Target Positioning and Tracking in WSNs Based on AFSA
Round 1
Reviewer 1 Report
The authors have evaluated indoor space target location and tracking by using AFSA. I recommend the following revisions.
1. The results show that the more the number of AF used in algorithm, the better the accuracy of target positioning. This is understandable but two important things to address is the complexity and edge processing. The authors should comment on the hardware complexity of the WSNs. And how the improvement in the results is compare to previous methods used in the literature.
2. The authors should also elaborate the reason of not getting the improved performance with adaptive step size algorithm, despite of the improvement reported in ref [22]. What is the limitation in current scenario?
3. Table 8 needs explanation on the comparison of fixed vs. adaptive step size.
Author Response
Response to Reviewer 1 Comments
The authors have evaluated indoor space target location and tracking by using AFSA. I recommend the following revisions.
Point 1: The results show that the more the number of AF used in algorithm, the better the accuracy of target positioning. This is understandable but two important things to address is the complexity and edge processing. The authors should comment on the hardware complexity of the WSNs. And how the improvement in the results is compare to previous methods used in the literature.
Response 1: Thank you very much for this comment. A description of the complexity and edge processing has been added in Section 1 of the revised edition.
“Compared with traditional wireless sensor networks, with the rapid development of the Internet of Things (IoT), multiple data collection and computational task wireless sensor networks will face some new challenges. For example: WSN hardware complexity and software computing requirements will increase. Therefore, combining edge computing in the future is crucial for many WSN applications. It can reduce latency and bandwidth usage, although it may lead to higher costs. However, it is helpful to improve overall performance and reduce energy consumption. “
The performance comparison between the proposed methods and the traditional ones has been described in Section 5.1 and 5.2 for target positioning and tracking respectively.
“ For target positioning, the proposed AFSA algorithms, P7 and P8, combined with RSM and HAVP have better performance in both average error and average positioning time than the traditional methods, P1 and P2, in random deployment for target positioning. Especially in P8, the average positioning time is reduced by 81.9% compared with P2. “
”The average performance of the related modes using HAVP is shown in Table 17. It shows that using HAVP alone can improve the tracking success rate without increasing the positioning time. When using both DAFS and RMS for target tracking, the average positioning time of K8 is shorter by 42.47% than that of the traditional one, K1, and the average success rate remains the same as that of K1."
"Table 17 Average positioning time and success rate for K1, K3, K7, K8.
|
Parameter |
K1 |
K3 |
K7 |
K8 |
|
Average positioning time (sec) |
0.0584 |
0.0574 |
0.0422 |
0.0336 |
|
Average success rate |
97.8% |
99.8% |
97.2% |
97.4% |
“
Point 2: The authors should also elaborate the reason of not getting the improved performance with adaptive step size algorithm, despite of the improvement reported in ref [22]. What is the limitation in current scenario?
Response 2 : Thank you very much for this comment. The reason why the adaptive step size algorithm does not improve the performance is that in the simulation environment of the wireless sensing network in this study, when there are a large number of AFs in the algorithm, whether it is fixed or adaptive step size algorithm can satisfy the need to move AFs in the algorithm to the region of the global optimum solution within 100 iterations. On the contrary, when the number of Afs in the algorithm is fewer, such as reducing to 12, no matter the fixed step size or the adaptive step size, the AFs in the algorithm cannot move around the global optimum solution after 100 iterations. It has been described in Section 5.1 in the revised edition.
“ There is no significant difference between fixed step size and adaptive step size and visual field on positioning efficiency. According to Table 5 and Table 6, the values of all methods using fixed and adaptive step size are averaged respectively, and the results are shown in Table 8. It can be seen from Table 8 that the simulation results of the fixed step size algorithm or the adaptive step size algorithm have little difference in the average positioning error and positioning time. The reason is that in the simulation environment of the wireless sensing network in this study, when there are a large number of AFs in the algorithm, whether it is fixed or adaptive step size algorithm can satisfy the need to move AFs in the algorithm to the region of the global optimum solution within 100 iterations. On the contrary, when the number of Afs in the algorithm is fewer, such as reducing to 12, no matter the fixed step size or the adaptive step size, the AFs in the algorithm cannot move around the global optimum solution after 100 iterations. ”
Point 3: Table 8 needs explanation on the comparison of fixed vs. adaptive step size.
Response 3: Thank you very much for this comment. The explanation on the comparison of fixed vs. adaptive step has been added in the revised version in Section 5.1.
“There is no significant difference between fixed step size and adaptive step size and visual field on positioning efficiency. According to Table 5 and Table 6, the values of all methods using fixed and adaptive step size are averaged respectively, and the results are shown in Table 8. It can be seen from Table 8 that the simulation results of the fixed step size algorithm or the adaptive step size algorithm have little difference in the average positioning error and positioning time. The reason is that in the simulation environment of the wireless sensing network in this study, when there are a large number of AFs in the algorithm, whether it is fixed or adaptive step size algorithm can satisfy the need to move AFs in the algorithm to the region of the global optimum solution within 100 iterations. On the contrary, when the number of Afs in the algorithm is fewer, such as reducing to 12, no matter the fixed step size or the adaptive step size, the AFs in the algorithm cannot move around the global optimum solution after 100 iterations. ”
Reviewer 2 Report
The article is devoted to using uses an artificial fish swarm algorithm and the received signal strength indicator channel model for indoor target positioning and tracking.
In general, the structure of the article is good. But some of the results need to be improved
- "4.2. Simulation Analysis on Target Location"
It seems to me that formula (12) is incorrect - is the letter X large or small, or is this a criterion? If a criterion, then there must be a maximum or minimum. The entire section 4.2 should be rewritten to be readable by the reader.
- "4.3. Simulation Analysis on Target Tracking"
Lots of obscurity. How are K1 ... K8 counted? What exactly is done in the MATLAB program
- Discussion section required
Author Response
Response to Reviewer 2 Comments
The article is devoted to using uses an artificial fish swarm algorithm and the received signal strength indicator channel model for indoor target positioning and tracking.
Point 1: In general, the structure of the article is good. But some of the results need to be improved
Response 1: Thank you very much for this comment. Section 4 has been rewritten. The Section 5 Discussion have been added in revised version to make the paper easily readable.
Point 2: "4.2. Simulation Analysis on Target Location" , It seems to me that formula (12) is incorrect - is the letter X large or small, or is this a criterion? If a criterion, then there must be a maximum or minimum. The entire section 4.2 should be rewritten to be readable by the reader.
Response 2: Thank you very much for this comment. The letter X in formula (12) is an uppercase letter standing for the position of estimated or target point. The lowercase letters have been corrected in the revised edition. Section 4.2 has been rewritten into two parts 4.2 simulation results and 5.1 Target Positioning discussions in the revised version.
Point 3: "4.3. Simulation Analysis on Target Tracking" , Lots of obscurity. How are K1 ... K8 counted? What exactly is done in the MATLAB program
Response 3: Thank you very much for this comment. Due to the moving characteristics of the target in the target tracking system, the target moving speed is 4~7 m/s. In other words, the AF needs to move as much as possible within one second and reach the set threshold value alpha, which affects the system success rate. If the signal strength received by the AF is greater than or equal to alpha, it is considered that the AF has found the target position and ended this search. The information, such as the current estimated point position and fitness value, is recorded in the record matrix for subsequent success rate calculation. On the other hand, the AF can perform the algorithm behavior within one second. If the fitness value of any AF does not reach alpha within one second, the AF with the highest fitness value among all AFs in the algorithm is selected as the estimated point. The procedure of simulation on target tracking has been described in Section 3.2.2. in the revised edition.
“3.2.2. AF Movement Restriction in the Algorithm.
Due to the moving characteristics of the target in the target tracking system, the target moving speed is 4~7 m/s. In other words, the AF needs to move as much as possible within one second and reach the set threshold value alpha, which affects the system success rate. If the signal strength received by the AF is greater than or equal to alpha, it is considered that the AF has found the target position and ended this search. The information, such as the current estimated point position and fitness value, is recorded in the record matrix for subsequent success rate calculation. On the other hand, the AF can perform the algorithm behavior within one second. If the fitness value of any AF does not reach alpha within one second, the AF with the highest fitness value among all AFs in the algorithm is selected as the estimated point. Figure 8 is the flow chart of the movement restriction of AFs in the algorithm. Equation (13) is the definition of success rate. “
Point 4: Discussion section required
Response 4: Thank you very much for this comment. The discussion section has been added in the revised edition in Section 5.
“5. Discussions
5.1. Target Positioning
According to the simulation results in Tables 5 and 6, some discussions for the target positioning system are as follows.
The average positioning error decreases as the number of AFs increases. The results in Table 10 are obtained based on the number of AFs. The average error is minimum at 100 AFs and becomes higher with the decrease of AF. However, the average positioning time will increase with the increase of AFs because the more AFs are, the longer the calculation time of the algorithm will be.
…….
5.2. Target Tracking
From the simulation results in Tables 8 and 9, the further discussions for target tracking can be depicted as follows.
RSM can use less time to achieve a better tracking success rate when there are many AFs, and the tracking performance of the modes without RSM is better when the number of AF is fewer. Table 15 shows the total average positioning time and success rate of using RSM and not using RSM, which shows that the average positioning time of the analysis mode with RSM is less than that of the analysis mode without RSM. However, as the number of AF used decreases, the difference between the average positioning time using RSM and those without RSM is smaller. The reason is that after using RSM, the number of AF used at the same time is less than that without RSM, so AF in the AFSA can obtain less information from other AFs at the same time, so AF needs to perform more algorithmic actions to reach the global optimal solution. Therefore, when the number of AF is 12, the average positioning time using the RSM is longer than that without using the RSM.
……. “
Round 2
Reviewer 1 Report
Most of my initial concerns are addressed by the authors. I recommend to accept the article.
Reviewer 2 Report
The authors have taken my comments seriously. All inaccuracies that I pointed out have been corrected in the updated version of the manuscript.
