Dieser Artikel ist in der englischen Sprache verfasst um SSL-Teams auf der ganzen Welt einen Einblick in unser System zu geben.
Just having good robot hardware won’t score a goal, instead someone has to tell the robots what to do. And that’s where our artificial intelligence comes into play. It’s running on a computer next to the field and uses the vision data to generate commands which then are sent to our team.
To this day, many teams use the famous Skills-Tactics-Plays (STP) design to break down the complex task of deciding, what each robot shall do. STP consists of three layers which each abstract a different part of the game play. The lowest level is the skill layer that provides a basic robot-independent abstraction of the hardware functionality. Building on this there is the tactics layer that provides the necessary means to have a robot work on it’s own, that is to execute an action like shooting, being the keeper or catching the ball. These tactics are combined into a multi-robot play to form a coordinated maneuver, like passing or defending.
In 2013 however, we replaced the „Plays“ layer of our A.I. architecture with an agent-based design. In short each robot is controlled by an independent agent who makes decisions for his robot only. Instead of using predefined global blueprints, we are now able to generate dynamic behavior. All agents can communicate via a messaging system, while team decisions (like the defense coordination or which robot will try to catch the ball) are handled by an external module called „Trainer“.