Andrzej Tutaj

Improving quality of control in networked control systems by co-design of sampling period and buffer size

The paper studies the control of networked control systems (NCS) with sensor-to-controller and controller-to-actuator time delays. The delays are bigger than one sample and random. By using of buffers, the original problem can be transformed into linear system with constant delay. The optimal size of the buffer is an open problem. Increase of the buffer length deteriorates the quality because of longer delay in the control loop. At the same time longer buffer decreases the probability of packet dropping and thus improves the quality. The idea is to make the system more robust by buffering and simultaneously adjusting the control loop sampling and actuation intervals. The paper presents this “co-design” approach and demonstrates it for LQ control problem of cart-pendulum model.

Buffer length selection for distributed control systems

In this paper we consider a state-feedback discrete control of the continuous linear process, where the actuator is connected to the controller via data transmission network. The main complication of this control architecture is the presence of variable time delays. The idea is to reduce temporal dependency of the individual parts of this model by introducing buffer at the actuator. As long as variable delay is shorter than buffer length the loop delay is constant and is equal to the buffer length. If this condition is not fulfilled for some steps, the system operates in open-loop. Introduction of a longer buffer eliminates open-loop operations but decreases control quality. Finding a balance between these two tendencies is the main research problem of this paper.

State and parameter estimation in a hydraulic system - moving horizon approach

In most control systems the best quality of control can be achieved if an exact model of a plant is known and if disturbances-free and continuous (instead of discrete) measurements of all its state space variables are available. Since it is seldom the case in real-life installations, state observers, adapting (parameters identification) algorithms, anti-noise filters and continuous signal estimators must be often employed. The paper describes an application of the moving horizon estimation (MHE) and continuous state estimation from discrete output measurements (CSE-DOM) based method to a laboratory hydraulic three-tank system. The resulting algorithm encompasses all the four above-mentioned mechanisms (state observation, parameters identification, disturbances filtration and continuous state estimation) and can be used for nonlinear systems as well, which is the case in the presented article. The paper provides a detailed explanation of how the MHE and CSEDOM methods were adapted to be applicable to the three-tank system. Results of computer simulations and practical laboratory experiments are also included.

Analogue electronic active filter implementation of an approximate state space realisation of a fractional-order system

The paper presents a procedure of approximate implementation of a given fractional-order system as an electronic analogue active filter. Employment of LIRA method of impulse response approximation allows the filter to be only composed of standard RC components and operational amplifiers. A proposed state variable transformation helps to simplify filter structure and reduce component count while improving its noise properties and accuracy of integer-order prototype frequency response reproduction. The paper provides and compares results of realistic LTspice IV simulations for both implementations based on the LIRA realisation and its transfromed variant respectively. They confirm effectiveness of the presented approach involving LIRA approximation and state transformation steps.

Packet buffering, dead time identification, and state prediction for control quality improvement in a networked control system

Time-varying communication delays, data packet dropping, and other network-induced phenomena are inherent in networked control systems. Their presence can deteriorate noticeably control quality and narrow stability margins forcing designers to adopt conservative controller tuning. Control quality drop can be to some extent remedied by employment of network packet buffering or queuing, finite horizon state estimation, and continual time delay identification methods applied to networked control loops. The paper presents a modular structure for a networked control system where the named techniques are deployed in a network node located on the actuator site. A case study for a DC motor servomechanism and a periodic trajectory tracking problem is given. Simulation results are provided.

Comparison of real-time industrial process control solutions: Glass melting case study

Two solutions supporting advanced control algorithms in real-time are described and compared in the paper: a SCADA-based system and a custom QNX-based system. The control of the glass melting and conditioning process, which is considered to be difficult and demanding for control applications in the glass industry, is provided as an implementation example in the paper.

Active buffer design method for networked control systems

The paper studies the control of networked control systems (NCS), considering controller-to-actuator variable time delays and packet dropout. The idea presented in our previous papers - to reduce delay variability in the loop by introducing buffer at the actuator - is now extended. The controller sends less frequent but more informative packets. The data in each control packet contains the control signal for the current time step as well as the sequence of predicted future control signals, next buffered at the actuator. The active buffer operation algorithm is proposed in the paper. The performance of NCS as a function of the prediction horizon and sampling period, under assumed communication schemes, is investigated. The results of simulation experiments for both unstable and stable equilibrium point stabilization of a pendulum on the cart system are presented.