在线分布式优化研究进展

Recent Advance in Online Distributed Optimization

在线分布式优化是通过多智能体之间有效的协调合作以实现实时优化的任务,其在多无人系统实时编队跟踪、大规模分布式机器学习、传感器网络的动态参数估计等领域中有着广泛的应用。分别从决策空间、性能指标、信息反馈模型等角度对近年来国内外在线分布式优化的代表性研究工作进行了梳理和总结,着重从算法设计思路和收敛性结果两个角度进行了剖析,同时也指出不同算法的优势和不足。最后,对在线分布式优化算法的应用进行了简单讨论并对全文进行了总结和展望。

Online distributed optimization is a real-time optimization task fulfilled by a system of multiple agents through effective cooperation,and it can be applied to fields such as formation and tracking of multiple unmanned system,large-scale distributed machine learning,and dynamic parameter estimation in sensor networks.This paper presents an overview of the state-of-the-art research on online distributed optimization in recent years,from the angle of decision space,performance index,information feedback model,with special focus on the intuition behind the algorithm design and the convergence analysis results.Meanwhile,discussions about the pros and cons of the algorithms have been provided.Finally,we present some applications of online distributed optimization,make a summary of the paper,and discuss possible directions for future research.