Approach

Jump to bottom Edit New page

jeromecremers edited this page Mar 5, 2013 · 1 revision

This is a brief outline of the approach. Please feel free to comment and/or add stuff.

State space:

Which team possesses which control points
Do I have ammo? (maybe we should rephrase this to 'How much ammo do I have?'
How many foes do I observe?

Possible actions/subgoals:

Get ammo
Defend control point (which is already in our possession)
Capture a certain control point

Now we try to do Q-learning on these state-action pairs.