Variable Compliance Control for Robotic Peg-in-Hole Assembly: A Deep-Reinforcement-Learning Approach
and Kensuke Harada 1,2Round 1
Reviewer 1 Report
The authors of this paper presented d a learning framework for position-controlled robot manipulators. The topic of the paper seems relevant to scope of journal and interesting. But, there are some issues to be addressed for revised version.
1) The English need to be improved.
2) There is no comparison with scheme withing RL. It is better to show how much improvement has achieved rather than scheme with RL.
3) There are many references without any detailed explanation. Thus, it is demanded to make reference short with essential literature and describe their contribution clearly.
Reviewer 2 Report
This is a very interesting study applying reinforcement learning to industrial robots. Furthermore, it addresses an important issue related to uncertainty in robot manipulation.
I think this is a well-worth research effort that deserves to be published.
I have just some minor issues. While the authors state they develop the present study based on their previous work, there is more emphasis on that study than on the new contributions (which exist). For example, figure 4 could be adapted to clearly exhibit the novel contributions, pinpointing what is previous work and where are the new one.
I suggest a careful read by the authors as there are some minor language issues. E.g., “… We build on top of our previous work [31] <by> adding a more…”
