Google Research Football: A Novel Reinforcement Learning Environment — mentioned by nickswalker

Wed, 29 Apr 2020 23:41:30 GMT

That's a fair characterization; the 3D league is meant to incorporate physical constraints, and learning "through" those to the higher level aspects of the game is challenging. That's what the 2D league exists for though. The Google simulator lets you learn from pixels if you want to, but the agent that you're controlling only has 8 actions [1] available to it, so the learning problem here really has no bearing on a robotics application or anything in the real world that I can think of. [1] https://arxiv.org/pdf/1907.11180.pdf

Paper and discussion history. A mention is not an endorsement.

Google Research Football: A Novel Reinforcement Learning Environment · HackerPapers discussions

Google Research Football: A Novel Reinforcement Learning Environment — mentioned by nickswalker