On-line Prediction Method of Random Delay in Forward Path of Networked Control System

Abstract

The invention discloses an online prediction method for the random time delay of a forward channel of a network control system, and the method gives the consideration to a condition that the time delay of the forward channel cannot be directly obtained at a controller end, and predicts the unknown time delay in advance so as to reduce the impact on the performance of the system from the uncertainty of the time delay in the network control system. The method comprises the steps: firstly enabling a buffering queue to collect time delay historical data through a feedback channel, ordering and updating usable information for a controller; secondly enabling the controller to initialize an online prediction model based on enough time delay historical data H(k), updating parameters through a recursive least-squares method if there is new usable data xn+1 so as to adapt to the real-time change of a network, and regularly carrying out the rolling prediction, wherein a trend terminal dk is added to a prediction value after the inverse differential processing of the prediction value, i.e., the prediction value (shown in the description) of the time delay of the forward channel. The method employs the recursive least-squares method to update the model parameters in real time, has a capability of tracking time-varying parameters, and can adapt to the real-time change of the network better.

Description

technical field [0001] The invention relates to a prediction method for network-induced time delay of a network control system, in particular to an online prediction method for a random time delay of a forward path of a network control system. Background technique [0002] In Networked Control Systems (NCSs), the research goal of control strategy is to overcome the influence of network-induced delay, data packet loss and timing disorder on the control system, and achieve the required control performance. According to the impact on system performance, the network-induced delay is divided into two categories: the feedback channel delay from the sensor to the controller and the forward channel delay from the controller to the actuator. Both types of delay are generated by the network, which may be bounded or unbounded, time-varying and random, and its randomness will lead to disordered data timing at the receiving end of the network. For the control system, the former can be m...

AI Technical Summary

Problems solved by technology

In the paper "Research on Network Delay Analysis of Internet-Based Remote Control System", Xu Shuping et al. trained a generalized regression neural network based on the sliding model of spline mean to establish a prediction model of network-induced delay, which has good generalization ability, but The training time is too long, and the parameters cannot be updated online

Images