Autonomous Driving Decision Making and Controlling Based on Federated Deep Reinforcement Learning
Deep reinforcement learning shows good performance in autonomous driving but has issues with limited data. In privacy-sensitive cases, data exchange is restricted. This paper proposes a federated framework, discusses its learning strategy and policy, and explores multi-scenario adaptation. TORCS platform experiments confirm its adaptability and better driving performance in lane keeping with data privacy ensured.
这是拖了很久的一篇论文...
最开始是想在毕业前投在期刊上的,但最终由于达不到期待值迟迟没有投出..
后来因为毕业之后edu邮箱被回收,想在arxiv上投稿的想法也破灭了..
Latex转markdown的效果也十分得糟糕,在博客上很难体现出论文的格式..
最终采取了折衷方案的折衷方案——索性直接把代码和论文全都一股脑放到github上!
