Versions Compared

Key

  • This line was added.
  • This line was removed.
  • Formatting was changed.
Comment: Migrated to Confluence 4.0

Abstract

I read a paper online about a Strategy-based making decision making robot soccer system that utilized a real-time self-organizing fuzzy decision tree (SOFDT). This is because the environment of a robot soccer game is highly dynamic and the system needs to be very flexible and adaptive. The strategy is also event-driven which means that the robots take action based on the factors of their environment, i.e the other participating robots, the ball and their goal. The system was with the NTU-Formosa multi-agent soccer robot system.

System

The image information of the system was captured by a CCD camera ans sent to a corresponding host computer that analyzed the data based on the soccer field. According to the determined situation, the host computer decides a strategy and plans the motion modes and velocities for each soccer robot of the same team. Each robot of the NTU-Formosa then receives velocity commands from the host computer that assigns rotational velocities for its right and left wheels.

Decision Making Process

The system implements decisions in multiple stages. This paper proposed the self-organizing fuzzy decision making system to be applied on the NTU-Formosa system to decide the appropriate action. To generate a practical decision tree, training data was necessary and fuzzy sets were generated from the training data. In order to adapt to the changing environment, the parameters were tuned online in order to improve performance in terms of decision results and input. The basic decision tree looked like this:

Where is the ball?

- If in the offense zone

     - Who is in control of the ball?

            -if teammate

                 do something

             -if opponent

                 do something else

              -if uncontrolled

                 do another thing

-If in Defense Zone

....

Eventually, there were 12 possible options based on the four positions of the ball on the field and the 3 possible control positions of the ball(teammate, opponent or uncontrolled).

Conclusion

The paper went into detail about the development of the SOFDT and compared to traditional decision trees, the SOFDT utilizes less memory space and has a faster generation speed and the ability for realtime adaptability. From their simulations, the results were good.

Link to paper - Robot Soccer System

Robot Soccer System