CAJ | 학술논문

基于量化状态信息的异步随机Gossip算法大多以均匀选择概率的时间模型为基础，未充分考虑网络拓扑结构对局部信息传递的影响。为此，该文提出了一种以非均匀选择概率为时间模型的改进算法。首先给出了非均匀选择概率下的多智能体系统时间模型，在随机性量化策略下给出了一致性误差的收敛性质；并讨论了量化精度和概率化权重矩阵第2大特征值对一致性误差收敛速度的影响，进而利用投影次梯度给出了选择概率的分布式优化方法。仿真结果表明，该基于量化状态信息的算法可通过选择概率的分布式优化，提高一致性误差的收敛速度。
기우양화상태신식적이보수궤Gossip산법대다이균균선택개솔적시간모형위기출，미충분고필망락탁복결구대국부신식전체적영향。위차，해문제출료일충이비균균선택개솔위시간모형적개진산법。수선급출료비균균선택개솔하적다지능체계통시간모형，재수궤성양화책략하급출료일치성오차적수렴성질；병토론료양화정도화개솔화권중구진제2대특정치대일치성오차수렴속도적영향，진이이용투영차제도급출료선택개솔적분포식우화방법。방진결과표명，해기우양화상태신식적산법가통과선택개솔적분포식우화，제고일치성오차적수렴속도。
As the traditional quantized asynchronous randomized gossip consensus algorithm is based on uniform selection probability time mode, the impact of network topology on local information transfer is not been fully considered. Thus, an improved quantized asynchronous randomized gossip consensus algorithm with non-uniform selection probability is proposed in this paper. Firstly, the asynchronous time model with non-uniform selection probability is proposed. Then the convergence of the algorithm is analyzed with randomized quantized information. The impact of the quantization resolution and the second largest eigenvalue of the probabilistic weighted matrix on convergence rate is also discussed. Furthermore, this paper proposes an optimization algorithm for selection probabilities with projection subgradient method in a distributed manner. The numerical example indicates that, the proposed algorithm improves the convergence rate by optimizing selection probabilities of agents.