A wizard vs. warrior demo game for reinforcement learning, using the Unity ML-Agent.

The red wizard is controlled by AI or Player and the blue warrior is hard-code, which will also rush for the wizard and use his sword to attack the wizard. The wizard has to keep running to keep distance from the warrior and find a chance to fight back with his adept fire ball magic.

1. Get reward when defeat or hurt the warrior.
2. Get punishment (negative reward) when being hurt.
3. Get small reward when still alive.

Two different behaviors of the wizard. the left one is aggressive. the right one is conservative.

Another game. This time the wizard is facing with the attacks from some future civilizations. Unfortunately his magic does not work for the high-tech aircraft. Just try to run away!