Joint Policy Search for Multi-agent Collaboration with Imperfect Information

Neural Information Processing Systems 

To learn good joint policies for multi-agent collaboration with incomplete information remains a fundamental challenge.