CollaQ: Multi-Agent Ad-hoc team play through Reward Attributional Q-functions

CollaQ: Multi-Agent Ad-hoc team play through Reward Attributional Q-functions

Oct 23, 2020
|
142 views
|
|
Code
Details
CollaQ, a novel way to decompose Q function for decentralized policy in multi-agent modeling. In StarCraft II Multi-Agent Challenge, CollaQ outperforms existing state-of-the-art techniques (i.e., QMIX, QTRAN, and VDN) by improving the win rate by 40% with the same number of samples. In the more challenging ad hoc team play setting (i.e., reweight/add/remove units without re-training or finetuning), CollaQ outperforms previous SoTA by over 30%.

Comments
loading...