Next Article in Journal
Automatic Cause–Effect Graph Tool with Informal Korean Requirement Specifications
Next Article in Special Issue
Embedded Federated Learning for VANET Environments
Previous Article in Journal
Probabilistic Joint Importance-Based Retrofit Strategy for Seismic Risk Mitigation of Transportation Networks
Previous Article in Special Issue
Coordinated Multi-Platooning Planning for Resolving Sudden Congestion on Multi-Lane Freeways
 
 
Article
Peer-Review Record

Task Offloading Strategy of Vehicular Networks Based on Improved Bald Eagle Search Optimization Algorithm

Appl. Sci. 2022, 12(18), 9308; https://doi.org/10.3390/app12189308
by Xianhao Shen 1,2,3, Zhaozhan Chang 1,2, Xiaolan Xie 1,2,* and Shaohua Niu 3
Reviewer 1:
Reviewer 2:
Reviewer 3: Anonymous
Reviewer 4:
Appl. Sci. 2022, 12(18), 9308; https://doi.org/10.3390/app12189308
Submission received: 19 July 2022 / Revised: 26 August 2022 / Accepted: 14 September 2022 / Published: 16 September 2022
(This article belongs to the Special Issue Vehicular Edge Computing and Networking)

Round 1

Reviewer 1 Report

In this paper, the authors propose a computation offloading method based on improved bald eagle search optimization (IBES) to reduce computing delay and energy consumption in vehicular networks. In general, the presentation of this paper is not good, and the scientific contribution is not clear or very limited since there are a lot of similar works in the topic. I also suspect the correctness of the model and solution. Some major concerns are summarized as follows:

1.       The offloading scenario is not clear and motivated. The offloading workflow must be presented. For example, where is the decision-making entity within your system topology and how are the vehicles interacting with this entity? In other words, which component solve the optimization problem and how to obtain the information for solving the optimization problem?

2.       The authors consider the priority relationship of tasks, but it is only mentioned and not clear how it is addressed in both the problem formulation and solution. Typically, this priority is presented under directed acyclic graphs and takes different approach than the proposed solution.

3.       The authors try to optimize S and F. However, from the presentation of the IBES algorithm, it is not clear how to obtain the optimal S and F? Moreover, in the performance evaluation, f is fixed. So, I think the problem formulation is not correct.

4.       The target environment is vehicular environment; however, the effect of mobility is not considered.

5.       The space and time complexity of the proposed algorithm are not analyzed.

6.       The simulation setting is not sufficiently described. The simulation results are not convincing since there are no reasonable explanations for the improved performance over existing algorithms.

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Reviewer 2 Report

The authors have presented their approach comprehensively and neatly. One minor suggestion is that referring to the simulations as experiments can be misleading to the readers. So, simply presenting the simulation results would suffice as no real vehicular network experiments were performed.

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Reviewer 3 Report

The article proposes IBES, an algorithm to reduce delay and energy consumption during task processing in vehicular networks. For this, the solution defines a framework involving a bio-inspired genetic algorithm that aligns the implementation with the hunting behaviour of the bald eagle. Then, the results are compared with alternatives, concretely PSO and BES algorithms under a set of situations. These contexts study the impact of some variables such as the density of the vehicles, the memory size or iteration times. As a result, the proposed solution provides enhanced outcomes, improving the alternative results and justifying the advance of the paper.

The paper is a very good contribution which reflects in an effective way the proposal and the value of the work. For this, the manuscript defines in a clear way the approach and reviews relevant alternative advances which up the reader to date regarding the matter. Additionally, the integration of the algorithm into a framework relates easily to the application of the solution. Finally, the obtained results manifest better values than the literature.

Considering the contribution and the value reflected in the manuscript, I recommend the paper for publication. However, I would like to invite authors to apply some minor changes to improve the quality of the work. Next, some of these points are described.

In the abstract, concretely at line 11, “MEC” initials are mentioned, however, its meaning has not been explained. To tackle this, I would recommend the following change: “[…] Multi-Access Edge Computing (MEC) […]”.

In the Introduction section, at line 24, when VANET initials are described, “AD” is capital letters. I would recommend respecting the definition of initials: “Vehicular Ad-Hoc Network (VANET)”. This issue is presented multiple times throughout the paper (i.e. line 69). Please, be constant in the way initials are described and avoid redefining them. Initials can be described once at the introduction and referenced in the other sections, excepting the abstract and conclusions

Also, there are some other concepts, reflected as initials, which are not correctly defined and assume that the author already knows them. This is the concrete case of EH at line 55, RSU at line 93, and V2R at line 130…

In the case of the simulation process, I can’t find the range of the communication interface (IEEE 802.11P). Is it the minimum to reach the wide of the lanes?

Thank you very much and good work!

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Reviewer 4 Report

Section 1 does not satisfactorily describe the weaknesses of existing works with respect to the newly proposed solution; authors should add a table, where the features of the analyzed works are detailed, and, then, highlight the novelty introduced by the new schema, in terms of performance, reliability, strategy, and so on.

The system model in Section 2 is a bit trivial. Also, the quality of the figures is poor (mainly concerning Figure 1). In the Same Section, a sequence diagram (or something similar) should help in understanding how the data flow is managed.

In Section 4, the network topology/scenario is not detailed. How many entities are involved? Which could be the behavior in a real environment? Is there an impact on scalability? 

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Round 2

Reviewer 1 Report

None of the previous concerns is addressed in the revision. The authors ignored or answered the questions very superficially. In addition, there are serious flaws in problem formulation and solution. Therefore, the quality of this paper may not be able to be improved to warrant publication.

Point 1: The offloading workflow is still not be illustrated for the practical application of the algorithm in real vehicular network scenarios. For example, where to deploy the algorithm and the interaction between entities involving in offloading to collect enough information for running the algorithm.)

Point 2: As the authors present, it seems just a first come first served model, and no proper task dependencies are considered. Therefore, it cannot be claimed as the contribution of this paper as the authors stated. It may be useful to refer to an example existing work that addresses task dependencies as follows:

X. Fu, B. Tang, F. Guo and L. Kang, "Priority and Dependency-Based DAG Tasks Offloading in Fog/Edge Collaborative Environment," 2021 IEEE 24th International Conference on Computer Supported Cooperative Work in Design (CSCWD), 2021, pp. 440-445.

Point 3: There are serious flaws in the problem formulation and solution. The vehicle-MEC association is not formulated although there are multiple MEC servers. The energy consumption model in (14) seems not correct. No variable can be found in (15), (16), (17), and the objective function in (18), and there is a duplicated sum notation in the objective function. What is the meaning of the third equation (line 267) in C1. The model is full offloading, so C2 is not correct. No resource constraints on the cloud server.

 

For the proposed solution, how to map the variables s and f to the bald eagle position is not clearly presented. Moreover, in the simulation settings (Table 2), the resource allocation f^i_MEC and f^i_cloud is fixed.

Point 4: The authors mentioned that: “Vehicles are randomly distributed on the road, traveling at speeds of 30 to 50 kilometers per hour.” However, the effect of mobility is not considered. Therefore, the performance of the proposed algorithm in vehicular networks cannot be verified. Moreover, one can easily argue that it is not for a vehicular network but just for a general mobile network.

Point 5: I don’t know why the authors refuse to provide the space and time complexity information of the proposed algorithm since they are necessary factors to evaluate the effectiveness of an algorithm. Typically, computation offloading paradigms in vehicular network needs low-complexity algorithms to satisfy latency constraints.

Point 6: The simulation setting is still superficially described. For examples, the arrangement of BS and MEC servers on the lanes (overlapped or non-overlapped communication range), task parameters (di, ci, Timax), cloud transmission latency, energy consumption parameters (e.g., local device power, coefficients), algorithm parameters, etc., all of these are not mentioned. What is the priority matrix that can improve the performance of the proposed algorithm as the authors mentioned in Figure 3? Moreover, there are also no proper references to the simulation settings.

 

As I mentioned, none of my six concerns in the first round has been properly addressed by the authors. First, the manuscript has serious flaws that the authors haven't revised. Second, in a well-studied field such as computation offloading strategies for VEC networks, both the theoretical and practical contributions of this manuscript are limited. More specifically, the limitations can be summarized as follows:


1. No workflow to illustrate the applicability of the proposed algorithm in vehicular network.

2. Task dependencies problem is only mentioned, but not properly addressed.

3. There are serious flaws in the problem formulation and solution as I pointed out.

4. The proposed algorithm is for vehicular networks, but the characteristics of vehicular networks is not considered (e.g., vehicle mobility).

5. The complexity of the proposed algorithm is not presented. So the feasibility of the proposed algorithm cannot be verified for vehicular networks that typically require low-complexity offloading algorithms.
6.  The simulation setting is superficially described. As I pointed out, many important parameters are missing. Also, no exact references to the simulation settings are provided. 

 

In addition to the limitations pointed out, the manuscript also has severe flaws related to performance evaluation.

First, the result of energy consumption for the Local algorithm (Figure 5) is incorrect. As calculated in equation (4), the local energy consumption Eil doesn't depend on the input data size. So, the local energy consumption of the Local algorithm must keep unchanged when we increase the input data size. However, Figure 5 shows that it increases linearly with the input data size, which is surely incorrect.

Second, the objective of the manuscript is to minimize the total cost that is the weighted sum of the latency and energy consumption cost. In other words, it is the tradeoff between the latency and energy consumption. However, the performance evaluation mainly focuses on the latency and energy consumption separately. The tradeoff between the latency and energy consumption is not known.

Third, since [23] addressed a different problem from the manuscript, how to make a performance comparison between the proposal with the PSO algorithm in [23]?

 

 

Comments for author File: Comments.pdf

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Reviewer 4 Report

Author Incorporated all the comments

Author Response

Thank you for your reply.

Back to TopTop