Contrastive Initial State Buffer for Reinforcement Learning.
Nico MessikommerYunlong SongDavide ScaramuzzaPublished in: ICRA (2024)
Keyphrases
- initial state
- reinforcement learning
- optimal policy
- state space
- markov decision process
- situation calculus
- markov decision processes
- average cost
- stationary distribution
- function approximation
- belief space
- decision problems
- dynamic programming
- partial observability
- heuristic function
- goal state
- partially observable markov decision processes
- finite state
- conformant planning
- heuristic search
- optimal control
- infinite horizon
- action theories
- long run
- markov chain
- machine learning
- planning problems
- dynamical systems
- multistage
- probability distribution
- search space
- multi agent
- learning algorithm