I am proud to anounce, that I have finished my bachelor thesis on april 11th. In the practical part now I developed a simulation of a crowded marketplace with autonomous agents that follow the behaviors: Wander, Idle, Sit, Buy something and dynamically start a dialog.
The simulation mainly uses the AI community solutions: Behave (AngryAnt), UnitySteer (Arges Systems) and A* Pathfinding (Aron Granberg). On top of these librarys I developed some extensions.
The main feature behind the scenes is a new Containment Behavior for UnitySteer, that now can handle Obstacles of any shape. I extended the original idea of the steering inventor Craig Reynolds to handle concave corners in the obstacle shape too. In addition to that it is now possible to define the container line for this behavior via the A* grid that is used by the A* pathfinding module. This way there can be made any changes in the world geometry while the container automatically adjustes its position.
I will give more insights of my work in the following weeks and intent to make a stable release of the Containment Behavior to share with the community.