PODDP: Partially Observable Differential Dynamic Programming for Latent Belief Space Planning.
Dicong QiuYibiao ZhaoChris L. BakerPublished in: CoRR (2019)
Keyphrases
- belief space
- partially observable
- dynamic programming
- state space
- markov decision processes
- markov decision problems
- infinite horizon
- reinforcement learning
- partial observability
- belief state
- decision problems
- dynamical systems
- partially observable markov decision processes
- optimal policy
- classical planning
- fully observable
- planning under uncertainty
- planning problems
- latent variables
- heuristic search
- orders of magnitude
- dynamic environments
- long run
- optimal control
- state variables
- symbolic model checking
- finite state
- initial state
- reward function
- planning domains
- sufficient conditions
- multi agent