Thank you for your visit. Please explore this page's sections:

Real Time Genetic Algorithm project has been developed by Mihai Virtosu, virtosumihai_at_yahoo_dot_com, student at the Politehnica University of Bucharest. It combines research from the areas of Control Theory, Evolutionary Computing, Genetic Algorithms as well as Software Engineering. It was build as an ASP.NET web application, which can be accessed anywhere from the Internet at this address: http://rtga.crinsoft.ro

This project's purpose is to fine tune PID controllers' parameters (Kr, Td, and Ti) and uses genetic algoritms as a method of optimization. We optimize the overshot and the settling time in such proportion dictated by the user.

As a technical note, the application that wrappes the genetic algorithm is written in C#.NET and ASP.NET, with the algorithm computation back end written in Matlab R13.

Keywords: Control Theory, Evolutionary Computing, Genetic Algorithms, Software Engineering.

Please click here to access the RTGA application now. It is highly recommended, if you use the 1024 X 728 screen resolution or less, that you view the page in Full Screen mode (you can enter Full Screen mode in Microsoft Internet Explorer by pressing F11. You can exit Full Screen Mode by pressing F11 again.)

Back to Top

The input data area is displayed here:

• Process parameters

We assume that our model for the process in a second order transfer function (with two poles) and a non-dominant time constant. In the above picture, 8 and 2 are the dominant time constants and 0.1 is the less dominant time constant (this less dominant time constant makes our system a quick process, to which we can apply the Kessler algorithm in finding a controller).
• Restrictions

Lower and upper limits for each of the PID controller's parameters: Kr, Ti, and Td. These limits are needed by the genetic algorithm to limit its search.
• Achievements

Here is where users define what they want to optimize. They can choose to compute a controller that has the best settling time, or the best overshot, or a combination of these two.
Please note that, generally, a good settling time is harder to achieve than a good overshot.
• Number of generations

This is the number of generations for the genetic algorithm. The greater the number of generations is, the better the results are. Please note, however, that a large number of generations will take more time for the genetic algorithm to complete.
If we want to use this algorithm for on-line optimizations, we should use a small number of generations and the previous information (this feature of using the previous information is explained next in this section)
• "Do not use history" option

If we check this option, the algorithm will start with a random initialization. Otherwise, it will use the previous information.
Beeing conceived as a real time algorithm, at each run it uses, by default, the previous acquired information to produce a better response. For instance, if you set the number of generations to 10 and run this algorithm 5 times it would be equivalent to setting the number of generations to 50 and running it only one time.
This feature has also a drawback. The genetic algorithm uses the previous information when it makes its initialization. Initialization is very important to genetic algorithms. If we choosed the initialization point close to a local optimum, the algoritm would be unlikely to find the global optimum. In our case, if our genetic algorithtm uses the previous information over and over again, it may not find the optimal solution ever. In this instance, if we feel that our algorithm has started on the wrong path, it is better to check "Do not use history" and let the algorithm do a random initialization.

Please click here to access the RTGA application now. It is highly recommended, if you use the 1024 X 728 screen resolution or less, that you view the page in Full Screen mode (you can enter Full Screen mode in Microsoft Internet Explorer by pressing F11. You can exit Full Screen Mode by pressing F11 again.)

Back to Top

Users can interact by pressing these two buttons:

Display the response

Pressing this button will compute and display the step response for the system composed of the process and the current controller, without computing any controller.

This action is useful when users want to see the effect that perturbations have upon the system. By slightly changing one of the time constants (for example, 8 becomes 9), you can press this button and visualize the effect this perturbation had on the system. Typically, it will result in an increased overhead or settling time.

Please note that when the page is displayed for the first time this button is disabled because no controller has been computed yet.

Compute a new controller

Pressing this button will compute a new controller and display the step response of the closed loop system.

Please click here to access the RTGA application now. It is highly recommended, if you use the 1024 X 728 screen resolution or less, that you view the page in Full Screen mode (you can enter Full Screen mode in Microsoft Internet Explorer by pressing F11. You can exit Full Screen Mode by pressing F11 again.)

Back to Top

The step response for the closed loop system is displayed in the right pane. In addition to that, in the left lower pane are displayed some of the parameters for the PID controller and step response.

Kr, Ti, and Td are the PID controller's computed parameters.

The settling time and overshot are the parameters for the step response of the closed loop. A variation from the previous parameters is also displayed in red, beneath each value.

Please click here to access the RTGA application now. It is highly recommended, if you use the 1024 X 728 screen resolution or less, that you view the page in Full Screen mode (you can enter Full Screen mode in Microsoft Internet Explorer by pressing F11. You can exit Full Screen Mode by pressing F11 again.)

Back to Top