Oct, 2022
VER:基于策略的强化学习扩展导致在具身重组中出现导航
VER: Scaling On-Policy RL Leads to the Emergence of Navigation in Embodied Rearrangement
Erik Wijmans, Irfan Essa, Dhruv Batra
TL;DRVariable Experience Rollout (VER) is a reinforcement learning technique that scales on-policy learning in heterogeneous environments to many GPUs, leading to faster navigation and mobile manipulation tasks with surprising out-of-distribution generalization.