Risk Assessment of Unmanned Aerial Vehicle (UAV) Operations based on Bayesian Network

UAVs have a wide range of applications while there are risks in the operations. In this paper, we identify the risk factors influencing drone accidents from the perspectives of environment, technology, and personnel through the analysis of the drone accident reports dataset. Moreover, we construct a Bayesian network that reveals the dependency relationship among risk factors and the propagation mechanism leading to accident. In terms of parameter learning, prior probabilities for each cause are estimated based on reported occurrences and official drone pilot registration data, and a least squares method is used to estimate conditional probabilities. The sensitivity analysis validates the effectiveness of the Bayesian network. Through the diagnostic inference, the risk factors that have a significant impact on drone accidents are discovered.