You are here: Home / RTLWS 1999-2017 / 
2023-01-28 - 05:34

Real Time Linux Workshops

1999 - 2000 - 2001 - 2002 - 2003 - 2004 - 2005 - 2006 - 2007 - 2008 - 2009 - 2010

Ninth Real-Time Linux Workshop on November 2 to 4, 2007, in Linz, Austria

Implementation of Model Based Networked Predictive Control System

Ahmet Onat, Emrah Parlakay

Networked control systems are made up of several computer nodes communicating over a communication channel, cooperating to control a plant. Since the stability of the plant depends on the timely measurement of plant outputs and timely computation and application of control signals, end-to-end real-time performance is required. Although methods of guaranteeing real-time performance of the computer nodes exist, this is much more difficult to guarantee for communications if readily available network infrastructure must be incorporated. Such networks promise high average performance but fail to provide guarantee on the timely delivery of data. A method which aims to protect the stability of the plant under communication delays and data loss, Model Based Predictive Networked Control System (MBPNCS), has previously been proposed by the authors. This paper aims to demonstrate the implementation of a networked control system on a non-real-time communication network; Ethernet.

MBPNCS approaches the problem by using a plant model to predict a predefined number of future states of the plant and respective control signal for each, to compensate for the possible delay and data loss that can take place during the communication between modules. The system is made up of a sensor node, a controller and an actuator node which periodically sense and transmit the plant outputs, calculate and transmit the control signals, and periodically apply the received control signals to the plant respectively. The predicted control signals are to be applied in the case of loss or delay in communication. A state machine ensures seamless transition from the moment of loss of connection between the nodes, to restored communication status.

For the implementation of MBPNCS, first a compact Real-Time Linux distribution was prepared and its performance measured. Then the necessary hardware and software to control a DC motor was prepared and their parameters identified. The networked control method running on the prepared RT Linux distribution was then implemented on three industrial computers taking the roles of sensor controller and actuator nodes on the network. Communication was done using real-time sockets to improve the response times.

In this paper, we first briefly describe the MBPNCS method, then discuss the implementation, detailing the properties of the operating system, communications and hardware, and later give results on the performance of the Model Based Predictive Networked Control System implementation controlling a DC motor.


Valid XHTML 1.0 Transitional