Rules-Based Learning AI

Personal Project (Artificial Intelligence)

This ant simulator tracks two ant colonies. The ants must avoid poison and gather food and water. Ants on succesful trips return to the colony with a record of their movements. Ants leaving the colony attempt to duplicate past successful trips.

The ants demonstrated some learning behavior, but tended to adapt only once. A time-based decay system for past successful trips or active negative reinforcement may have alleviated this. Another interesting phenomenon was the occasional development of a lethargic ant colony, which moves at a slower rate than normal.

---

DEVELOPMENT STATUS:

Completed 2007

---

VITAL STATS:
Language: C++
Platform: PC
Dev Time: 1-2 Weeks
Designer: Tim Turner

---

NOTEABLE FEATURES:
*Finite State Machine (First Attempt)
*Rules-based learning system (First Attempt)

---

DEMOS AND CONTENT:

Evasion Demo (Demo software):

Instructions: Watch the ant colonies struggle for survival. Tap a key to change performance: Any key - Single update Space - Toggle continuous run mode s - Toggle character graphics on/off r - Display ant colony rules P (Caps) - Drop Poison W (Caps) - Drop Water F (Caps) - Drop Food