CHAPTER 7 - CONCLUSION

7.1 Chapter Introduction

This final chapter attempts to conclude the findings of the experiments conducted. Problems that were encountered in the project are also recorded and a short discussion on further works which could be done on this project.

7.2 Summary of Experiments

The agent architecture designed and implemented with the testbed is a simple system with a number of features to facilitate conducting of experiments. It is generally adequate for conducting experiments on emotions. However, a list of possible improvements as stated in the later section can further enhanced the system

A summary of the experiments were,

A high value in emotions variables such as Joy and Pride, can increase the team's aggression as players have more confidence in their team mates (Pride) and they feel happier (Joy) resulting in more throws, which is more offensive play. A low value can trigger off aggression as anger (Appreciation) increases and this can result in more runs forward. However, having aggression does not imply a definite win as physical abilities also have an effect on the results. A possible application of this feature is in a bid and award system where an agent may be aggressive in bidding but without the adequate resources, it will not be awarded the bid.

The third experiment shows that emotion stabilisation exists in the simulation when there are no extreme and unexpected hitch. This could be useful in a environment where agents are expect to be consistent unless provoked. However, more experiments are necessary for emotion stabilisation studies.

7.3 Problems encountered

There are numerous problems which were encountered during the course of the project. They were,

Basically, the scope of this project is rather extensive as it involved a number of decisions and tasks like choosing a emotions theory, selecting a type of agent and designing its architecture, designing and implementing a testbed with the architecture, testing the simulation and running experiments on it.
A compromise had to be reached between working on the realism of the simulation and spending time on experiments and emotions related work. A portion of the limited time had to be spend on making the game in the simulation realistic. This was a major problem as the game must be sufficiently real for experiments to be ran and thus leaving little time for the experiments itself.
Choosing the right values for the parameters in the parameter file was another problem. Experiments within experiments had to be carried out to see so that the parameters chosen will not yield extreme results and inaccurate which are impossible to be interpreted.

7.4 Future pointers

In view of the time-constraint and concern of this project, the following are some directions for future works.

In this simulation, only upper and lower thresholds were introduced to somewhat represent emotion intensity and form a guide as to when emotion affects actions. To simulate emotion intensity much better, a full spectrum of thresholds is necessary and can be incorporated.
Emotions has many facets. In this simulation, only one reaction to an event is explored. For example, when a goal is scored, the losing team can feel demoralise, desperate or reproach against the other team. Many different reactions are possible towards the same event and this can be an area of study.
The third category of attitude in the semantic-based theory can be implement. Players often have likes and dislikes for other players and it would be interesting to study the change in performance with this variable introduced.
The number of experiments conducted and the accuracy of the results were constraint by time. More experiments with different parameters settings are necessary to explore the possible undiscovered emotions trends and its relationship with behaviour.
A simple deliberative component can be added to model team strategies and emotions can be studied for its effect on these strategies. The goal-driven component in the Tok architecture can be adapted.

7.5 Final Word

This sums up the report and the project. Working on this project had been a extremely challenging, sometimes frustrating and at other times exhilarating experience. It has truly been a knowledge adventure, to venture into a topic that I know nothing about in the beginning, designing, implementing, testing and finally running experiments on it. I sincerely hope that others will continue this adventure, completing the tasks which time has forbade me to accomplish.