Research On Networked Control Systems Based On Predictive Control Strategy

NCSs(Networked Control Systems)are a class of special feedback control systems wherein the control loops are closed through network.Compared with the conventional control systems(CCSs),NCSs have shown many appealing advantages such as low cost,reduced complexity in system integration,simple installation and maintenance,high flexibility and enhanced reliability.Therefore,NCSs have been widely used in industry,such as field bus control systems,industrial Ethernet,etc.NCSs use the shared network instead of the traditional point-to-point connection to implement the data exchanges.The system configuration causes a series of problems never encountered in CCSs,such as random transmission delay,data dropout,packet disorder,etc.,which are referred to as network-induced constraints.These constraints can deteriorate the control performance,and even cause the system instable.Therefore,how to deal with these constraints becomes the basic problem in the NCSs research.Among the methodologies to cope with mentioned problems,the approaches based on predictive control strategy become increasingly popular due to that they take full advantage of the characteristics of NCSs.The most representative work in this regard is the networked predictive control(NPC)scheme,and its feasibility has been proved by both theoretical analysis and practical experiment.However,there still exists an important issue—the controller only have uncertain information about the actual control inputs applied by the actuator,which leads to the presence of the errors between actual control inputs and predictive control inputs.To solve this problem effectively,a novel system architecture—asynchronous update based predictive networked control system(AUBPNCS)is proposed,which can overcome the predictive input errors.Firstly,the issue for predictive control input errors in the NPC scheme is analyzed in depth,and then a novel networked predictive control system architecture based on proactive compensation strategy is proposed.According to the functions of different network nodes,the implementation of the system is divided into three sections – output data processing,control inputs calculation and control input update.The key aspects of the implementation process including hybrid driven modes,deterministic transmission mechanism and data buffering and selection scheme are further demonstrated through a comparative analysis of different operating states of the system.This system structure does not depend on control algorithms and network protocols.It is able to handle random delay,packet loss and packet disorder in both forward and feedback channel simultaneously,and eliminate the predictive control input errors in classic NPC scheme effectively.Next,in framework of AUBPNCS proposed in this paper,three different control algorithms are applied to complete the design of the networked controller,respectively.Stability analyses of the corresponding closed-loop systems are investigated,and some simulation examples are given to demonstrate the effectiveness of the control strategy.(1)In framework of AUBPNCS,a networked dynamic matrix control algorithm is proposed,which is suitable for the controlled plants described by finite step respond(FSR)models.At this time,traditional prediction control algorithm based on that single value(the first item of the control sequence)is applied is changed into a novel control algorithm with a variable applied length(including but not limited to the first value)in the control sequence.By transforming the closed-loop system into impulse response model,robust stability criterion of the variable applied length dynamic matrix control algorithm is derived analytically.The criterion is eatablished under the assumptions that both the delay and consecutive data dropouts are bounded and the model uncertainty meets a certain range.Finally,a simulation example based on the True Time toolbox verifies the effectiveness of the control strategy.(2)In the framework of AUBPNCS,a networked state feedback control algorithm suitable for the controlled plants described by state space models is proposed.In the algorithm an extended state observer is used to realize the model prediction and online correction.With some proper modifications,the closed-loop system is equivalent to a special kind of the model based networked control system,whose update cycle is the round-trip delay.The stability criterions of the closed-loop system are derived under fixed and random delay conditions,respectively.At last,under the True Time simulation environment,the plant is modeled based on the actual parameters of a linear two-stage inverted pendulum system.The results verify the effectiveness of the algorithm under different network load conditions.(3)In the framework of AUBPNCS,a networked generalized predictive control algorithm based on polynomial model is proposed.Based on the feedback data packet which contains the historical input/output sequence generated by the sensor node,an event-driven online parameters identification algorithm is proposed.The closed-loop system is expressed in a form of augmented state space model,and the stability criterion under random delay is derived based on switching system approach.Finally,the simulation examples show the effectiveness of the algorithm under different delay conditions.Finally,the main contents of this dissertation are summarized,the potential works are pointed,and the outlook on the development trend and application prospect of NCSs is made.