Content of review 1, reviewed on March 31, 2021

Overall comments: This paper addresses the motion planning algorithm, which is one of the main topics in robotic research. The authors proposed novel neural network architecture that generate a function to be used directly as motion trajectories without extensive collision avoidance or prolonged iterative procedures.

The proposed motion planning method sound very promising; however, some improvements in the paper presentation and numerical results are needed: 1) It is necessary to present the developed method in the form of step-by-step algorithm/pseudocode/flowchart. This is to ensure that the proposed motion planning method can be reconstructed by other researchers. 2) Since this paper focuses on the motion planning of the arm robot manipulator, it is necessary to ensure that the proposed method is applicable for other type of manipulators instead of only 2-link and 3-link manipulators, especially the industrial manipulator. For example, PUMA-560 industrial manipulator, which is a standard manipulator for testing the motion planning algorithm, can be used. 3) In the figure 2, the authors compare the configuration results obtained from the proposed method with the Dijkstra’s algorithm. It is better to represent them in the form of a joint angle. For the arm robot manipulator, the configurations may look similar, but they may represent significantly different values of the joint angle. 4) Comparison with A* search algorithm, and Rapidly Random Tree (RRT) algorithm, has been conducted in terms of planning time and trajectory length. It is better to present results of the comparison parameter in the form of table. 5) Equations must be written structurally with numbers. For example: at page 3, a cost-to-go function : f1=Q→C

There are also some minor corrections of the paper as follows: •First section: Introduction and problem definition should be separated to two different sections. •Page 2: Another method is [17] …. Please write the name of method instead of directly write the reference number. There are some similar issues that need to be checked carefully in other paragraphs.

Source

    © 2021 the Reviewer.

References

    Jinwook, H., Galen, X., Ziyun, W., Volkan, I., Daniel, D. L. 2020. Learning to Generate Cost-to-Go Functions for Efficient Motion Planning. ArXiv.