Research on AGV Path Planning Integrating an Improved A* Algorithm and DWA Algorithm
Round 1
Reviewer 1 Report
Comments and Suggestions for AuthorsThe paper presents a proposition of combining A* and DWA algorithms as a planner for an indoor mobile robot. The paper is clearly written and easy to follow.
I would wish the authors emphasized the advantage of their proposition over the existing options. Both A* and DWA were studied in numerous papers, moreover, in ROS1 A* is one of the two default algorithms of the global_planner in navigation package and DWA is the default local_planner, so this combination was used very often. Therefore I would expect the authors to compare their result with the version previously published, at least with the ROS1 implementation.
I would however indicate several issue that in my opinion should be changed or modified in a revision to improve the paper:
1. The title contains "in Intelligent Printing Factory", but the paper does not provide any specific characteristic or requirements related to such environment and the algorithm does not include any steps that would be related to such factory - I suppose it is an application the authors aim to, but it seems unnecessary limitation - I would suggest to remove reference to printing factory from the title and conclusions (l.407)
2. The authors seem to ignore the difference between A* algorithm and its popular implementations. In l.69 they claim the paper presents "improved A* algorithm", but it seems they focus on a modification of implementation. The features/issues indicated in paragraph l.114-119 and later on are related with an implementation using square grid with 8-way neighbourhood, not the A* algortihm itself. I would suggest the authors to modify the text to indicate, that an alternative implementation of A* is proposed, and not claiming that the A* algorithm is improved.
3. In their comments about optimality of various modification of A*, including their proposition, the authors seem to ignore the admissibility of the heuristic function. I would suggest to include that in comments and especially - to analyze the admissibility of the proposed function.
4. The description of the proposed function for A* (l.126-131) is unclear and the formula (3) seems incorrect or overcomplicated. In the current form (3) could be written as Q=N/((|X_i-X_t|+1)(|Y_i-Y_t|+1)) with ambiguous definition of N ("N as the total number of obstacle grids between the current position and the target position" - along atraight line? within a rectangle defined by two points?)
5. Paragraph l.132-144 comments on properties of the proposed heuristic function, but the function itself (how Q is incorporated in it) is not defined until l. 153-155. Similarly in l. 149 the authors write about "adjusting the cost function G()" (what sound strange as G is the real cost from the start to current position, so it is fixed, with no possibility to be adjusted), and the fact that it is indeed the actual cost we find in l.156.
6. What do the authors mean by "adjusting G and H in different spaces"? (l. 150)?
7. In the comments about H(n) l.159-166 the authors should refer to admissibility and potential influence on optimality of found path
8. In paragraph l.216-225 the authors provide path smoothing procedure. The procedure is not clear - is it looped? Current description seems to be valid only for the path consisting of 3 segments - could the authors provide full algorithm?
9. It is not clear how the distance in DWA is calculated - eq. (10), l. 282-288; I have not found also other parameters of DWA, especially - time horizon for single iteration planning.
10. In. l.313 dist_sta() is described as "the nearest distance between the mobile robot and known obstacles in the global path nodes" - in is not clear how the distance is calculated if obstacles are unknown; similarly as in the previous comment - the method of calculation is not clearly defined
11. Paragraph l. 334-344 compares "traditional" and "improved" DWA, but the performance of both depend on the parameters used. It would be good to indicate what parameters were used for both, especially as "traditional" version presents a very bad performance (collision)
12. Algorithm in Fig. 8 - it is not clear if path smoothing of A* is used; as the A* is a deterministic approach, how the first loop (in the left part of the chart) is supposed to work(change anything) if the initial and final point remain the same? What does the condition "Global optimal path?" mean and how it is verified?
13. Fig. 9 seems to not match description, according to which the DWA should be passing through the points found by A*, while the path seems to be passing by them - probably a part of the description on the dependance between A* and DWA is ommited and should be explained more clearly
Minor issues:
- l. 187 it seems it should be "n3" not "n7"
- Fig. 5 - segments named in the text (b,c) should be marked in the figure
- eq (7) - to fit the text, velocity should be marked "v_t", not "V"
- eq (13) should be dist_dyna()
- Fig 7 - subfigures (a) and (b) seem switched
- Table 2 unnecessarily split
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Reviewer 2 Report
Comments and Suggestions for AuthorsAuthors propose the improvement of the traditional A* algorithm to improve the search efficiency and path smoothness by adding the raster obstacle rate and improving the heuristic function in the evaluation function of the A* algorithm; to introduce the evaluation subfunction of the nearest distance between the AGV and the known obstacle and the unknown obstacle in the global path in the Dynamic Window Approach (DWA algorithm), to reduce the interference of obstacles to AGV in global path planning. Finally, authors combine the improved algorithms into a new fusion algorithm. The presented experimental results show that the search efficiency of the fusion algorithm is significantly improved and the transportation time is shortened.
The article was written correctly and it raises a very interesting problem.
On the other hand, there are several elements that need improvement:
- References given in the text should be preceded by a space. E.g. text [1].
- Equations presented in the article should be written according to the style of the article. The numbering should be aligned to the right and the spaces before and after the equation are missing.
- Sometimes the figures are aligned to the right and sometimes centered (figures 4 and 6). Please align this according to style.
- Line 334 - suddenly the text enters the left paragraph.
- Figure 7 and 9 alignment problem.
- Line 376 - the table should be numbered and captioned.
- Why is table 2 divided into two?
- The authors present results for improved algorithms. Results for basic algorithms are missing. This would make it easier to evaluate the proposed improvements.
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Round 2
Reviewer 1 Report
Comments and Suggestions for AuthorsThe issues indicated in the previous review were carefully addressed by the authors and clarified in the revised version and submitted response.
But I would ask the authors to verify yet the following, mostly related with editing.
While the authors agreed that the proposed change refers rather to modifying
implementation of the A* algorithm than proposing an alternative to it, the old
wording was kept in l.18-19 ("paper proposes an alternative to the A* algorithm") and l. 70 ("paper proposes an alternative to the A* algorithm"), l.380 ("an alternative to the A* algorithm").
I would suggest to modify to "alternative implementation (or version) of the A* algorithm" (and not an alternative _to_ ).
The authors' proposision is also called "alternative A* algorithm" in the text later - but while not being rigorously precise, it is acceptable after the explanations in l.85-88 and can be used for brevity.
Some minor editorial issues
- missing "*" at A* Algorithm in the title
- I would suggest moving the formula defining H(n) (l.146) up, immediately after the first mention (l.136-137).
- it is not clear why in l.290-293 for X's capital letters are used and for y's - small letters, please consider to unify it if the distinction is not needed
I have also noticed some editorial defects, like missing spaces (especially around mathematical symbols in the text), table splitted between pages or caption on another page than the referred Figure - it maybe only an issue with a commented pdf sent for the review, but I suggest the authors to double check for those issues with the final version.
Author Response
Please see the attachment.
Author Response File: Author Response.docx
Reviewer 2 Report
Comments and Suggestions for AuthorsIn my opinion, the article in its current form can be accepted for publication.
Author Response
Please see the attachment.
Author Response File: Author Response.docx