Skip to content

Latest commit

 

History

History
30 lines (13 loc) · 1.7 KB

random_papers.md

File metadata and controls

30 lines (13 loc) · 1.7 KB

Intro

In this note, I will briefly summarise some random papers I have read regarding RL.

this article tackles the issue of partial observability of the environment using SLAM to create the game map.

critical components of DRL is memory which remembers the observations from the environment. but it is still too simple in current research. so in this paper, they develop a memory system, which is using a spatially structured 2D memory image to learn to store arbitrary information about the environment over long time lags, with an adaptable write operator that is customized to the sorts of 3D environments that DRL agents typically interact with.

they proposed “Active Neural Localizer”, a fully differentiable neural network that learns to localize accurately and efficiently. and Active Neural Localizer is trained end-to-end with reinforcement learning. similar concept to the neural map.

navigation with RL in the real-world.