面向网络诱导时延和数据包丢失补偿的网络化广义预测控制  

Networked generalized predictive control for compensation of network-induced time delay and packet loss

作　　者：唐斌[1,2] 章云[1] 刘国平[2,3] 桂卫华[2] 

机构地区：[1]广东工业大学自动化学院,广东广州510006 [2]中南大学信息科学与工程学院,湖南长沙410083 [3]格拉摩根大学工程学院

出　　处：《控制理论与应用》2010年第7期880-890,共11页Control Theory & Applications

基　　金：国家自然科学基金重点资助项目(60634020);国家自然科学基金面上资助项目(60874008);广东省自然科学基金研究团队项目(8351009001000002);广东省自然科学基金博士启动项目(9451009001002702)

摘　　要：针对网络化控制系统中存在的网络诱导时延和数据包丢失,考虑了网络化广义预测控制问题.基于多个数据打包传送的通讯方式以及网络诱导时延和数据包丢失预先可知的假设,提出了一种采用最小预测步长和预测控制增量向量分别补偿网络诱导时延和数据包丢失对系统性能影响的新方法,给出了相应的网络化模型预测算法和网络化滚动优化算法,对于被控对象参数未知或缓慢变化的情况,给出了基于递推最小二乘辨识改进算法的网络化反馈修正算法,通过仿真验证了所提出网络化算法的有效性.A generalized predictive control is considered for networked control systems with time-varying network- induced time delay and packet loss. Because multiple data are sent in one packet from the source node to the target node, and the network-induced time delay and the packet loss are known a priori, a new method is proposed to compensate the influences of network-induced time delay and packet loss on control performance using minimal prediction horizon and predictive control increment vector, respectively. Thus, we develop the corresponding networked model prediction algorithm and the algorithm of networked receding-horizon optimization. For the controlled system with unknown or slowly varying parameters, the networked feedback correction algorithm is discussed based on a modified recursive leastsquares identification algorithm. Simulation results are given to show the effectiveness of our proposed algorithms.

关 键 词：网络化广义预测控制 网络诱导时延 数据包丢失 最小预测步长 预测控制增量向量 

分 类 号：TP273[自动化与计算机技术—检测技术与自动化装置]