Reinforcement Learning Based Neural Controllers for Dynamic Processes without Exploration.
Frank-Florian SteegeAndré HartmannErik SchaffernichtHorst-Michael GrossPublished in: ICANN (2) (2010)
Keyphrases
- reinforcement learning
- neural network
- action selection
- network architecture
- exploration strategy
- autonomous learning
- multi agent
- state space
- function approximation
- evolutionary robotics
- learning problems
- temporal difference
- balancing exploration and exploitation
- fitted q iteration
- robot control
- process model
- dynamic programming
- learning algorithm