A Real-time Coverage and Tracking Algorithm for UAVs based on Potential Field Hosein Khandani Control and Intelligent Processing Center of Excellence School of Electrical and Computer Engineering University of Tehran Tehran, Iran E-mail: h.khandani @ ece.ut.ac.ir Hadi Moradi Control and Intelligent Processing Center of Excellence School of Electrical and Computer Engineering University of Tehran Tehran, Iran E-mail: moradih@ut.ac.ir Javad Yazdan Panah Control and Intelligent Processing Center of Excellence School of Electrical and Computer Engineering University of Tehran Tehran, Iran E-mail: yazdan@ut.ac.ir Abstract—One of the important applications of the UAVs is the coverage of an area and tracking possible targets. When the area is large, optimal coverage and smart search will be an important issue. On the other hand, in real time applications, its computational time is as important as the optimality of the algorithm. In this paper, an algorithm based on Potential Field is proposed to search an area in a way to result good performance. One of Potential Field pitfall is getting stuck in local minima, which is alleviated using Random Moves, avoid Past, and Globally Oriented Search in this approach. The approach has been simulated and compared with two common benchmark search space methods. Finally, the method is implemented on a quadrotor and its performance is evaluated. Keywords— Optimal Real Time Search; Rescues Operation; Potential Field. I. INTRODUCTION Recently UAVs are utilized for border patrol, persistent surveillance, and search and rescue operations which involve coverage and tracking tasks. To provide coverage and tracking in a certain area, it is required to constantly cover the whole area, under surveillance, and look for possible targets. After finding a target, higher priority should be devoted to tracking in comparison to the coverage task, for instance, the application of UAV in lifeguard’s duty which is depicted in fig. 1. In this task, 1) in the first step the lifeguard covers the allocated area to it and looks for possible swimmers. 2) After detecting the target, the lifeguard would start tracking it to ensure that it would safely reaches the coast or it passes the area under the lifeguard’s surveillance. It is important to note that the lifeguard would still try to cover the areas in the vicinity of the tracked target to cover any more swimmers that could be found in the area. Consequently, the problem becomes the optimization of a function with higher priority for tracking and lower priority for coverage. 3) After the swimmer quitted from the allocated area, the UAV should return and cover the other regions that were not covered previously. A large and growing body of literature has reported in this area recently. For instance, target search and track were discussed in [1], with the aim of achieving the maximum likelihood of recovering a lost target. In this study, four path patterns are introduced to search an area. These patterns are claimed to maximize the expectation of rediscovering the target. Two examples of these patterns are the Spiral and the Lawnmower. In [2], Genetic Algorithm (GA) is used to control a group of UAVs. By combining the pheromone spraying method with the GA, the overall performance get improved. Each UAV sprays pheromone on the path that it has covered, sends its observations to a control center and moves toward the direction with minimum pheromone. This process can be controlled by changing the rate of pheromone evaporation. GA is not grantee for optimal solution and for real time decision is poor performance. An approach is proposed in [3], to minimize the amount of time spent on finding victims. UAV is supposed to search the area for victims while avoiding collisions with obstacles and other UAVs. Three methods are used for evaluation: Greedy heuristics, Potential-based heuristics and Partially Observable Markov Decision Process (POMDP) based heuristics. The third method gained higher performance in comparison to the other two heuristics, but its computation time increases as the number of cells grows. In this method, it is impossible to achieve the optimal solution for large environments due to the long computation time. In [4], search algorithms are divided into two classes: Traditional approaches and Modern approaches. The traditional approaches include Lawnmower and Hill Climbing, and modern approaches take into account Simulated Annealing and Fig. 1. A UAV in lifeguard duty switching between coverage and tracking. Proceeding of the 2nd RSI/ISM International Conference on Robotics and Mechatronics October 15-17, 2014, Tehran, Iran 978-1-4799-6743-8/14/$31.00 ©2014 IEEE 700