planning approach to enable a team of unmanned aerial vehicles (UAVs) to efficiently conduct surveillance of sensitive areas. The proposed approach, termed PARCOV (Planner for Autonomous Risk- sensitive Coverage), seeks to maximize the area covered by the sensors mounted on each UAV while maintaining high sensor data quality and minimizing detection risk. PARCOV uses a dynamic grid to keep track of the parts of the space that have been surveyed and the times that they were last surveyed. This information is then used to move the UAVs toward areas that have not been covered in a long time. Moreover, a nonlinear optimization formulation is used to determine the altitude at which each UAV flies. The efficiency and scalability of PARCOV is demonstrated in simulation using complex environments and an increasing number of UAVs to conduct risk-sensitive surveillance. Fig. 1. Snapshots of PARCOV at different iterations showing how the quadcopters cover the designated area. The risk model is shown as a heatmap with red indicating high risk and blue indicating low risk. Figures better viewed in color and on screen. I. I NTRODUCTION UAVs are seen as providing a viable way to enhance automation in environmental monitoring, search-and-rescue missions, package delivery, target tracking, and many other applications. UAVs, such as ARDrone and AscTec Pelican quadcopters, are becoming more commercially available, making them also an economically-feasible option for de- ployment in autonomous aerial missions. Towards increasing the autonomy of UAVs, this paper describes an algorithm for persistent area coverage using multiple cooperative quadcopters while accounting for the risk and sensor data quality involved in the coverage. The proposed approach, PARCOV, seeks to move the quadcopters to promote informed coverage and adjusts the altitude to maximize sensor data quality while minimizing the associ- ated risk. Risk plays an important role in many autonomous aerial missions, especially when seeking to reduce the likeli- hood of being detected by a possibly hostile agent. Although this paper focuses on detection risk, PARCOV is general and can minimize other risk metrics that decrease in value as the altitude increases. For instance, risk can also be used to model a brushfire. In such scenario, PARCOV can provide risk-sensitive aerial coverage of a wildfire while maximizing the sensor data quality. There is a burgeoning body of work focusing on aerial missions using one or several UAVs [1], [2]. Ergezer and Leblebiciog ˘ lu [3] describe an algorithm for 3D path planning using UAVs that seeks to avoid forbidden regions and max- imize information collection from desired regions. Nikolos A. Wallar is with the School of Computer Science, University of St Andrews, Fife KY16 9AJ, Scotland, UK. E. Plaku is with the Dept. of Electrical Engineering and Computer Science, Catholic University of America, Washington DC 20064 USA. D. A. Sofge is with the Naval Research Laboratory, Washington, DC 20375 USA. et al. [4] develop evolutionary algorithms for offline/online path planning for UAVs. Kuhlman et al. [5] provide an algorithm that optimizes a closed-loop trajectory path for persistent area coverage by a single UAV that maximizes the information gained from information-rich areas. Beard and McLain [6] take into account communication-range constraints in order to ensure that UAVs always remain in communication range as they visit desired regions and avoid forbidden regions. Cheng, Keller, and Kumar [7] derive a control policy that generates time-optimal UAV trajectories for urban structure coverage. Chandler, Pachter, and Ras- mussen [8] propose cooperative control techniques in order to minimize team exposure to radar detection. Distributed- task allocation procedures are developed in [9]–[11] in order to enhance cooperative searching. Sydney, Paley, and Sofge [12] provide a physicomimetic method for target detection using a group of UAVs. This approach continuously searches the area by following the gradient of an information surface to track targets using mutual information between the UAVs. A bio-inspired approach is proposed in [13] which seeks to model the information of a search space as a field for grazing and the UAVs as grazing animals that seek to eat the available information. This approach has shown to converge more quickly to total information collection than traditional lawnmower methods. Genetic algorithms are used in [14] to design evolving behaviors that could increase the autonomy of a swarm of UAVs in carrying out search-and-destroy missions. Huynh, Enright, and Frazzoli [15] analyze the persistent-patrol problem for a team of UAVs and propose several policies to minimize the expected waiting time be- tween the occurrence and detection time of an incident. The proposed approach, PARCOV, offers several contribu- tions. In particular, it utilizes simple interactions between UAVs to promote an emergent behavior that maximizes