This paper addresses the problem of efficiently managing clean-energy aerial vehicles for advanced air mobility (AAM) in a distributed manner. A concept of operation for AAM vehicles is considered to deliver packages for a number of customers at different locations. The objective of the problem is to minimize the overall energy consumption of all AAM vehicles. A feed-forward neural network (FFNN) is proposed to predict the fight energy consumption of a delivery drone, whose flight data were used to train the neural network. To optimize the allocation of AAM vehicles to service stations (e.g., for charging and maintenance) with limited service bays, a distributed limited resource allocation algorithm (DLRAA) is proposed based on the Hungarian method. DLRAA was compared with a mixed integer linear programming (MILP) solver based on the average run time and average cost in 1000 simulation runs. The results show that the proposed FFNN produced an accurate prediction given noisy flight data, the proposed DLRAA generates allocation that satisfies the given constraints efficiently, and DLRAA outperforms MILP based on average cost in all testing cases. DLRAA outperforms MILP based on the average run time when the number of charging stations is less than 30 and the number of drones is less than 160.
Allocating Limited Resources and Learning Flight Energy Consumption for Advanced Air Mobility
AIAA Journal ; 63 , 3 ; 1049-1061
2025-03-01
Conference paper , Article (Journal)
Electronic Resource
English