Real-Time Networked Control with Multiple Clients

Date

2010-01-14

Journal Title

Journal ISSN

Volume Title

Publisher

Abstract

In this thesis closed-loop control strategies over a communication network with multiple clients are developed. To accomplish this objective, a steel-ball magnetic-levitation system, a DC motor speed-control system, and an autonomous wheelchair robot referred to as Clients 1, 2, and 3, respectively were used as Networked-Based-control (NCS) test beds to validate the proposed strategies. For real-time operation, Linux with Real-Time Application Interface (RTAI) and Control and Measurement Interface (Comedi) were used as the operating system for Clients 1 and 2. Client 3's software was written in Microsoft Visual Basic 6.0 on the Windows XP operating system (OS). User datagram protocol (UDP) was used as the communication network protocol in this research due to its better real-time performance instead of transmission control protocol (TCP). Although UDP has no guarantee for transferring data, it has smaller overheads and less time delay than TCP. Since the robotic wheelchair and the server are run on different OSs, Samba was used to put both systems into the same LAN with a fast data-transmission speed. Using Samba, the round-trip communication time between the robotic wheelchair and the server is only 11.2 ms whereas 30.8 ms is taken without using Samba. When the server receives the sensor data from multiple clients at the same time, the NCS stability may be deteriorated due to the limitation of the system bandwidth. The NCS stability is affected by the sampling period of the system, and the reduction of the sampling period improves the control loop's performance. However, a shorter sampling period requires more network bandwidth to transmit more sensor data or control data, which increases the network traffic load. Using the PING test, the transmission time for each control loop was measured. The processing time for each system was also measured by a time-stamp function, and the operation time for each control loop was obtained. In order to maintain the NCS stable, several combinations of the sampling periods for each client are suggested and verified. The bandwidth utilization of Client 1 is set to be 43.5% and the range of the bandwidth utilization of Client 2 with guaranteed stability was found to be between 9.1% and 45.3%. Thus, the bandwidth utilization of Client 3 is from 11.8% to 46.8%. The multiple-client NCS test bed could maintain its stability within these ranges of the bandwidth utilizations of all clients.

Description

Citation