Local and soft feature selection for value function approximation in batch reinforcement learning for robot navigation.
Fatemeh FathinezhadPeyman AdibiBijan ShoushtarianJocelyn ChanussotPublished in: J. Supercomput. (2024)
Keyphrases
- robot navigation
- reinforcement learning
- feature selection
- continuous state
- temporal difference
- temporal difference learning
- state space
- batch mode
- approximate dynamic programming
- state action
- function approximation
- autonomous mobile robot
- autonomous robots
- machine learning
- reinforcement learning algorithms
- feature extraction
- markov decision processes
- text classification
- basis functions
- map building
- model free
- scene understanding
- policy iteration
- function approximators
- feature space
- evaluation function
- multi class
- dynamic programming
- knn
- optimal policy
- initially unknown
- action selection
- active learning
- training set
- training data
- landmark recognition