A Re-Entry Path Planning Method for Service Robots Based on Dynamic Inver-Over Evolutionary Algorithm
, Chaoyong Chen 3, Tianmiao Wang 1,2, Youdong Chen 1, Hegen Xiong 3, Fan Ren 1 and Yu Zou 3Round 1
Reviewer 1 Report
The paper deals with an interesting trajectory planning problem occurring in the application of cleaning robots working in an uncertain dynamically changing environment. The paper is clearly written and understandable. The main contribution comes from the proposed heuristic adjustment of the planned route by the proposed “occupy” operator intended to cope with the blocked cells in the operating space. The provided simulation study indicates the practical applicability of the proposed scheme, although a more thorough examination for different cell configuration and rate of changes would probably be needed to draw any general conclusions regarding the overall performance. Nevertheless, the paper is recommended for publication and some minor comments are given to be considered when preparing the final camera-ready version of the paper:
- Consider replacing “Kinetic Models” in the heading of Section 2.3 by “Kinematic Models”, which is more appropriate in your situation studying only the transforms between the joint and machine coordinates. Kinetics is usually used in the context of dynamic analysis, involving a study of forces and torques acting in a mechanism
- Consider replacing “walking at” at the end of section 2.3 by some other verb, e.g. following. The wheeled robot probably cannot walk in this case.
- The paper aims at providing a solution for real-time capable motion planning. Nevertheless, achieved computational times are not mentioned. Try to add this information in your simulation study to support the plausibility of your results. The mentioned reduction of computational demands compared to restarted execution of the static algorithm could be quantified as well.
- The use of „at home“ in the first sentence of par. 2 on page 1 seems peculiar. The paper is intended for the international audience.
- “Either” is usually used in conjunction with “or” in the next sentence.
- The simulation study presents just one particular planning scenario. Is it possible to include a more thorough study of the algorithm performance e.g. by randomizing the non-cleaned cells locations and the probabilities of cell occupation changes? Without this, it is hard to make any judgment about the overall performance in various situations.
Author Response
Point 1: Consider replacing “Kinetic Models” in the heading of Section 2.3 by “Kinematic Models”, which is more appropriate in your situation studying only the transforms between the joint and machine coordinates. Kinetics is usually used in the context of dynamic analysis, involving a study of forces and torques acting in a mechanism.
Response 1: As suggested by the reviewer, we have changed the title of Section 2.3 to “Kinematic Models for Cleaning Robots”. In addition, the remaining three in Section 2.3 have been corrected.
Point 2: Consider replacing ‘walking at” at the end of section 2.3 by some other verb, e.g. following. The wheeled robot probably cannot walk in this case.
Response 2: As suggested by the reviewer, we have replaced “walking at” by “following”. This change can be found in the revised manuscript of page 5, and line 4.
Point 3: The paper aims at providing a solution for real-time capable motion planning. Nevertheless, achieved computational times are not mentioned. Try to add this information in your simulation study to support the plausibility of your results. The mentioned reduction of computational demands compared to restarted execution of the static algorithm could be quantified as well.
Response 3: We agree on the importance of computational time and have added this point to the simulation study. We made 10 sets of experiments to compare the running time of the dynamic algorithm and the static algorithm. The dynamic algorithm uses Occupy operator to calculate the results, while the static algorithm is to restart by inputting the cell position. Finally, the average computational time of the dynamic algorithm is 0.00526s, while the static algorithm is 0.05847s. This information shows the rationality of the results in the paper. This change can be found in the revised manuscript of page 11, and paragraph 2.
Point 4: The use of “at home” in the first sentence of par. 2 on page 1 seems peculiar. The paper is intended for the international audience.
Response 4: Thank you for pointing this out. Therefore, we have replaced “scholars both at home and abroad” by “domestic and foreign scholars”. This change can be found in the revised manuscript of Section 1, paragraph 2, and line 2.
Point 5: “Either” is usually used in conjunction with “or” in the next sentence.
Response 5: Agree. We have made the correction. This change can be found in the revised manuscript of Section 1, paragraph 2, and line 3.
Point 6: The simulation study presents just one particular planning scenario. Is it possible to include a more thorough study of the algorithm performance e.g. by randomizing the non-cleaned cells locations and the probabilities of cell occupation changes? Without this, it is hard to make any judgment about the overall performance in various situations.
Response 6: We agree that the simulation study is only for the case where the cell position is unchanged and the number of cells is changed. This takes into account that the position of uncleaned cell won't change during the actual cleaning task of the robot. As for the cell occupation probability, we introduced the probability(p) and frequency(f) of environmental change in the “Experimental Setup” in Section 4.2, where f indicates the ratio of the number of cells occupied at the time of environmental change to the overall number of cells. During experimental operations, f is randomly set to be positive between 0 and 0.4. We believe that this initiative demonstrates varying degrees of environmental change.
Reviewer 2 Report
See files
Comments for author File: 
Comments.pdf
Author Response
Point 1: In section 2.3 a kinematic model for cleaning robot is proposed. Skid and slip phenomena are considered negligible. Some comments should be introduced.
Response 1: Thank you for pointing this out. We agree with this comment. Therefore, we have added the point “skid and slip phenomena are considered negligible” to the manuscript. This change can be found in the revised manuscript of page 4, and lines 9-10.
Point 2: From practical point of view, in case of non negligible error between path planned and robot position, has a re-planning step to be considered? In positive case how do the authors manage such a situation?
Response 2: In the actual scene, the cleaning robot needs to reach the uncleaned cell to continue the coverage task. If the error between path planned and robot position isn't negligible, no re-planning steps are required at this time. Because the cleaning robot doesn't need to have as high accuracy as traditional path planning to perform coverage tasks, we accept a certain range of errors when the robot reaches the position of the uncleaned cell. At the same time, the influence of the error between path planned and robot position can be eliminated by sacrificing a bit of repeated coverage.
Point 3: In experimental simulations is not clear if the robot tracks the path planned ? What do the authors mean by “experimental results”? Some details should be clarified.
Response 3: In the experimental simulation, we have given the position of all uncleaned cells, and obstacles between the cells aren't considered. In this case, the robot should track the path planned. Because no route can be shorter than the line connecting two points. We know that the static Inver-Over algorithm can solve the shortest routes of the robot to traverse all uncleaned cells. But it cannot handle the change of the number of cells over time, while the dynamic Inver-Over evolutionary algorithm proposed in the paper can track the changes and calculate the results. In order to verify the dynamic performance of the algorithm, we give the evaluation index of the dynamic Inver-Over evolutionary algorithm in Section 4.1. In an environment where the number of cells is changing, we intercept the number and position of cells at any time as the input of the static Inver-Over algorithm to obtain the optimal solution. And the dynamic Inver-Over evolutionary algorithm with the Occupy operator can track this change to automatically calculate the results. Therefore, “experimental results” is the error of the solution between the static algorithm and the dynamic algorithm. The smaller the error, the closer the solution of the dynamic algorithm at any time to the optimal path solution, the better the algorithm performance.
This manuscript is a resubmission of an earlier submission. The following is a list of the peer review reports and author responses from that submission.
