Guided goal generation for hindsight multi-goal reinforcement learning
作者: Chenjia BaiPeng LiuWei ZhaoXianglong Tang
作者单位: 1School of Computer Science and Technology, Harbin Institute of Technology, Harbin 150001, China
刊名: Neurocomputing, 2019, Vol.359 , pp.353-367
来源数据库: Elsevier Journal
DOI: 10.1016/j.neucom.2019.06.022
关键词: Guided goal generationConditional generative modelHindsight experience replayMulti-goal reinforcement learning
原始语种摘要: Abstract(#br)Typical reinforcement learning (RL) can only perform a single task and thus cannot scale to problems for which an agent needs to perform multiple tasks, such as moving objects to different locations, which is relevant to real-world environments. Hindsight experience replay (HER) based on universal value functions shows promising results in such multi-goal settings by substituting achieved goals for the original goal, frequently giving the agent rewards. However, the achieved goals are limited to the current policy level and lack guidance for learning. We propose a novel guided goal-generation model for multi-goal RL named G-HER. Our method uses a conditional generative recurrent neural network (RNN) to explicitly model the relationship between policy level and goals, enabling...
全文获取路径: Elsevier  (合作)
影响因子:1.634 (2012)

  • policy 政策
  • learning 学识
  • generation 世代
  • reinforcement 放大
  • replay 重放
  • generative 生产的
  • agent 因子
  • moving 移动
  • enabling 启动
  • current