Tittle: A hybrid airspace modeling for using drones in logistics operations
Abstract:
Drones, or Unmanned Aerial Vehicles (VATs), are considered the future of air transport, especially in logistics operations. They can be integrated with trucks, enhancing strong bridges and mitigating weaknesses of both. This creates an optimization problem between two simultaneous vehicles, known as the Flying Helper Traveler Clerks Problem (FSTSP). This concept finds applications in the transport of vaccines and medicines, in intelligence, surveillance, recognition (IVR), in material deliveries, both military and civil operations. The objective of this research is to implement the Genetic Algorithm of Bias Random Key (BRKGA), in the context of the Traveling Cler Problem (TSP) and the FSTSP, testing it from data in two different scenarios, one civilian - case 1 Florianópolis, and one military - case 2 Advanced Army Aviation Course. The methodology for implementing meta-heuristics can be followed by a method that consists of understanding the problem using a mathematical model, using Ponza instances (2016) and two real-world scenarios, implementing and comparing the meta-heuristics using solution quality gap analysis and execution time. The algorithm demonstrates good performance in the TSP, reaching an average gap of 1.4% compared to solutions in the literature, with execution times ranging from 4s to 4463s, depending on the complexity of the instance. In FSTSP, BRKGA surpasses Simulated Annealing (SA) in some small instances and reaches an average runtime of 0.41s. When BRKGA is applied to medium and large instances of Ponza (2016), it shows reasonable performance, surpassing the SA of Ponza (2016) at runtime, with an average time of 40.9s compared to SA’s 336.06s. However, in terms of solution quality, SA reached 0.10 gap compared to 0.18 BRKGA. BRKGA achieves a time performance close to the average time of HTGVNS (18.06s) and HGVNS (16.39s), however, in terms of solution quality, HTGVNS and HGVNS remain state-of-the-art. In case 1, the BRKGA optimization algorithm saved 102.6s in the drone’s 1 tour and allowed access to inaccessible locations for trucks in the drone 2 tour. In case 2, the BRKGA algorithm saved 36.5s on the drone tour. In conclusion, BRKGA can achieve reasonable results when applied in TSP and FSTSP, however, the improvement of these results can be achieved through a more refined implementation of drones. In addition, factors such as isolated, sensitive and altitude-varying areas can be better used by drones in future research.