Objective

    This project is concerned with the development of a system capable of demonstrating high performance closed-loop control without any physical connection between the control unit and the plant (including actuators and sensors).  It is our feeling that this area of research will prove to be both fruitful and challenging, offering great potential for future work.  At the same time, it is an area in which surprisingly little has been published in the way of results.  This can be attributed to the many complexities involved in an undertaking that requires the fields of communication, computing, and controls to come together seamlessly.  There are a number of difficulties with performing closed-loop wireless control in real-time, which means that there are also great opportunities to contribute to this emerging field.  This project aims to address many of these problems in the years to come, and will hopefully lead to a continuing advancement in the capabilities of wireless control systems.

Why Wireless?

    There are a number of reasons to be interested in closed-loop wireless control.  In traditional industrial control applications, the implementation of a wireless control system offers some obvious benefits.  Consider the advantage gained by simply eliminating the need for any physical connection between the controller and plant.  Wires impose strict limitations, such as their allowable length, as well as requiring environmental factors, such as temperature constraints, to be met.  In addition to these constraints, wires can be costly. Damage to the wiring of a system can often be difficult to locate and expensive to replace.  We compare this against a wireless control solution, in which the limitations on placement of the controller can be expressed simply as a specified radius from the plant.  This allows for a much more mobile and accessible control unit, making modification and maintenance far simpler.  Further, ideas such as the centralization of all major controllers for a single factory could become a reality.  Beyond such applicability to control in industry, closed-loop wireless control has the potential to open up whole new venues to technologically advanced control application.  By removing the constraints that have historically held controllers tied to the systems that they control, a world of new control possibilities come into reach.

Experimental Setup

    The components chosen for this project are aimed at versatility and ease of use rather than specifically addressing a single problem.  For this type of versatility in control and communication, the obvious choice was the personal computer.  The basic idea was to have one computer act as the controller with the other acting as the actuator and sensor for a physical system.  Due to hardware/software constraints the actuator/sensor machine must be a desktop PC, but in order to demonstrate mobility a notebook PC was chosen for the controller.  For communication between these computers, we chose 802.11b wireless network cards rated at 11Mbps.  For the calculation and execution of control and sensing, we settled on the user-friendly environment of MATLAB/Simulink.  To achieve real-time performance with these components, we originally intended to use RTX, which acts as a real-time kernel made for Windows operating systems.  Due to insufficient network card support, however, RTX cannot be used in conjunction with a wireless network.  As a result, WinCon, which runs over the RTX kernel, cannot be used either.  Instead, we have chosen to continue to run on the Win2000 operating system and are using a separate timer board for improved clock resolution.  This setup is now up and running, and has been successfully applied to the control of a furuta pendulum.  For more information on this topic, view the technical approach section of this site.

Results

    The first step in this project involved writing the necessary code in order to communicate data back and forth over a network between the Controller and the Actuator computers.  For this project we make use of User Datagram Protocol (UDP) for all communications.  This protocol provides a connectionless and unreliable method of data transfer, but communications can be carried out much faster than is possible under Transmission Control Protocol (TCP).  Due to the importance of rapid communication rates, this is a tradeoff that should prove beneficial.  The communication code for the controller is encapsulated in a Simulink S-function block which can be treated as a black box capable of sending feedback and receiving control, or vice versa.  On the Actuator computer we use MATLAB-compiled C-code to perform the communications as well as the computationally simple tasks of sampling feedback and outputting the control actions to the DAC board.  Before applying this setup to the control of a physical system, we did a great deal of experimentation with the effects of network delays and data loss, as well as possible remedies for these problems.  Now that a physical plant is in use, we are applying some of these ideas to enhance the performance of our controller.  A more detailed explanation of this work, including the major findings of our simulation and experimental work, can be found on the results page.