An Efficient Framework for Probabilistic Search with Multi-UAVs

Abstract

Abstract. This paper focuses on the task of searching a stationary target using a team of multiple unmanned aerial vehicles (UAV) with limited communication ranges and sensing capabilities. It is based on a probabilistic representation model of the search environment, in which each UAV keeps a probability map on the presence of the target for each cell. The UAVs make observations and update the map for multiple rounds according to the Bayesian rule. In this process, each UAV can exchange data with its neighbors who are in its communication range. The decision of finding the target can be made once its probability of presence in any cell reaches a threshold. In this work, we design an efficient framework to decide the movement of each UAV at each time step, and propose a novel data fusion strategy among UAVs. The simulation result shows that our framework achieves improvements on both the success rate and the time taken to complete the searching task. We also explore the impact of different parameters, like the sensor model, moving mobility model of each UAV, and the data fusion strategies associated with the moving patterns. The simulation results show that the correct rate on target identification can vary greatly under different parameter settings, indicating the importance of the selection of UAV searching components such as the sensor model and the mobility model.