Asymmetry Elliptical Likelihood Potential Field for Real-Time Three-Dimensional Collision Avoidance in Industrial Robots

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors The presented work is interesting and carries a novelty to a large extent. However, my major concern is that the algorithm has been tested considering a simple scenario.  - Include more complex scenarios that a robotic manipulator encounters in a real scenario such as objects of different sizes, shapes and structures. - It is claimed in the abstract that '... real-time collision avoidance system'. A rigorous analysis of to which extent the algorithm is real-time could be added. - Include a flow chart summarising the proposed AELPF algorithm for clarity in presentation and to maintain the interest of the readers. - How do you compare your method with the one reported in 10.1109/RCAE56054.2022.9995798? - Abstract could quantitatively characterize the performance achieved. Words like 'significantly', 'more efficient', etc. are subjective. How much efficient is efficient? - Page 3, Line 64 mentions that the robot moves in a straight line for 8 milliseconds .. How does this number of 8 ms come? - Pivotal role of robots in industrial setups, mentioned at the start of Section 1 could benefit from the reference 10.1109/SYSoSE.2012.6384195. - Please combine small paragraphs in the paper since they give the impression of a scattered manuscript. e.g. please see Section 2.2. There are paragraphs containing one sentence.  - The caption of Figure 1 should also mention [8] though I appreciate that the text on Line 36 lists the reference. - In Sub-section 3.2.1 (Simulation), mention the specifications of the laptop/machine on which simulations were run. Are they the same as given in Figure 13? Also, please mention the software tool used for simulation.
- Please thoroughly proofread the paper for typos and linguistic improvements e.g. Line 31, '...as negatively impacting '' should be '...as negatively impact ...'

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

The manuscript presents a novel collision avoidance method, Asymmetry Elliptical Likelihood Potential Field (AELPF), for industrial robots and compares it with traditional approaches such as the Vector Field Histogram (VFH) and Follow the Gap Method (FGM). While the study addresses an important issue in robotics and provides an interesting perspective inspired by autonomous driving, several critical aspects require substantial improvement to meet the standards of a high-quality research paper.

One major concern is the lack of an in-depth literature review and contextual positioning of the proposed method within existing research. The paper briefly discusses conventional approaches but does not provide a comprehensive comparison with modern AI-driven solutions, such as deep reinforcement learning-based obstacle avoidance or Dynamic Window Approach (DWA). Recent advancements in data-driven control and real-time adaptive systems suggest that hybrid methods could offer superior adaptability and robustness. It would significantly improve the paper if the authors provided a broader review of advanced learning-based obstacle avoidance strategies and clearly justified why AELPF is preferable over such approaches.

Another critical issue is the mathematical formulation of the AELPF. The manuscript lacks a rigorous derivation of the non-symmetric elliptical potential field, which is essential to understand the underlying principles and stability of the method. While the equations provided describe the repulsive and attractive fields, they do not sufficiently explain how the non-symmetry parameters are determined or how the elliptical shape is optimized dynamically. Furthermore, there is no discussion on whether AELPF could suffer from local minima problems or unstable behaviors in complex obstacle configurations. To strengthen the theoretical foundation, the authors should present a more detailed mathematical derivation, along with a formal stability analysis to demonstrate that the approach avoids undesirable oscillations or deadlocks in real-world applications.

The experimental design and comparative analysis also require substantial improvement. The authors chose VFH and FGM as baseline methods for comparison; however, both of these approaches primarily operate in 2D planar spaces, whereas AELPF is designed for 3D obstacle avoidance. This choice raises concerns about the fairness and completeness of the evaluation. More advanced benchmark methods, such as DWA and deep reinforcement learning-based algorithms, should be included to better position AELPF against state-of-the-art techniques. Additionally, the computational efficiency and real-time performance of AELPF remain unclear. The manuscript does not provide any analysis of computational complexity, execution time, or scalability, making it difficult to assess the feasibility of deploying the method in real-time industrial environments. The authors should include a detailed discussion on the algorithm’s processing speed, computational load, and suitability for embedded systems.

To further enhance the scientific impact of this research, the authors should consider leveraging ideas from recent works on data-driven control and real-time distributed control architectures. For instance, the study by "Data-Driven Learning for H∞ Control of Adaptive Cruise Control Systems" demonstrates how machine learning techniques can optimize control strategies in a dynamic environment. This suggests that AELPF could be further improved by incorporating data-driven learning mechanisms to adapt to varying industrial conditions. Additionally, "Distributed Real-Time Control Architecture for Electro-Hydraulic Humanoid Robots" provides insights into how real-time feedback control architectures can enhance robotic motion planning. The authors should explore whether AELPF can benefit from a more structured real-time control framework to improve its responsiveness and adaptability in practical settings.

Finally, the presentation and writing quality of the manuscript need refinement. The title could be made more explicit to emphasize the novelty of the work, such as highlighting the real-time 3D collision avoidance aspect of AELPF. Moreover, some sections contain redundant explanations and minor grammatical issues, which reduce the clarity of the paper. The conclusion section should also be expanded to include specific directions for future improvements, such as integrating deep learning-based enhancements, improving computational efficiency, and extending AELPF to dynamic, multi-agent scenarios.

Author Response

Please see the attachment

Author Response File: Author Response.pdf

Reviewer 3 Report

Comments and Suggestions for Authors

Very interesting work.   Asymmetry Elliptical Likelihood Potential Field (AELPF) algorithm can reduce the risk of collisions while maintaining operational efficiency. Could you add more results to demonstrate this claim?

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 4 Report

Comments and Suggestions for Authors

This paper proposed an Asymetry Elliptic Likelihood Potential Field (AELPF) algorithm for collision avoidance of industrial robots. In particular, the main feature of this study is that it is designed to overcome the limitations of existing algorithms to enable more sophisticated 3D obstacle avoidance.

However, the proposed method was tested in a relatively simple environment (fixed obstacle), and experimental verification on whether the same performance can be achieved in a complex industrial environment in which multiple obstacles exist is insufficient. In addition, since the setting process of parameters (γ, Ak, etc.) is not clearly described, discussion on whether these values are optimized only in a specific environment or can be applied in various environments is insufficient.

Therefore, the following improvements are needed to increase the effectiveness of the proposed measures.

First, the operational complexity and processing speed of AELPF should be quantitatively compared and analyzed with existing algorithms to more clearly present its applicability as a real-time collision avoidance system.

Second, it is necessary to discuss the parameter optimization method more systematically. Although the current study only states that the appropriate value has been set through experiments, improvement is needed to ensure optimal performance even in various environments by utilizing techniques such as genetic algorithms and Bayesian optimization.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

Comments and Suggestions for Authors

The revised paper has been polished and is in a more presentable form. Recommended for acceptance.

Reviewer 2 Report

Comments and Suggestions for Authors

The current version can be accepted.