BARReL: Bottleneck Attention for Adversarial Robustness in Vision-Based Reinforcement Learning.
Eugene BykovetsYannick MetzMennatallah El-AssadyDaniel A. KeimJoachim M. BuhmannPublished in: CoRR (2022)
Keyphrases
- reinforcement learning
- multi agent
- computer vision
- function approximation
- state space
- real time
- human computer interaction
- autonomous mental development
- temporal difference
- model free
- learning problems
- augmented reality
- vision system
- learning process
- machine learning
- optimal policy
- dynamic programming
- visual attention
- learning algorithm
- genetic algorithm
- temporal difference learning
- high robustness
- robotic control