Scheduling and resource allocation method for high-performance federated learning in UAV bee colony

Abstract

The invention relates to a scheduling and resource allocation method for high-performance federated learning in UAV bee colony, and belongs to the technical field of wireless communication. The method comprises the following steps: S1, establishing a UAV bee colony network model supporting SWIPT; S2, establishing a federated learning system model based on the SWIPT technology, and establishing a mathematical model of the UAV scheduling number maximization problem under the condition of meeting the constraint conditions of energy consumption and time delay; S3, introducing a simplified variable, and changing a mathematical expression of a constraint condition to obtain a simplified UAV scheduling number maximization problem mathematical model; S4, acquiring an optimal equipment scheduling and resource allocation method enabling the UAV scheduling number to be maximum by using GBD; and S5, acquiring a suboptimal equipment scheduling and resource allocation method which enables the UAV scheduling number to be maximum by using a low-complexity algorithm. According to the method, the performance almost the same as that of an optimal algorithm can be achieved under various network settings.

Description

technical field [0001] The invention belongs to the technical field of wireless communication, and relates to a scheduling and resource allocation method for high-performance federated learning in a UAV bee colony. Background technique [0002] In recent years, UAV swarms have been widely used in fields such as logistics, agriculture, and surveillance. While supporting these services, UAVs in swarms generate large amounts of data that can be used for machine learning tasks such as localization, trajectory planning, and object recognition. Traditionally, ML model training needs to transmit the data in the bee colony to a remote computing center (such as cellular base station BS, etc.) for centralized processing. Due to the high mobility of UAV and the high randomness of wireless channels, the communication overhead of data transmission will be huge. Therefore, UAV swarms may be more suitable for model training with federated learning techniques proposed in recent years. Us...

AI Technical Summary

Problems solved by technology

However, research conclusions about which UAVs should be scheduled to participate in model training and how much computing and communication resources should be allocated to these scheduled UAVs to obtain high-performance federated machine learning models have not been reported yet.

Images