Optimal Placement of Drone Delivery Stations and Demand Allocation using Bio-inspired Algorithms

The last-mile delivery problem has recently received more attention due to dramatic increase in e-commerce and one-day delivery options. One delivery method has been in the forefront of last mile delivery research: drone delivery. Particularly, hybrid truck-drone delivery systems which attempt to overcome the limitations of only using drones. Researchers have attempted to improve drone routing and scheduling, but not many have studied the required infrastructure, including drone stations for recharging and pick-up of packages. This paper tackles the placement of drone delivery stations using bio-inspired optimization algorithms. The solution framework consists of two stages. The first stage tackles the location planning problem of stations, while the second stage deals with the allocation of delivery demand to located stations. The conventional k-means algorithm is used as a baseline for the location planning problem, while the greedy algorithm is used as a baseline for the demand allocation problem. The results shows that the simulated annealing algorithm achieves a 14% reduction in total cost and 6.2× run time improvement, whereas the genetic algorithm achieves a reduction of 10% in total cost and 1.8× run time improvement.