Research on Location Selection of Takeoff and Landing Points for Auxiliary Rescue UAV in Hazardous Materials Transportation Accidents

Planning and deploying unmanned aerial vehicle (UAV) takeoff and landing sites in areas with concentrated hazardous material transportation routes, combined with UAV-assisted joint emergency rescue operations alongside ground-based emergency rescue stations, can significantly enhance the emergency response capabilities for such accidents. This paper investigates the location and allocation problem of UAV landing sites for auxiliary rescue in hazmat transportation accidents. First, a location-allocation model for UAV landing sites under deterministic scenarios was constructed. Second, considering the uncertainty of potential risks along road network segments, distributionally robust optimization (DRO) theory was introduced, and a robust location-allocation optimization model for UAV landing sites was established. Third, an easily solvable approximation method was adopted. Under the constraint of a zero-mean bounded perturbation ambiguity set, the original distributionally robust optimization model was equivalently transformed into an integer programming model, which was then solved using a branch-and-cut algorithm based on Benders decomposition. Finally, numerical examples were conducted to verify the effectiveness of the proposed models and solution algorithm. The results indicate that although the output of the distributionally robust optimization model tends to be relatively conservative, it exhibits stronger robustness and can effectively enhance the overall reliability of the emergency rescue system. The coverage effectiveness of UAV landing sites increases with the number of planned takeoff and landing points, generally following a pattern of diminishing marginal returns. Under conditions where probability distribution information of uncertain parameters is incomplete, the distributionally robust optimization approach outperforms traditional stochastic programming methods. Compared with conventional robust optimization, distributionally robust optimization can incorporate prior information on partial probability distributions, thereby effectively avoiding the issue of overly conservative location-allocation solutions.