Broker-Mode Tree-Structured Cooperative Offloading of Splittable Tasks for Dynamic Multi-UAV-Assisted MEC Systems

Abstract

The rapid development of uncrewed aerial vehicles (UAVs) has expanded the application scenarios of mobile edge computing (MEC). However, existing multi-UAV collaborative MEC schemes usually rely on fixed cooperation structures or limited collaboration mechanisms, making it difficult to fully utilize distributed UAV computing resources in dynamic network environments. To address this issue, this article considers splittable tasks and proposes a broker-mode tree-structured multi-UAV collaborative computing framework for dynamic UAV-assisted MEC systems, aiming to minimize total system energy consumption. In the proposed framework, any UAV can dynamically act as a broker once a task is offloaded to it. The broker UAV determines whether the task can be executed locally and, if necessary, partitions the task and forward subtasks to other UAVs according to available computing resources and network conditions. Through recursive cooperation, a tree-structured collaborative topology is formed, enabling efficient task computing while improving resource utilization and cooperation flexibility. To jointly optimize task offloading decisions, UAV trajectory, multi-UAV cooperation, and resource allocation, the original nonconvex optimization problem is decomposed into three subproblems under the block coordinate descent (BCD) framework and solved using successive convex approximation (SCA) and related techniques. Simulation results show that the proposed algorithm reduces task computation energy consumption by over 10% and improves the task completion ratio by more than 9% compared with the suboptimal scheme.