After #440 and #441 are complete, and some sort of basic agent can be successfully evolved for a specific game, try seeing how behavioral diversity or something similar works in the same game. If this is at all successful, then the long-term idea is to scale up to novelty searching across all of the games, or to do something like MAP-ELITEs where each separate game is one of the bins/niches.
After #440 and #441 are complete, and some sort of basic agent can be successfully evolved for a specific game, try seeing how behavioral diversity or something similar works in the same game. If this is at all successful, then the long-term idea is to scale up to novelty searching across all of the games, or to do something like MAP-ELITEs where each separate game is one of the bins/niches.