Since the first video I have implemented more goal behaviours (specifically Fight Opponent and Escape Complex) for the AI. Once an agent has the four items it will make its was to the exit and "complete the game". It can get the items from the containers or from another agent if he beats it when fighting. If an agent gets beat while fighting it respawns in a random room after a cool-down period.
The short video below demonstrates a four room complex with two AI agents. It is a bit contrieved to quickly show the agents searching for items, the basic combat and escaping the complex once all the items are acquired. It is very rough looking at the moment but all the AI behaviours are working. Functionality before polish...
My next objective is to implement the player object and HUD so that the game can actually be played and interacted with rather than watching the AI entites fight it out. I also intend to tweak the goal evaluation/selection behaviour of the AI entities.
As an aside, I also wrote a blog entry on my own site about working with fonts in Unity3D from C# scripts. I had been meaning to do this since I implemented the text scroll functionality above the AI entities heads as I could not get all this information in one place when searching myself at the time... hope it is useful:
A game I am creating using the Unity3D game development application. It is based on the same concept as Spy Vs Spy (on the NES) except I am expanding the scope by having more agents, randomly generated levels of variable size and two different game modes (search and escape and capture the complex).
I started this game to become a better AI game programmer. I am implementing the techniques from the book "Programming Game AI By Example" by Matt Buckland and am having to convert from C++ to C# which is a lot of fun (in a sad, programmer way).
In this demo you can see Steering Behaviours, Navigation Graph, Path Following and Goal Based Behaviours: