ABOUT US
First and main field of our activity is RoboCup Simulation League. But we also prepare LEGO-Mindstorms' laboratories, where we show students how to build their own robot. If you want to read more about our activities in other fields, go to our main page http://autonom.ict.pwr.wroc.pl. There every field has their own web-page concerning only on it. Here you can read only about RoboCup Simulation League player.

The team 'EKA-PWR' representing Wroclaw University of Technology is the only team representing Poland that has participated in official RoboCup Simulation League competitions. So far we take part in:

1. RoboCup European Championship 2000 (competed as WROCŁAW UT),
• 2. RoboCup German Open 2001 (competed as WROCŁAW UT),
• 3. RoboCup World Championship 2002 - Fukuoka (Japan) (competed as WROCLAW 2002).

Team beginning
Robocup simulation league is an interesting example of multiagent system which facilitates testing and comparing various techniques of creating algorithms controlling agents operating in an environment having features similar to the "natural world". Trials have been undertaken at Institute of Engineering Cybernetics, Wrocław University of Technology, to create such algorithms through evolution. Initial research works utilising evolutionary techniques for creating Robocup player control algorithms, were undertaken in the year 1999. They constituted automatic optimisation of player control algorithm parameters using the conventional genetical algorithm. A simple reactive algorithm performing low-level behaviour such as "run to the ball" and "shoot towards the goal", was then used as a test task. Parameters specifying distance of the player to the ball and angle to the goal were used as input (activation) signals. Program structure was fixed and set a priori by the programmer. Only the parameters were subjected to optimisation. This method enabled fine-tuning the program to current characteristics of the objects (player and ball) but required setting general structure of the algorithm. Basic inference of this stage was to indicate the need for automatic program structure generation. In the second stage of testing, genetical programming algorithm was used for automation of reactive controller programming. Unfortunately, due to very large size of solution space and long duration of experiments (evaluation of individual program-person requiring multiple repetition of a section of the match), it became necessary to introduce rough digitization of parameter space. The effect of such an approach was to accelerate the process of evolution to a period of few days, but at the cost of sub-optimality of solutions obtained.

Back to top

TEAM DESCRIPTION
The basic code of 'EKA - PWR' team was created to perform experiments with evolution and genetical programming of control algorithms. The final version is a combination of part of results of these experiments as well as manually coded heuristic behaviours created in conjunction with students of the Electronic Department. The crucial point of our architecture is parallel connection of several low level reaction modules and high level deliberative action planning. Actions elaborated by various modules are subsequently evaluated and arbitrated by superior Task Manager. Such architecture enable us to test and directly compare different control paradigms. At this stage of development, both approaches (reactive and deliberative) are applied mainly in action planning for single player. Cooperative sequences and deliberative planning for small formations are still under construction. Our strategic formation module generates default formation (arrangement of positions of players on the field) depending upon the situation on the field. At that moment, we utilize two basic formations (attack and defence) selected on the basis of which team is in possession of the ball. Default position of the player is parameterised by the current formation, position of the ball on the field and position of the player with respect to the ball.
Team 'EKA-PWR' consists of ten homogeneous soccer players utilising the same general control algorithm and the eleventh specialised player performing the role of the goalkeeper.

Back to top

RESEARCHES WE MAKE
Research works concerned the possibility of simultaneous evolution of both structures as well as parameters of the program. The first experiments with such a complete approach to genetical programming of controller showed that in more complex cases, the problem becomes practically uncomputable (on our computer equipment). We therefore assumed that an important element ought to be the capability of entering the initial knowledge (given by an expert) which could direct and considerably accelerate the course of evolution process. In such a case, the evolution process could have interactive character and would perform three simpler tasks:

• a) verification the quality of solution set by the human expert,
• b) fine-tuning of structural and genetical recombinations of solution sections set by the expert, to improve general structure of the algorithm,
• c) fine-tuning of parametrical and genetical mutations of algorithm parameters to find local matching.

• 1. Evident setting of whole or part of initial population (e.g. by adding programs written by the expert to this population or by generating these programs from sections of code acknowledged by the expert as the most interesting).
• 2. Specific definition of matching function rewarding the solution preferred by the expert.
• 3. Application of specialised genetical operators which instead of "blind" recombinations would utilise the recombination directed by the knowledge set by the expert (e.g. selection of mutation points and intersections in such a way so as to avoid breaking up of useful sections, and simultaneous focusing genetic operators on useless sections or sections generating weaker evaluations).
• 4. Proposing such a program representation that would disable creating solutions incorrect from the expert's point of view (e.g. eliminating endless loops or conditions that will never be true).

Back to top

DOWNLOAD
You can download our Fukuoka' player, to check his behaviour on the sport field. Using this one we went through preliminary (started 85 teams) and achieved 32 position among 42 starting teams.
Read team description for RoboCup 2002 Championship: long and short version.

Back to top

CONTACT US
Pawel Rogalinski & Marek Piasecki
Wroclaw University of Technology
Institute of Engineering Cybernetics
ul. Janiszewskiego 11/17
50-372 Wroclaw, Poland

or e-mail:
pawel@ict.pwr.wroc.pl
trociu@autonom.ict.pwr.wroc.pl

Back to top

BACK TO MAIN AUTONOM PAGE