World Modeling for Autonomous Wheel Loaders
Round 1
Reviewer 1 Report
Comments and Suggestions for Authors
This paper introduces a new method using deep learning models to predict the loading performance and pile state changes of automated wheel loaders. This approach can be utilized to optimize the operation of automated loaders, maximizing loading efficiency. Overall, the data is comprehensive, the writing clear and well-organized, effectively conveying the technical information and research findings.
In Figure 5, the difference between "Predicted h'" and "Post-processed h'" appears to be subtle. Is there a way to make the distinction more clear and obvious?
Images and videos from the simulator are found in Fig. 1, Fig. ?? and Supplementary. Is there a typographical error on line 219?
The reviewer suggests changing Figure 8 from a top-down view to an axial side view, as non-professionals may find it difficult to understand the actual shape of the material pile as it is configured.
Post-processing is most beneficial in cases where the…Is there a typographical error on line 405,408?
Although Figure 11 visually compares the effectiveness of the post-processing step through data and error distribution, the discussion of these results in the paper is somewhat indirect. This might lead to an inadequate understanding of the importance of the post-processing step among readers, particularly the significance and necessity of part (b), which should be further explained.
The reviewer expresses skepticism about this statement: "plan the movement of both wheel loaders and haul trucks in a way that is optimal for coordinating … other machines." Please further explain whether the model has this capability.
Referring to the study below will be useful in your discussion of innovations in loader automation technology or the application of trained models in selecting attack pose:
(1) Magnusson, M.; Almqvist, H. Consistent Pile-Shape Quantification for Autonomous Wheel Loaders. In 2011 IEEE/RSJ International Conference on Intelligent Robots and Systems; IEEE: San Francisco, CA, 2011; pp 4078–4083. https://doi.org/10.1109/IROS.2011.6095031.
(2) Chen, G.; Wang, Y.; Li, X.; Bi, Q.; Li, X. Shovel Point Optimization for Unmanned Loader Based on Pile Reconstruction. Computer aided Civil Eng 2024, mice.13190. https://doi.org/10.1111/mice.13190.
The reviewer suggests providing more details about the computational resources required for model training and real-time operation. Can it be deployed on hardware with limited processing capabilities? This would be beneficial for readers to understand the trade-offs between accuracy and computational efficiency in operational environments.
Author Response
Please see the attachment.
Author Response File: 
Author Response.pdf
Reviewer 2 Report
Comments and Suggestions for Authors
The authors developed a method for learning world models for wheel loaders performing automatic loading actions on a pile of soil. The method enables one to create computer models based on deep neural networks trained on data from 3D, multibody dynamics simulation of over 10,000 random loading actions in gravel piles of different shapes. The pile state predictor model was trained in a two-stage process. The accuracy and inference time for predicting the loading performance are relatively good.
My comment are as follows:
1. A glossary of abbreviations is necessary.
2. Better define the symbols - because it is difficult to detect the relationship between the general idea and detailed parameters and formulas, for example, are the a and h parameters time and length or height or something else? – Page 3 - beginning of the article - for readers.
3. What are the units for "work" and what does this term mean?
4. What do you understand by the term „shape k’” – Figure 1?
5. How do you define four scalar parameters for slope and curvature of the piles? – Line 103.
6. Define the parameters from Eqts 1 and 2 better. Which parameters do you give values ​​and what are the units? Better describe the differences between conditions and values.
7. Is the state of the prism expressing its volume? How? Is the relationship between the volume of the pile sought? with which parameters?
8. Algorithm 1 is unreadable - a conceptual, accurate drawing would be good.
9. You must first thoroughly explain the problem in three-dimensional space and time, and only then can you discuss the improvements and rationalization of actions – Figures 3-4.
10. What parameters do the values ​​refer to?: „ 8km/h towards a pile 5m” – Line: 242?
11. Is the term: a (Eqt.12) the four-dimensional vector used in Figure 1. What do the subscripts mean?
12.To understand the text _ Lines 258-271, appropriate drawings and definitions are necessary.
13. What is the relationship between the parameters from formula (12) and the dimensions and markings in Fig. 7?
14. I think that it is necessary to decide on more detailed explanations of the presented relationships and precisely define the parameters - dependent and independent variables and relate them to physical quantities and values, e.g. {hn, an, h′n,Pn}10718n=1.
15. For example, what is the relationship between the angle q and the above parameters?
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Reviewer 3 Report
Comments and Suggestions for Authors
Comments for author File: Comments.pdf
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Round 2
Reviewer 2 Report
Comments and Suggestions for Authors
A glossary of abbreviations seems to be necessary, for example the dependent variable in Equation (15) is not defined.
More detailed quantitative explanations of the existed relationships between dependent and independent variables could be made in the form of analytical dependencies.
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Reviewer 3 Report
Comments and Suggestions for Authors
The authors have not addressed the comments 1 and 4.
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
