Population density estimation for dynamic ground risk assessment of drone operations

Abstract

Unmanned Aircraft Systems (UAS), also known as drones, have promising potential for integration into intelligent transportation systems to facilitate enhanced cargo and passenger flow. The escalating prevalence of UAS operations raises concerns regarding third-party ground risks, particularly within densely populated urban landscapes. Current mitigation strategies propose avoiding areas of high population density; however, these methodologies predominantly operate under the assumption that population density within specific locations remains static, which overlooks critical spatial-temporal population dynamics. This study introduces a prediction method with random forest regression to enhance the accuracy of population density estimations integral to risk assessment models. Obtained results were integrated into a gravity model to diffuse the predicted population volume at various time intervals, thereby generating dynamic risk maps. Finally, a graphic user interface is developed with backend computations to facilitate the decision makings for: 1) regulatory bodies to determine if a flight plan can be approved based on its risk cost and the target level of safety, and 2) UAS operators (e.g., drone delivery companies) to evaluate the overall risk level of their fleet by compute the risk cost of each flight plan. This work also contributes to the safety risk management of urban air mobility (UAM) in time-dependent ground risk mitigation.