Structured World Belief for Reinforcement Learning in POMDP.
Gautam SinghSkand Vishwanath PeriJunghyun KimHyunseok KimSungjin AhnPublished in: ICML (2021)
Keyphrases
- reinforcement learning
- belief space
- belief state
- partially observable markov decision processes
- state space
- partially observable
- continuous state
- markov decision processes
- function approximation
- hidden state
- reinforcement learning algorithms
- model free reinforcement learning
- optimal policy
- partially observable markov decision process
- temporal difference
- markov decision process
- partial observability
- point based value iteration
- learning algorithm
- belief revision
- model free
- multi agent
- state and action spaces
- reinforcement learning methods
- machine learning
- policy evaluation
- dynamic programming
- structured data
- planning under uncertainty
- belief functions
- optimal control
- learning process
- control policies
- evidential reasoning
- markov chain
- planning problems
- hidden markov models
- expected utility
- reward function