DOCF: A Task-Agnostic Dispersed Collaborative Computing Framework for Aerospace Multi-Agent System
Jialin Hu, Zhiyuan Ren · 2023
Collaborative computing for aerospace Multi-Agent System (MAS) has become an attractive technology in both academia and industry. Most researches focused on the performance enhancement for specific mission scenarios, which limits their versatility, increases their burden, and leads to a lot of duplicate investment. To relieve this tension, a task-agnostic dispersed collaborative computing framework (DOCF) for aerospace MAS is proposed. The core idea is the separation of strategy generation and data computation. In the strategy generation phase, the strategy that supports various missions with adaption to the features of aerospace MAS (i.e., decentralized and dynamic environment) can be computed distributedly by diffusing the task model and network information to the network. In the data computation phase, the data is processed step by step and the mission can be accomplished by orchestrating the resources of agents based on the strategy. Moreover, the sequence diagram of the proposed framework is demonstrated and the use cases are discussed. Through analysis, the proposed framework can provide a generic approach that can be applied to performing various missions with adaptation to the decentralized and dynamic environment.