Automatic Control Telelab

Frequently Asked Questions

I tried to execute a remote experiment but the Java interface to control the experiment did not appear.
There are 2 known reasons for this kind of problem.
1. Your computer is behind a firewall. Since ACT tries to connect through a specified TCP port (port 2000), firewalls do not allow this. You have to set your firewall to allow communications on such a port.
2. Your browser does not a have a Java Virtual Machine (JVM), or your JVM is not working properly. A solution consists in installing a working JVM for your browser. You can download it at the following address
www.java.com.

When I try to execute an experiment, an error occurs (Connection with server lost. Program aborted).
Usually, this error is caused by the reasons described in the previous answer. Sometimes this error can happen the second time you run an experiment; in this case, probably, your browser does not close properly the previous connection or does not refresh its cache, and so it fails. Try to close and reopen your browser.
In case of user-defined controllers, please look at the following point.

I design a user-defined controller. The compilation was successful. When I start the experiment an error occur (Connection with server lost. Program aborted). Why?
This error is probably due to a runtime error during the controller execution. A common reason regards continuous-time controllers containing (stable) poles too far from the imaginary axis, i.e., with time constants too small with respect to the execution sample time. A solution consists in changing the position of such poles.
Brief explanation. Suppose the sample time is set to 0.01 seconds. This means to set the field "Fixed Step -> fixed step size" with such value in the Simulink Solver Option. If your controller has one pole in -10000 (i.e., a time constant of 0.0001), it is likely that Simulink will produce an error since it is not able to compute the controller output by using this sample time.

During an experiment, time plots do not seem to reflect the real experiment dynamics. For example, if I set a sinusoidal wave (1 Hz) as reference, the reference plot is a line and not a sinusoid. Why?
Since online data plots are updated every second, signals at high frequency may produce aliasing. Download the data file at the end of the experiment to draw more accurate plots (in this file, data are recorded at a smaller sample time).

During an experiment, live video is not in real-time. Why?
This is a common problem due to video streaming over the Internet. The number of fps (frame per second) of the video can vary depending on your Internet connection (by a modem, a LAN, ...) and on the network load.

Some process models provided in the description pages do not correspond to the real processes. A controller which works well on simulation, sometime has poor performance on the real process. Why?
The reported models are in general simple, and they could not be very good. Moreover, some processes are (slowly) time variant, i.e., their parameter change with time. If a user needs more accurate models, we invite him/her to build a new model by performing ad hoc experiments. In fact, the ACT may be easily used for system identification experiments by using a special interface provided for some processes or by choosing the input signals manually (in this case the user can use the controller template and connect the input blocks to use directly to the "command" port).