Nonlinear Model Predictive Congestion Control for Networks
Cunwu Han*, Mengqi Li*, Yuanwei Jing**, Lei Liu*, Zhonghua Pang*, Dehui Sun*
*Key Laboratory of Beijing for Fieldbus Technology and Automation,
North China University of Technology, Beijing 100144, China,
(e-mails: cwhan & zhpang & sundehui@ncut.edu.cn,limengqi0909@163.com,liulei_sophia@163.com)
**College of Information Science and Engineering, Northeastern University,
Shenyang 110004, China, (email: ywjjing@mail.neu.edu.cn)
Abstract: This paper addresses congestion control for networks. The main focus is the modeling and
compensation of nonlinear disturbance in the network, which is seldom considered in the literature. A
new congestion control method is presented to compensate for the effects of nonlinear disturbance,
uncertainty, time-varying delay, and input constraint. A state feedback congestion controller is designed
via model predictive control approach. The stability of the closed-loop system is analyzed by using
Lyapunov-Krasovskii functional. The effectiveness and feasibility of the proposed controller are verified
by simulation results.
Keywords: Congestion control, model predictive control, nonlinear disturbance, uncertainty, time-varying
delay, input constraint.
1. INTRODUCTION
Congestion is a serious problem for networks (Jacobson,
1988; , which will reduce the quality of
service (QoS). In order to avoid congestion in transmission
control protocol (TCP), active queue management (AQM) is
an effective congestion control approach, and a lot of AQM-
based congestion control methods have been reported in the
literature, such as random early detection (RED) (Floyd and
Jacobson, 1993; Chen and Yang, 2009), proportional-integral
(PI) (Hollot 2002), proportional-integral-derivative
(PID) (Fan 2003), coefficient diagram method (CDM)
(Bigdeli and Haeri, 2009a), adaptive technique (Barzamini
2012; Patete 2008; Zhang 2003), robust
control (Chen and Yang, 2007; Quet and Özbay, 2004; Tan
2007; Zheng and Nelson, 2009), and sliding mode control
(Ignaciuk and Bartoszewicz, 2011, 2012). However, all the
proposed controllers did not consider constraint nonlinearity
in input. In fact, the drop probability considered as an input is
limited between 0 and 1 in real TCP/IP networks. Therefore,
the effect of a saturating input should be considered.
In order to deal with the input constraint, Chen (2007)
presented a robust congestion controller for TCP/AQM
system, but did not consider the uncertainty and disturbance;
moreover, they only considered the constant time delay in
state, while did not consider the time delay in control input.
Additionally, the congestion controllers mentioned above are
designed for continuous time systems. Sall (2011)
presented a robust digital congestion control algorithm, but
did not consider the uncertainty and disturbance.
It is well known that the model predictive control (MPC) has
strong ability to cope with time delay, uncertainty, and input
constraint (Kothare 1996; Mayne 2000, Yan and
Bitmead, 2005). A congestion controller based on MPC was
presented in Wang (2012). An implementation method
of MPC as a digital controller was investigated in Marami
and Haeri (2010). A predictive functional control (PFC) was
introduced as a new AQM controller in Bigdeli and Haeri
(2009b). However, the nonlinear disturbance, uncertainty,
input constraint, and time-varying state and control input
delays are not considered in these predictive controllers.
In the previous work (Han 2014), we presented a model
predictive congestion control for networks with uncertainty,
input constraint, and time-varying delays in both state and
control input, but did not consider the nonlinear disturbance.
In fact, from the dynamics of the AQM (see details in the
next section), the nonlinear disturbance should be considered
in the design stage of the controller. To the best knowledge of
the authors, the nonlinear disturbance in congestion control
system is seldom considered in the literature.
This paper presents a model predictive congestion control
method for networks with nonlinear disturbance, uncertainty,
input constraint, and time-varying state and input delays. The
state feedback congestion controller is designed by using
linear matrix inequality (LMI). Simulation results show that
the proposed method has good performance. In addition, the
discrete-time congestion control model is used to obtain a
digital MPC controller directly.
This paper is organized as follows: The problem formulation
is described in Section 2. A model predictive congestion
controller is designed in Section 3. Simulation results are
displayed in Section 4. And conclusion remarks are
addressed in Section 5.
This work is supported by National Natural Science Foundation of
China (61573024, 61673023) and Beijing Natural Science
Foundation (4154068, 4152014).
Preprints of the 20th World Congress
The International Federation of Automatic Control
Toulouse, France, July 9-14, 2017
Copyright by the
International Federation of Automatic Control (IFAC)
554