Font Size: a A A

Research On Estimation And Control For Networked Control Systems

Posted on:2017-05-09Degree:DoctorType:Dissertation
Country:ChinaCandidate:H LinFull Text:PDF
GTID:1108330485992762Subject:Control Science and Engineering
Abstract/Summary:PDF Full Text Request
With the rapid advances in computer science, communication and control techniques, net-worked control systems (NCSs) have gained widely applications in industry, military, and daily life. The introduction of network brings numerous advantages, such as reduction of wiring, lower cost in maintenance, increased system agility, ease of information sharing, etc., but also causes some new issues, such as delay and packet losses in data transmission. If the occurrence of control packet losses is available to the estimator, the NCS is usually called the NCSs with packet loss and acknowledgment (ACK), and it is also known as the Transmission Control Protocol-like (TCP-like) system; otherwise, it is called the NCSs with packet loss without acknowledgment, and it is also known as the User Datagram Protocol-like (UDP-like) system. And the UDP-like system is stud-ied in this thesis. The ACK mechanism does simplify the technical analysis, but in the networked systems it is difficult to implement due to the delay and packet loss. The UDP-like system, at the price of less delivery reliability, avoids unnecessary energy consumption and causes lower latency, making it a preferable choice for NCSs. However, it is technically involved to study the estimation and control issues for this system, many fundamental problems are still unsolved, and there are few effective methods available for the UDP-like systems.This thesis investigates the estimation and control issues for the UDP-like systems. The main work and contributions are summarized as follows:(1) For the UDP-like system, we derive the optimal estimator, and show that its error covariance (EC) can be represented as two parts:one is the EC of the optimal estimator for its correspond-ing TCP-like system and the other is a summation part. By developing an auxiliary estimator, it is shown that there exists a critical observation packet arrival rate determining stability of the expected EC (EEC), and surprisingly it is identical to its counterpart for the corresponding TCP-like system. It is revealed that whether there are ACK signals or not has no effect on the stability of the EEC. Furthermore, under some conditions the EEC converges exponentially.(2) We study the optimal LQG problems for the UDP-like system via dynamic programming ap- proach. We show that for general UDP-like system the separation principle does not hold, and the solution is difficult and sometime impossible to obtain. There are two special conditions under either of which the optimal control exists and it can be recursively computed.(3) We investigate the optimal estimation problems for the smart-sensor-based UDP-like system. We first propose two conditions:1) the sensor has some computation abilities, namely, the smart sensor; 2) the control command and the incidence of control packet loss are known to the sensor. For the UDP-like systems with these two conditions, we derive the optimal esti-mation, and then construct the finite and infinite product probability measure spaces for the estimation error covariances (EEC), respectively. Based on them, we give the stability condi-tion for expected EEC, show the existence of a measurable function to which the EEC converge in distribution, and propose some practical methods to evaluate the estimation performance.(4) We investigate the optimal LQG control problems for the smart-sensor-based UDP-like system. We first derive the LQG control law for the finite horizon case, and show that the optimal LQG control is a linear function of the state estimate, and it can be recursively computed by coupled Riccati equations. Then we provide the condition for the existence of the infinite horizon LQG control, and prove that the closed-loop system is mean square stable whenever the infinite horizon LQG control exists.(5) We study the state estimation issues of networked control systems in which observations, con-trol commands and acknowledgement (ACK) signals are randomly dropped. We develop a sub-optimal but computationally efficient filtering algorithm, and establish a sufficient con-dition for the stablity of the expected EEC. The effects of the packet losses of observations, control commands and ACK signals on the estimation performance are analyzed.(6) We study the LQG problems for the UDP-like systems. First, we propose a framework for the suboptimal-estimator-based LQG control design. Under this framework, we design a subopti-mal estimator and then solve the finite horizon LQG problem. This LQG controller is optimal under this framework, and its computation is similar to that for the classic LQG controller. Then we give the condition for the existence of infinite horizon LQG control, and prove that the closed-loop system is mean square stable whenever the infinite horizon LQG control exists.Among these aforementioned works, for the UDP-like system we solve some fundamental estimation and control issues, and develop a suboptimal but effective estimator and a suboptimal LQG controller. For the smart-sensor-based UDP-like system, we investigate some fundamental issues on optimal estimation and LQG control.
Keywords/Search Tags:networked control systems, optimal estimation, optimal control, LQG control, packet loss, acknowledgment, smart sensor
PDF Full Text Request
Related items