Kernel Taylor-Based Value Function Approximation for Continuous-State Markov Decision Processes.
Junhong XuKai YinLantao LiuPublished in: Robotics: Science and Systems (2020)
Keyphrases
- markov decision processes
- continuous state
- action space
- state action
- finite state
- state space
- reinforcement learning
- average reward
- markov decision process
- partially observable markov decision processes
- optimal policy
- policy iteration
- stochastic games
- reward function
- planning problems
- approximate dynamic programming
- control policies
- temporal difference
- continuous state spaces
- dynamic programming
- average cost
- partially observable
- kernel matrix
- state dependent
- kernel function
- infinite horizon
- basis functions
- kernel methods
- heuristic search
- markov chain
- finite horizon
- real valued
- function approximators
- reinforcement learning algorithms
- dynamical systems
- markov decision problems
- policy gradient
- robot navigation