A Centralized Routing for Lifetime and Energy Optimization in WSNs Using Genetic Algorithm and Least-Square Policy Iteration

Round 1

Reviewer 1 Report

1.The authors examine the design of a centralized routing protocol for lifetime and energy optimization using GA and LSPI for WSNs.

2.The presented simulation results show that the suggested centralized routing protocol has improved performance in network lifetime and network energy consumption when compared to an existing Centralized Routing Protocol for Lifetime Optimization with GA and Q-learning.

3.The authors have pointed out the directions of future work as emulating the proposed protocol using Mininet considering the real-world parameters of a typical WSN.

4.In general, the paper material can be considered as a basis for a good tutorial material on routing in WSN.

5.It may be reasonable to do some improvements: (i) to add a scheme of the research (in the introduction); (ii) to add a structural presentation of the material on similar works (i.e., as a table and/or figure of approach taxonomy); (iii) to describe the used optimization problems (and to point out the classes of the optimization problems); (a) variables, (b) objective function(s), (c) constraints; (iv) to add simplified illustrative examples for the solving process; (v) to point the suggested novelty; (vi) to add references on additional survey papers on the considered topic(s).

Author Response

Thank you so much for your Comments and Suggestions. These are addressed as follows:

(i) To add a scheme of the research (in the introduction): This is addressed on Line 119 to Line 133 on page 3.

(ii) To add a structural presentation of the material on similar works (i.e., as a table and/or figure of approach taxonomy): This is addressed on pages 11 to 13.

(iii) To describe the used optimization problems (and to point out the classes of the optimization problems); (a) variables, (b) objective function(s), (c) constraints: This is addressed from the line 461 to 474 on page 16.

(iv) To add simplified illustrative examples for the solving process: This is contained in the designing of the proposed protocol from pages 17 to 19.

(v) To point out the suggested novelty: The novelty is the usage of LSPI instead of Q-learning.

(vi) To add references on additional survey papers on the considered topic(s): No additional survey papers were considered.

Reviewer 2 Report

1. Genetic algorithms can be slow to converge on an optimal solution. Do the authors assert that their GA implementation always converges to a solution "quickly"? There is a cost to the sensor nodes for waiting on the sink to provide the routing information.

2. The authors restate a conjecture that GA removes NP-hardness from MST problems. Do the authors assert that their approach is guaranteed to work on all MST problems in polynomial time?

3. Line 218. Should that read 1/M and not 1/m?

4. There does not appear to be room for a prioritization scheme. Some sensor nodes (e.g., safety critical) may be waiting to transmit more valuable than others (e.g., housekeeping). The authors need to address this.

Author Response

Thank you so much for your Comments and Suggestions. These are addressed as follows:

1. Genetic algorithms can be slow to converge on an optimal solution. Do the authors assert that their GA implementation always converges to a solution "quickly"? Yes as provided in Figure 5 on page 21.                  There is a cost to the sensor nodes for waiting on the sink to provide the routing information. Yes, but the proposed protocol only considers the data packets in the modeling. But, the cost of the control packet will be considered in future work as stated in lines 629 to 632 on page 26.
2. The authors restate a conjecture that GA removes NP-hardness from MST problems. Do the authors assert that their approach is guaranteed to work on all MST problems in polynomial time? Yes as provided on lines 539 to 542 on page 21.
3. Line 218. Should that read 1/M and not 1/m? This is addressed on line 233 on page 7.
4. There does not appear to be room for a prioritization scheme. Some sensor nodes (e.g., safety-critical) may be waiting to transmit more valuable than others (e.g., housekeeping): The proposed protocol was implemented assuming an equal prioritization scheme. However future work we considered different prioritization as stated in lines 629 to 632 on page 26.

Reviewer 3 Report

The paper presents the design of a centralized routing protocol for lifetime and energy optimization using GA and LSPI for WSNs. This is an extension from a previous published work "Centralized Routing for Lifetime Optimization Using Genetic Algorithm and Reinforcement Learning for WSNs ". Although this last paper is mentioned and compared with the present design, the content extracted from the previous work must be highlighted. In addition, you should add results of the computational time and CO2 footprint of your designs. A comparison table might help to differentiate all the related work. In order to understand when a node is deactivated, you can explain it within the Figure 4, and where the sink is located as well. Figures 2, 3 and 4 look blur and too big, please correct. 

Author Response

Thank you for your comments and suggestions. These are addressed as follows:

1. The content extracted from the previous work must be highlighted: This is captured in line 408 on page 13.
2. Results of the computational time and CO2 footprint of your designs: The computation time results are captured in Figure 9 and Figure 12 on page 24 and page 26, respectively. Statement on CO2 footprint is contained on lines 579 to 580 on page 23 and lines 605 to 607 on page 25 for increasing initial sensor node energy and packet generation rate, respectively.
3. A comparison table might help to differentiate all the related work: This has been provided in Tables 1 to 3 on pages 11 to 13.
4. In order to understand when a node is deactivated, you can explain it within the Figure 4, and where the sink is located as well: This is done on lines 549 to 550 on page 21.
5. Figures 2, 3 and 4 look blur and too big, please correct: Corrected.

Round 2

Reviewer 3 Report

The manuscript has improved considerably wrt the previous version. No more comments must be addressed.