Conceitos Básicos
A hybrid genetic algorithm with type-aware chromosomes (HGA-TAC) is proposed to efficiently solve the Traveling Salesman Problem with Drone (TSPD) and the Flying Sidekick Traveling Salesman Problem (FSTSP). The key contribution is the discovery of how decision-making processes should be divided among the layers of genetic algorithm, dynamic programming, and local search.
Resumo
The paper presents a hybrid genetic algorithm (HGA) for solving the Traveling Salesman Problem with Drone (TSPD) and the Flying Sidekick Traveling Salesman Problem (FSTSP). The algorithm consists of three main components:
-
Genetic Algorithm (GA) Layer:
- Generates incomplete solutions by determining the customer nodes served by the truck and the drone.
- Utilizes a novel type-aware chromosome (TAC) encoding to distinguish truck and drone nodes.
- Employs type-aware order crossover operations to support the TAC encoding.
- Uses two or three subpopulations depending on whether the drone range is limited.
- Proposes an escaping strategy to prevent the GA from being trapped in local optima.
-
Dynamic Programming (DP) Layer:
- Completes the solutions by determining the optimal launch and landing locations for the drone.
- Devises a dynamic programming algorithm, called Join, with a time complexity of O(n^2).
-
Local Search Methods:
- Applies various local search techniques, including 7 new methods, to improve the generated solutions.
The proposed division of decision-making among the three layers, along with the TAC encoding, brings reductions in the objective function values, computational time, or both, in most benchmark instances compared to existing methods.
Estatísticas
The minimum time for the network in Figure 2a solved by TSP is 39.
The minimum time for the network in Figure 2b solved by TSPD is 25.
Citações
"Unlike traditional trucks, drones do not have to follow the same network, giving them faster delivery times. They are also more eco-friendly since they are electrical and less expensive since they do not require any human involvement."
"Our HGA approach consists of three layers in its algorithmic structure. First, a regular genetic algorithm (GA) layer generates incomplete solutions by determining the customer nodes served by the truck and the drone. Second, a dynamic programming (DP) layer completes the solutions by determining the combined nodes, where the drone departs from the truck or returns to the truck. Third, we use various local searches to improve the generated solutions."