3
MT-Dreamer: Efficient Multi-Task Replay for Model-Based Deep Reinforcement Learning
We introduce a taxonomy of replay strategies and propose a novel approach to shared environment multitask replay, and combine it with the deep model-based RL algorithm DreamerV2.
Shane Parr
,
Mete Akhbulut
,
Stephanie Milani
,
George Konidaris
PDF
Cite
DOI
Cite
×