Multiagent reinforcement learning (MARL) has achieved a remarkable amount of success in solving various types of video games . A cornerstone of this success is the auto-curriculum framework, which shapes the learning process by creating new tasks for agents to adapt to, thereby facilitating the acquisition of new skills . We argue that behavioural diversity is a pivotal, yet under-explored, component for real-world multiagent learningsystems . We recommend modelling realisticinteractive behaviours in autonomous driving as an important test bed, and recommend the SMARTS benchmark . We list four open challenges for Auto-Curriculum techniques, which we believe deserve more attention from thiscommunity. Towards validating our vision

Author(s) : Yaodong Yang, Jun Luo, Ying Wen, Oliver Slumbers, Daniel Graves, Haitham Bou Ammar, Jun Wang, Matthew E. Taylor

Links : PDF - Abstract

Code :

https://github.com/mtrazzi/two-step-task


Coursera

Keywords : learning - multiagent - curriculum - auto - world -

Leave a Reply

Your email address will not be published. Required fields are marked *