Enhancing Believability of Virtual Soccer Players: Application of a BDI-Model with Emotions and Trust - Developing Concepts in Applied Intelligence - page 125

shows the soccer field, including two teams of 11 agents (the colored circles) and the
ball (the white circle), which is held by the goalkeeper on the right. On top of each
agent, a small number is displayed, which indicates the player's personal number.
Below, some information is displayed in plain text, to give some insight in the mental
states of the agents. Obviously, there is no space to display all aspects of the agents'
mental states (emotions, moods, trust states, beliefs, desires, and intentions).
Therefore the user can select which information is shown. In the example shown in
Figure 4, five columns of information are shown, indicating, respectively, the current
time, and each agent's number, mood, emotion regarding itself, and current action.
Fig. 4. Screenshot of the RoboCup application.
An example animation illustrating the behavior of the application can be found at
http://www.youtube.com/watch?v=I-ZamC-louo. This movie shows a scenario where
player 6 passes the ball to player 10, who then dribbles with it towards the goal, and
scores. As a result of these successful actions, various aspects of player 10's mental
state change (see the textual information below the playing field): its emotion with
respect to itself increases (from 0.366 to 0.686), as well as its overall mood (from
0.398 to 0.557), its trust in itself (from 0.522 to 0.552), and its belief about its own
capability to shoot at the goal (from 0.509 to 0.602). As a result, in the rest of the
scenario the player will behave more confidently, and will try to score more often.
The application has been tested extensively, using a large amount of different
parameter settings (e.g. for initial trust, emotion and mood states, trust and emotion
flexibility, and action preferences, see also [6]). In all tests, the behavior of all players
was analyzed in a step-wise manner, and compared with expectations of the modelers
(based on common sense). Various interesting types of behavior were observed,