Adversarial reinforcement learning for procedural content generation and improved generalization

A system for training a first reinforcement learning, RL, agent and a second RL agent coupled to a computer game environment of a video game, the system comprising:

• a generation module for configuring a first RL agent to iteratively generate a sub-goal sequence in relation to an overall goal within the computer game environment, wherein the first RL agent module generates a new sub-goal for the sub-goal sequence after the second RL agent, interacting with the computer game environment, successfully achieves a current sub-goal in the sub-goal sequence;
• and
• a interaction module for configuring the second RL agent to iteratively interact with the computer game environment to achieve the current sub-goal, wherein each iterative interaction comprises an attempt by the second RL agent for interacting with the computer game environment to achieve the current sub-goal;
• a first update module for updating the first RL agent using a first reward issued when the second RL agent successfully achieves the current sub-goal;
• a second update module for updating the second RL agent using a second reward issued by the computer game environment based on a performance of the second RL agent attempting to achieve said current sub-goal;
• and
• an output module for outputting, once the first and second RL agents are validly trained, a final first RL agent for automatic procedural content generation, PCG, in the computer game environment, and a final second RL agent for automatically interacting with a PCG computer game environment.