Interactive RL via Online Human Demonstrations.
2020
In this paper, we propose a general approach that uses online human demonstrations to directly shape an agent's behaviors. This approach can alleviate the uncertainties caused by human critiques, while at the same time, removing the offline pre-training in most existing learning from demonstration approaches. Using this approach, we also investigate the interplay among different shaping methods for more robust and efficient interactive learning between humans and agents.
- Correction
- Source
- Cite
- Save
- Machine Reading By IdeaReader
12
References
0
Citations
NaN
KQI