差别隐私
计算机科学
量化(信号处理)
控制理论(社会学)
带宽(计算)
变量(数学)
脉冲(物理)
电信网络
控制器(灌溉)
通信系统
噪音(视频)
多智能体系统
投影(关系代数)
信息隐私
国家(计算机科学)
噪声测量
状态变量
控制(管理)
输出反馈
控制系统
脉冲噪声
算法
电话网
线性系统
全状态反馈
干涉对准
差速器(机械装置)
控制变量
分布式计算
作者
Xiaodong Lü,Yulin Zhao,Ji-Feng Zhang
标识
DOI:10.1109/tac.2026.3668109
摘要
This paper investigates the differentially private output consensus problem of continuous-time heterogeneous multi-agent systems. An algorithm blending estimation, control and privacy is proposed. To prevent privacy leakage, each agent constructs an auxiliary variable regarding to its own initial output, and then adds a non-decaying noise to this variable. To incorporate the network bandwidth limitation, each agent quantifies the noisy auxiliary variable to 1-bit binary-valued data, and transmits this data to its neighbours at some impulse time instants. Based on the received information, each agent builds the recursive projection algorithm to estimate its neighbours' auxiliary variables, and updates its own auxiliary variable. Then, each agent designs its controller by using its own state and auxiliary variable to achieve the mean-square output consensus. To characterize the degree of privacy protection, the differential privacy index of the proposed mechanism is derived. Compared with the existing works, this paper only needs 1-bit communication bandwidth, and the control signals are not required to constantly feed back to the original system. From the aspect of quantization and control, the communication resources can be conserved. An example is presented to illustrate the effectiveness of the theoretical results.
科研通智能强力驱动
Strongly Powered by AbleSci AI