Urban On-Demand Delivery via Autonomous Aerial Mobility: Formulation and Exact Algorithm

Abstract

The implementation of the autonomous unmanned aerial mobility is a game changer for the on-demand delivery service in the crowded urban setting. In this study, the first of its kind commercial unmanned aerial vehicle (UAV) urban delivery program in China is targeted. Different from the traditional ground pickup and delivery services, the aerial mode considers not only the time window constraints, but also the spatial conflicts incurred during the take-off and landing operations of UAVs. To obtain the optimal flying routes of the focused problem, a mixed integer programming model is formulated. Due to its inherent complexity, the optimal schedule cannot be attained within acceptable time via the off-the-shelf solvers. To help speed up the solving process, a branch-and-cut based exact algorithm is proposed, together with a series of customized valid inequalities. To further accelerate, a greedy insertion heuristic is designed to secure high-quality initial solutions. In the numerical section, it is observed that the algorithm proposed in this paper can help solve the real-life on-demand UAV delivery problem to near optimum (within 5% optimality gap) within reasonable computation time (in 5 minutes). <i>Note to Practitioners</i>&#x2014;With the increase of labor cost, the distribution cost increases very rapidly. In the meantime, the employment of automated vehicles for logistics reshapes the landscape of the urban last-mile delivery. As an efficient courier carrier, the unmanned aerial vehicle (UAV) is trending the autonomous delivery endeavour. When integrating UAVs into the urban delivery program, practitioners need to pay special attention to the scheduling of UAVs at the operational level in addition to the hardware of the UAVs. To help solve the UAV dispatch problem, we propose an online scheduling scheme, considering the spatial conflict constraints in the actual UAV operations. And an exact algorithm is designed to accelerate the solving process. Numerical experiments demonstrate that the proposed algorithm can achieve near optimal dispatch plan with 5% optimality gap in 5 minutes. Furthermore, it is discovered that the demand pooling is an essential decision to make for UAV-based delivery. Longer pooling time can increase the UAV efficiency with more realized demand information, but too much pooling could lead to prolonged customer waiting and a low service level.