.png){kind=icon}
.png){kind=icon}
Return
Drone Rostering Using an Evolving Hyper-Heuristic Algorithm With Average Fitness-Based Population Pruning
.png){kind=icon}
PRE
AI
DOI：10.1109/tase.2026.3716648.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Autonomous drones have been deployed in many harsh and extreme industrial environments as a crucial carrier for efficient inspection. Both the drones and the inspection instruments they carry need to be recharged periodically. Therefore, the drone rostering problem becomes a key issue. In this paper, a drone rostering problem with instrument reload considerations is modeled. An evolving hyper-heuristic algorithm with average fitness-based population pruning (EHA-AFPP) is proposed to solve the above problem. EHA-AFPP uses three types of low-level heuristic operators (LLHs), including random search operators, heuristic-based operators, and evolutionary operators, to explore the solution space. A genetic algorithm with average fitness-based population pruning is designed as the hyper-heuristic to select the sequence of LLHs. The experimental results show the effectiveness of the EHA-AFPP in solving the drone rostering problem, which can efficiently find the near-optimal solution using the fewest drones and instruments in all test scenarios. The C++ source code of EHA-AFPP can be downloaded from https://github.com/Jsy680/DroneRostering Note to Practitioners—Industrial facilities are increasingly employing drones for inspections in harsh environments. However, coordinating drone fleets for expansive, uninterrupted inspections is becoming a logistical challenge. This paper proposes a new evolving hyper-heuristic algorithm to address the drone rostering problem. Experimental results demonstrate that this method significantly reduces the number of drones and instruments needed. With this approach, companies can acquire fewer drones and instruments while preserving the same inspection capacity. It should be noted that current limitations include the assumption of unlimited charging station capacity and fixed drone operating parameters. In future research, we will consider more dynamic scenario characteristics.
Keywords:
Drone rostering problem
hyper-heuristic algorithm
evolutionary algorithm
average fitness-based population pruning
Journal
IF:
6.4
Papers:
4.9K
Citations:
1.6W