Joint Resource Allocation Optimization in Space–Air–Ground Integrated Networks

Round 1

Reviewer 1 Report

In this paper, the authors address the problem of computation resources sharing among various GTs (Ground Terminals), several UAVs (Unmanned Aerial Vehicles) and one satellite.

Based on the UAV-assisted Space-Air-Ground Integrated Network (SAGIN), a communications network architecture considering that the GTs communicate with the satellite through several UAVs, proposed a model for evaluating the computation delay costs in GTs when they demand computation resources from UAV and a satellite. Based on the simulation of the developed model, the authors prove that their model outperforms other existing approaches.

This paper addresses a relevant subject in an era where IoT devices are increasing, and MEC solutions based on aerial networks using UAVs and satellites are a reality.

The paper language is easy to understand, but I recommend that authors improve the written English language (consider, for example, the sentence in lines 109-111). There are various typos to correct.

The article includes an abstract, 5 sections, and the references.

The Abstract introduces the subject addressed in the paper. In the text, the authors refer to the 3-tier network architecture considered and the computational optimization techniques applied to achieve the presented results.

In the Introduction section, the authors present state-of-the-art articles that address different optimization problems in similar network architectures.

In the System Model section, the authors define the model for the considered architecture.

The authors should consider explaining or detailing the computation task concept. Does a computation task include the transmission/reception of data? Is traffic routing or forwarding not considered a computation task? 

In line 110, the ‘I_i’ defines the i-th GT computation task. Is that definition also valid for UAVs and satellites?

In line 115, the authors state that “… the computation tasks can be performed (…) by GTs and LEO satellites” shouldn’t the UAVs also be considered?

The authors referred in text and illustrated (in Fig 1) that UAVs include

a MEC server, but their computation tasks are not defined or explained. What is the role of these UAVs in this model? Why don’t authors consider that UAVs also contribute to the volume of computation tasks?

Line 118 and Equation 2 identify the data computation location. The authors shall explain the difference between D_i^S (eq 1) and D_i³ (eq 2).

In line 120, the authors state that “…  Each UAV is outfitted with a server, allowing each UAV to offer services for GTs …”. How does the defined model consider the computation task generated by UAVs?

In line 128, consider defining the ||  || operator used in the equation 3.

In line 133, consider replacing ‘M’ with a different character or symbol to avoid any confusion with the ‘M’ defined in line 107.

In equation 9, consider replacing the ‘N’ with the ‘M’ used also in equation 8.

The authors shall consider explaining equation 10 and define the ‘F_i’ variable. Several questions arise when analyzing that equation, for instance:

- Is there a cache variable latency only in downlink? What about the uplink?

- How can the satellite generate traffic without a GT request? Can the request latency time be neglected?

- In equation 10, does the local computation delay depend on satellite computation tasks and time spent transferring data from satellite to GT? What do the authors mean by ‘local’? Which local tasks cause that delay?

In equation 12, what is the role of the satellite in UAV service?

Referring to equation 15, consider defining P_i^hov and replacing it with P_j^hov.

In equation 16, the authors shall explain why the UAV hovering energy is 'i' indexed, referring to i-th GT and not to j-th UAV.

In equation 20, the authors shall explain why representing the total delay by L_i^c instead of T_i (referred to in line 110). Is there any difference between them?

Section 3 has the same name as section 2, consider to rename it.

In section 3, System Model, the authors present the formulated optimization problem for the defined model.

In C1, C2, C3, C4, C7, and C8 constraints, the authors shall consider using the sets N and M (refed to in line 109) instead of sets cardinality, N and M, respectively, for defining 'i' and 'j'.

In the C4, consider to include  ∀ j ∈ M.  

In lines 240-243, what is the meaning of '^' over the variables? If the definitions presented in lines 246-248 apply, consider introducing them sooner.

In the section Results and Analysis, the authors characterize a scenario for simulation, perform the simulations and compare the performance of the presented model with other models from the literature.

Related to Figure 2, the authors shall explain why the delay is higher when more UAVs serve the constant number of GTs.

Author Response

Dear Reviewer,
     Thanks a million for your valuable comments and detailed checking to improve the quality of our manuscript. Your insights and suggestions do help us a lot. Revised portions are marked in blue in the paper. The attached PDF contains my responses to your comments point by point. Please review it.            

Thank you and best regards.                                                        

Author Response File: Author Response.pdf

Reviewer 2 Report

The authors of this article focus on unmanned aerial vehicle (UAV)-assisted space-air-ground integrated network (SAGIN) scenarios. Thus, they propose a three-tier network model comprised of satellites, UAVs, and ground terminals (GTs). Edge servers are deployed on UAVs providing cache services. The aim is to minimize the computation task latency. Moreover, an optimization algorithm based on block coordinate descent (BCD) is proposed, the problem is divided into different subproblems this way. These subproblems are solved with McCormick's envelope theory, successive convex approximation (SCA), and convex optimization.

The computation model as well as the system model are adequately presented. Then, the optimization algorithms are sketched in the form of pseudocode. Three schemes are configured, local computation, satellite computation, and fixed UAV positions. These schemes are compared against the proposal regarding satellite computation, computation tasks, maximum transmission power of GTs, and bandwidth, all against the total delay. The results show that, as expected,  the proposal outperforms all three basic baseline mechanisms.

The results obtained are not surprising since the baseline settings are quite simple. So, one may expect applying an optimization framework like the one proposed in this paper would be beneficial in reducing the total delay. Having said that, the proposal is quite interesting regarding UAV-assisted SAGIN. The models proposed adequately characterize the problem stated in the paper. Using the CVX Matlab toolbox is a way to go. However, have you considered validating your proposal with an event-driven simulator? The dynamics of a network like the one in Figure 1 of the paper might be better captured to feed your models. Please, provide some comments on this. Also, consider giving your proposal a name rather than just referring to it as "Proposed solution".

English quality is good throughout most of the paper. I suggest verifying that the acronyms are adequately deployed the first time they are cited. Also, some sentences start after the dot without inserting a space, check line 7 on page 1 for an example. On line 169, it is written "the number of rotors are expressed ...", so please replace "are" with "is". Finally, the units in Table 2, cycles/bits, should be cycles/bit. In line 311, it is written "SGAIN".

Author Response

Dear Reviewer,

      Thanks a million for your valuable comments and detailed checking to improve the quality of our manuscript. Your insights and suggestions do help us a lot. Revised portions are marked in blue in the paper. The attached PDF contains my responses to your comments point by point. Please review it.

Thank you and best regards!

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

The authors improved the paper and explained most of the issues and questions identified in my previous review.

Nevertheless, I shall recommend that authors add some of the answers to my comments (such as your answer to comments #2, #12 and #13) to the text to clarify the system definition and better support the implemented model.

In line 121, the sentence “The computation … ,i.e.“ does not explain equation 2. Consider removing that sentence and moving equation 2 to line 119 before the sentence starting with “Where … “

In line 126, consider removing the “… on the ground…”

In equation 20, consider replacing “α_i,j” with “ a_i,j “

Consider applying my comment #18 corrections in section 3 by replacing the L_i^c and the T_i.

Consider applying my comment #19 corrections in lines 216 and 219.

Author Response

Dear Reviewer,

   Thank you very much for taking the time to review this manuscript. We thank the reviewers very much for the time and effort they put into reviewing the previous version of the manuscript. Your suggestions have enabled us to improve our work. We appreciate all your comments and suggestions.

    We have incorporated the suggestions made by the reviewers. Those changes are highlighted in purple within the manuscript. The attached PDF contains my responses to your comments point by point. Please review it. 

   Thank you and best regards. 

Author Response File: Author Response.pdf

Reviewer 2 Report

Thanks for improving your work according to the comments made in the previous round. I believe the paper is now in shape to be considered for publication.

Author Response

Dear Reviewer,

    Thank you very much for taking the time to review this manuscript. Your insights and suggestions do help us a lot. We thank the reviewers very much for the time and effort they put into reviewing the previous version of the manuscript. Once again, thank you very much for your comments and suggestions.
    
  Thank you and best regards.  

Author Response File: Author Response.pdf