Jinsheng Sun

Decentralized robust fault-tolerant control for a class uncertain large-scale interconnected systems

This paper deals with the problem of decentralized robust fault-tolerant controller design for class uncertain large-scale interconnected systems. The system under consideration involves parameter uncertainties. The purpose of this problem is to design the local feedback controllers such that, for all admissible uncertainties as well as sensor failures or actuator failures, the system remains globally stable. The sufficient condition is given for decentralized robust fault-tolerant controller against sensor failure, and the design procedure of decentralized robust fault-tolerant controller is presented. Then, the case of actuator failure is discussed. Finally, an illustrative example and simulation results are given to demonstrate the effectiveness of the proposed method.

D-stable fault-tolerant control for discrete time-delay uncertain systems

This paper focuses on the D-stable fault-tolerant control problem of the discrete time-delay uncertain systems with actuator failures. Using the linear matrix inequality method (LMI), a sufficient condition of the existence of the controller is given to maintain the discrete time-delay uncertain systems against actuator failures D-stable. The design of the D-stable fault-tolerant controller is proposed Finally, an illustrative example and simulation results are given to demonstrate the effective of the design method proposed in this paper.

Adaptive Neuron PID: A New Active Queue Management Algorithm

Active queue management (AQM) can maintain smaller queuing delay and higher throughput by purposefully dropping the packets at intermediate nodes. In this paper, a novel AQM algorithm, called adaptive neuron PID, is proposed, and the detailed statement of this algorithm is given. Simulation results show that the new algorithm is stable and robust for small time-delay network, and the response speed and the performance are obviously better than that of some existing algorithms such as ARED, PI and REM. At the same time, for the networks with large time-delay, this algorithm is also stable

New Active Queue Management scheme based on statistical analysis

Most Active Queue Management (AQM) schemes try to stabilize the instantaneous queue length through mapping the congestion measurement into packet drop probability. In this paper we proposed a new AQM scheme called Exact Dropping Probability (EDP) without visible congestion measurement and mapping. Its developed through statistical analysis for a dumbbell topology network. The simulation results demonstrate that EDP can control the instantaneous queue length converging to the desired value under various scenarios. We compare the performance of EDP with other existing AQM schemes like PI, REM and RaQ, and show that EDP has the fastest reactivity and the smallest oscillation.

An AQM scheme based on Adaptive Weight Cascaded PID controller

In this paper we propose a new Active Queue Management (AQM) scheme based on Adaptive Weight Cascaded PID controller (AWCPID). The inputs of the outer and inner loop PID are queue length error, difference between packets arrival rate and export link capacity respectively. The ratio of outputs of the two PID controllers are dynamically adapted according to the magnitude of their inputs. An adaptive reference packets arrival rate, which is adapted based on network condition, is defined to give the controller more adaptivity. Simulations show that AWCPID can make queue lengths to converge quickly to the target and is effective in a wild range of network conditions. Its advantage over Single Neuron PID is also demonstrated extensively.

A comparison of active queue management algorithms

Active queue management (AQM) is a very active research area in network congestion control. A number of active queue management algorithms such as PI controller, DRED, SRED, Blue, REM, and PD-RED have been proposed in the past few years. This paper presents a comparative study of these algorithms using simulations. The performance metric used in the study is queue length. Our results demonstrate that PD-RED is more effective at stabilizing the queue length, and the performance of PD-RED is obviously better than that of PI controller, DRED, SRED, Blue and REM.

Synthetic X chart for AR(1) autocorrelated processes

Assuming independent observations in the implementation of control charts in Statistical Process Control (SPC) are sometimes violated in practice, especially with high sampling policy. In this paper, we will investigate the properties of the synthetic X chart with data autocorrelation which can be represented as a first-order autoregression AR(1) model. Numerical results show the undesired effect of autocorrelation within subgroups on the charts performance. In order to reduce this negative effect, a good strategy is to get non-neighboring items in each sample to form a new one. The synthetic control charts properties are evaluated when applying this new sampling scheme. With the guideline of this paper, the practitioners can choose the appropriate sampling strategy to account for the autocorrelation on the synthetic X chart.

An AQM algorithm based on Variable Structure PID controller

From control theoretic perspective, the Internet is a time-varying, highly nonlinear system with large time lag. The classic, fixed parameter Active Queue Management schemes are incapable of dealing with this kind of system which makes it necessary to design adaptive algorithms. In this paper we propose a Active Queue Management scheme based on Variable Structure PID controller. Variable Structure PID is a PID-like controller with its parameters being adapted in such way that: when error is big, strengthen the effect of proportional term, weaken the effects of integral and differential terms, so as to accelerate the response speed of system; when error is small, weaken the effect of proportional term while strengthen that of integral and differential terms, so as to enhance the robustness of system and improve steady state behavior. Simulations show that Variable Structure PID achieves far better performance at the cost of only tiny increase of computational complexity.

State feedback control AQM algorithm based on T-S fuzzy model

Considering the nonlinear characteristic of network congestion control system, a network congestion control system model is constructed using T-S fuzzy model, which can approach nonlinear system very well. A new active queue management algorithm is designed using state feedback control based on this model. The design method of proposed AQM algorithm is described in detail. The stability analysis and parameters selection are given. Simulation results show that the proposed algorithm has fast convergence speed, better stability and robustness to network loads varying. The compare with PI, PID, REM shows that the proposed algorithm has fast response speed, smaller steady state queue length variation and better robustness.

On the simulation study of active queue management

Active queue management (AQM) is an active research area of network congestion control. However, the simulation results given by AQM references were obtained in different network scenarios so that they did not have comparability. In this paper, to compare the performance of PI, REM, Q-SAPI and SAPI, a lot of simulations are run under same network scenarios. The simulation results show that the performance of SAPI is obviously superior to that of other three AQM algorithms. At the same time, the results show that an adaptive AQM algorithm is necessary for the nonlinear, parameter time-varying network congestion control systems to obtain satisfactory performance.