Thông tin tài liệu
HANOI UNIVERSITY OF SCIENCE AND TECHNOLOGY ———————— MASTER’S GRADUATION THESIS Evolutionary algorithms to minimize the number of energy depleted sensors in wireless rechargeable sensor networks NGO MINH HAI hai.nm202661m@sis.hust.edu.vn Thesis advisor: Dr Nguyen Phi Le Department: Institute: Department of Software engineering School of Information and Communication Technology Hanoi, 2021 Signature of advisor HANOI UNIVERSITY OF SCIENCE AND TECHNOLOGY ———————— MASTER’S GRADUATION THESIS Evolutionary algorithms to minimize the number of energy depleted sensors in wireless rechargeable sensor networks NGO MINH HAI hai.nm202661m@sis.hust.edu.vn Thesis advisor: Dr Nguyen Phi Le Department: Institute: Department of Software engineering School of Information and Communication Technology Hanoi, 2021 Signature of advisor CỘNG HÒA Xà HỘI CHỦ NGHĨA VIỆT NAM Đôc lập – Tự – Hạnh phúc ———————— BẢN XÁC NHẬN CHỈNH SỬA LUẬN VĂN THẠC SĨ Họ tên tác giả luận văn: Ngô Minh Hải Đề tài luận văn (Tiếng Việt): Áp dụng giải thuật tiến hóa giải tốn tối thiểu số lượng cảm biến cạn kiệt lượng mạng cảm biến sạc không dây Đề tài luận văn (Tiếng Anh): Evolutionary algorithms to minimize the number of energy depleted sensors in wireless rechargeable sensor networks Chuyên ngành: Khoa học liệu trí tuệ nhân tạo Mã số HV: 20202661M Tác giả, Người hướng dẫn khoa học Hội đồng chấm luận văn xác nhận tác giả sửa chữa, bổ sung luận văn theo biên họp Hội đồng ngày 24/12/2021 vơi nội dung sau: • Sửa tiêu đề phần 2.2 từ ”Problem formulation” thành ”Problem description” (trang 18) • Giải thích rõ cách tính tham số ti (trang 19) • Vẽ lại hình để kí hiệu hiển thị rõ ràng (trang 39 - 42) • Sửa số lỗi tả (trang 20) • Hiệu chỉnh lại cách mơ hình hóa tốn (trang 18-19) • Thêm hình vẽ nhằm mơ tả rõ ký hiệu luận văn (trang 23) • Thêm số hướng nghiên cứu tương lai vào phần Kết luận (trang 43) • Sửa Chương 5: Kết luận thành phần không đánh số chương (trang 43) Hanoi, ngày Giáo viên hướng dẫn tháng Tác giả luận văn CHỦ TỊCH HỘI ĐỒNG năm 2021 REQUIREMENTS OF THE THESIS Student’s information: Name: Ngo Minh Hai Class: Data Science (Elitech) Affiliation: Hanoi University of Science and Technology Phone: 0353 852 045 Email: hai.nm202661m@sis.hust.edu.vn Duration: 01/10/2020 - 31/09/2021 Thesis statement: Declarations/Disclosures: I herewith formally declare that I — Ngo Minh Hai — have performed the work and presentation in this thesis independently under supervisions of Dr Nguyen Phi Le, Assoc Prof Huynh Thi Thanh Binh and Mrs Tran Thi Huong All of the results are genuine and are not copied from any other sources Every reference materials are clearly listed in the bibliography I will accept full responsibility for even one copy that violates school regulations Hanoi, date month Author year 2021 Ngo Minh Hai Attestation of thesis advisor: Hanoi, date month year 2021 Thesis Advisor Dr Nguyen Phi Le ABSTRACT Wireless Sensor Networks (WSNs) are one of the most core technologies of the Internet of Things (IoTs) They have a wide range of applications and have attracted lots of attentions from researchers However, a traditional WSN remains as an energy-constrained network because of the limited energy of each sensor node As a result, prolonging network lifetime has become an urgent challenge that directly affects the network performance In recent years, the appearance of a new sensor network generation, called Wireless Rechargeable Sensor Networks (WRSNs), has opened up a breakthrough in dealing with the energy issue In WRSNs, we employ a Mobile Charger (MC) equipped with a charging device to charge the sensors that have rechargeable lithium battery inside wirelessly Therefore, an effective charging scheme can enhance the whole network’s performance and minimize the energy depletion of sensor nodes Although the performance of the charging scheme is decided by some essential factors including charging path and charging time of the MC, most of the existing charging schemes only consider the MC’s charging path factor with a fully charging method Moreover, the previous works assume that the MC’s battery capacity is infinite or sufficient to charge all sensors in the network in one charging cycle This hypothesis may lead to the energy depletion of the energy-hurry sensors and unnecessary visiting for energy-sufficient sensors The charging time has not been considered thoroughly in the previous works This dissertation aims to minimize the energy depletion in wireless rechargeable sensor networks by optimizing both the MC’s charging path and charging time without the mentioned limitations Since the charging schedule optimization problem is NP-hard, the dissertation will propose an approximate algorithm to solve the investigated problem.Specifically, it proposes a novel network model in which the MC does not need to visit and charge every sensor node Furthermore, it also proposes a hybrid genetic-algorithmbased charging scheme to achieve the problem’s aim The thesis conducts various simulations and experiments to evaluate the proposed charging scheme performance Empirical evaluations have shown that the proposed charging scheme outperforms the existing solutions by a substantial margin CONTENTS List of Figures List of Tables Acronyms Introduction Preliminaries 1.1 Overview of wireless rechargeable sensor networks 1.2 The charging scheme optimization problem 1.3 Related works 1.3.1 On-demand charging scheme 1.3.2 Periodic charging scheme 1.4 Optimization Algorithms 1.4.1 Fuzzy Logic 1.4.2 Genetic algorithms 13 The problem of minimizing the number of energy depleted sensors in wireless rechargeable sensor networks 18 2.1 Problem statement 18 2.2 Problem definition 18 2.3 Mixed Integer Linear Programming Formulation 19 Hybrid fuzzy logic and genetic-algorithm-based charging scheme 3.1 24 Fuzzy logic-based preprocessing system 24 3.2 Genetic algorithm for optimizing the charging path 30 3.2.1 Individual encoding 30 3.2.2 Evaluation method 32 3.2.3 Genetic operators 32 3.3 Genetic algorithm for optimizing the charging time 34 3.3.1 Individual encoding and evaluation method 34 3.3.2 Genetic operators 35 3.3.3 Individual Adjustment 36 Performance evaluations 37 4.1 Experimental environment settings 37 4.2 Experimental evaluations 38 4.2.1 Comparison between the proposed algorithm and an exact solver 38 4.2.2 Impact of the fuzzy logic preprocessing on charging decisions 39 4.2.3 Impact of the sensor density of the network 39 4.2.4 Impact of the data packet transmitting rate of sensors 40 4.2.5 Impact of the charging rate of the MC 41 Conclusion and future works 43 Publication 44 Bibliography 45 LIST OF FIGURES 1.1 A wireless sensor network 1.2 A sensor structure 1.3 Applications of WSNs 1.4 An illustration of WRSNs 1.5 A fuzzy logic system 10 1.6 Fuzzy logic temperature 11 1.7 Common fuzzy membership function plots 12 1.8 The genetic algorithm methodology 15 1.9 Natural selection 16 1.10 Genetic operator applications 17 2.1 An example of a mobile charger being on duty 23 3.1 Fuzzy membership function plots 27 3.2 The defuzzifcation process 30 3.3 An example of the decoding procedure 31 3.4 Different values of Ác impacts the distance between offspring and their parents in Simulated Binary Crossover (SBX) 34 3.5 An example of the Single-Point and AMXO Hybrid crossover (SPAH) crossover 35 4.1 Impact of the fuzzy logic preprocessing to the number of dead sensors 39 4.2 Impact of the density of sensor on different criteria 40 4.3 Impact of the data transmitting rate of sensors to the number of dead sensors 41 4.4 Impact of the charging rate of the MC to the number of dead sensors 42 LIST OF TABLES 3.1 The parameters of input memberships 26 3.2 The parameters of output memberships 26 3.3 The parameters of input memberships 27 4.1 Parameters of energy model 38 4.2 Performance comparison between the exact solver MILP and HFLGA 38 ACRONYMS ADC Analog-to-Digital Converter BS Base Station 1, 4, 7, 8, 12, 18, 22, 24–27 COG Center of gravity 1, 12, 28 FGA Full-charging Genetic Algorithm based charging scheme 1, 39 FIS Fuzzy logic Inference System 1, 25 FLCDS Fuzzy Logic-based Charging Decision Support 1, 24, 25, 28 GA Genetic Algorithm 1, 8, 9, 37 GACS Genetic Algorithm based Charging Scheme 1, 37, 38, 40–42 HFLGA Hybird Fuzzy Logic and Genetic Algorithm based charging scheme 1, 37–42 HPSOGA Hybird PSO and GA algorithm 1, 37, 38, 40–42 INMA Invalid Node Minimized Algorithm 1, 8, 37, 38, 40–42 IoT Internet of Thing 1, MC Mobile Charger 1, 2, 6–9, 13, 17–20, 22, 24, 25, 30–43 MILP Mixed Integer Linear Programming 1, 19–22, 37, 38 MNED The problem of Minimizing the Number of Energy Depleted sensors 1, 2, 19– 22 PSO Particle Swarm Optimization 1, 8, 37 PW Priority Weight 1, 24–32 RGA Random-charging Genetic Algorithm based charging scheme 1, 39 – Generate a random number u Œ0; 1/, – Calculate ˇ ˆ then 2: if sum 3: for i D to m 4: O i O i O sum ; 5: end for 6: end if 7: for i D to m 8: if O > max then i i 9: O i 10: end if 11: end for 12: return O ; max i ; 36 ; O ;:::; O m g CHAPTER PERFORMANCE EVALUATIONS 4.1 Experimental environment settings This section will conduct some experiments with simulations to evaluate the performance of proposed algorithm, namely Hybird Fuzzy Logic and Genetic Algorithm based charging scheme (HFLGA) In addition, the performance of the proposed algorithm is also compared with an exact solver MILP and these most relevant algorithms: INMA [9], Hybird PSO and GA algorithm (HPSOGA) [12], and Genetic Algorithm based Charging Scheme (GACS) [17] Particularly, each approach can be described as follows: • The exact solver MILP is an solver that can determine the exact optimum solution of an optimization problem However, the exact solver consumes a large amount of time to calculate, even with the small scale network The solver is created based on the MILP model of the investigated problem • The INMA is an on-demand charging scheme that aims to minimize the number of invalid nodes in a WRSN An invalid nodes is a sensor node that does not have enough energy to operate functionally The INMA estimates the charging latency of the MC to decide which sensors in the charging queue will be the next charged sensor Though this method can tackle the energy problem in real time, the MC needs to charge each visited sensor to the maximum capacity of the sensor’s battery This approach may not suitable for large scale network, because sensors will have to wait for a long period of time before getting charged • The HPSOGA is an periodic charging scheme that aims to prolong the sensor network lifetime The HPSOGA combines the PSO algorithm and the GA algorithm to determine the fittest charging schedule for a variety of different network scenarios Although the experimental results is very promising, this approach has some limitations Firstly, the HPSOGA follows the fully-charging scheme Because of that, the second limitation is that the HPSOGA is only suitable for small scale networks The HPSOGA’s objective is keeping the network functional through time, which means no sensor can be exhausted If the network size is increased, the waiting time of each sensor for the MC will dramatically rise , and thus sensors will be out of energy • The GACS is another periodic charging scheme that aims to minimize the number of exhausted sensor nodes This approach use the partial charging scheme to the sensor network Therefore, the waiting time of a sensor will be decreased, and sensors can be charged more energy However, charging every sensor in the network is not always an efficient way If the charging scheme can consider which sensors need to be charged and which sensors not, The energy-hurry sensors can be charged even more energy To demonstrate the effectiveness of the HFLGA, each mentioned approach is evaluated concerning various performance metrics, such as the number of exhausted energy sensors and traveling cost The WRSNs is randomly deployed in a 1000 1000 m2 area The algorithm is implemented using the C++ and Matlab programming language and experiments 37 on a computer with the Intel® Core™ i3-6006U 2:0 GHz processor and GB of RAM The Table 4.1 is the utilized parameters of the energy model, which is similar as the previous works [10][12] Table 4.1: Parameters of energy model Parameters Value Parameters Value max 10800( J) e PM 1( J/m) e 540( J) U 5( J/m) EMC 108000( J) V 5(m/s) The following experiments evaluate three important parameters that directly affect the algorithm’s performance, which include the network density, data packet transmitting rate, and charging rate of the MC Each experimentation performed with 30 distinct random scenarios and calculated the average values of every scenarios 4.2 Experimental evaluations Firstly, this section will show the performance comparison between the HFLGA and an exact solver MILP on small network topologies Secondly, the section will demonstrate the experimental evaluation of the HFLGA with three other algorithms, including INMA, HPSOGA, and GACS 4.2.1 Comparison between the proposed algorithm and an exact solver An exact solver MILP can give an exact optimal output, but it also consumes a huge amount of time Furthermore, in some cases, a problem does not have any feasible optimal solution, leaving the solver to search in vain forever Therefore, the exact solver is only suitable for small-scale problems which always have at least one feasible optimal solution However, the exact solver still plays an important role in evaluating the efficiency of other algorithms Comparing our proposed algorithm to an exact solver will show the performance gap between the two methods The experiment compared the performance of the HFLGA to the exact solver on 20-sensor and 25-sensor networks The results are shown in the Table 4.2 Table 4.2: Performance comparison between the exact solver MILP and HFLGA Number of dead nodes Traveling cost (m) Running time (s) Dataset MILP HFLGA MILP HFLGA MILP HFLGA 20 sensors 0 1432 1614 300 15.6 25 sensors 0 1466 1537 2500 17.1 Table 4.2 shown that the number of dead nodes in both the HFLGA and the exact solver is equivalent Although there is a small gap between the traveling cost of the two methods, the results of the proposed algorithm are acceptable In contrast, the HFLGA is incredibly faster than the exact solver Specifically, the average running time of the exact solver is almost 86 times more than the average running time of the proposed algorithm Moreover, these are experiments on the small-scale network The running time of MILP will exponentially increase if the number of sensors soars It will take hours, days, even years to find an optimal solution Therefore, the HFLGA, which can produce a sufficient 38 solution in an acceptable time, is a more practical approach 4.2.2 Impact of the fuzzy logic preprocessing on charging decisions A key feature of the HFLGA is the fuzzy logic decision support system, which plays an important role in determining the optimal charging scheme In this section, we conduct experiments to compare the performance of the fuzzy logic decision support with some other decision-making approaches Specifically, the HFLGA is compared with Randomcharging Genetic Algorithm based charging scheme (RGA) and Full-charging Genetic Algorithm based charging scheme (FGA) The RGA and FGA are almost identical to the HFLGA, except in the preprocessing phase The RGA randomly selects sensors for visiting and charging The probability of sensors does not depend on any characteristic of a sensor, such as the remaining energy or the energy consumption rate On the other hand, the FGA decides that the MC have to visit and charge all sensors in the network The experiment varied the number of sensors from 150 to 300 sensors Then, each approach tried to minimize the number of dead sensors in each network scenario Fig 4.1 shows that the number of dead sensors increases when the number of sensors grows This increment will be considered later in the next section In this experiment, we only examine the gap between the HFLGA, RGA, and FGA Particularly, the number of dead sensor nodes of the HFLGA slightly increases, while the results of the RGA and FGA fluctuate wildly This is because the fuzzy logic decision support used the data from requested sensors including the residual energy and the power consumption rate to determine the charging scheme In contrast, the RGA is a random method, which is very unstable in decision making Besides, the FGA is a full-charging method, which is a common approach for an optimizing charging schedule problem Although this approach can be suitable for small networks, it can not remain its performance when the number of sensors increases 30 Number of dead sensors 25 20 15 10 150 175 200 250 Number of sensor nodes 300 Figure 4.1: Impact of the fuzzy logic preprocessing to the number of dead sensors Thanks to the outstanding fuzzy logic decision support system, the HFLGA is dominantly more efficient than the RGA and FGA Concretely, in Fig 4.2, with the number of sensors n D 300, the HFLGA lessened the dead sensor nodes by and commpared to RGA and FGA, respectively 4.2.3 Impact of the sensor density of the network This section generally appraises the effect of the number of sensors in the WRSNs by altering the number of sensors from 150 to 300 Specifically, each considered algorithm ran 39 independently to find the charging schedule of 150-, 175-, 200-, 250-, and 300-sensors network Fig 4.2a and Fig 4.2b are visualizations of the criteria which aim to compare the solutions of mentioned algorithms 40 20000 35 Traveling energy (J) Number of dead sensors 30 25 20 15 15000 10000 10 5000 0 150 175 200 250 Number of sensor nodes 300 150 (a) On the number of dead nodes 175 200 250 Number of sensor nodes 300 (b) On the total traveling cost Figure 4.2: Impact of the density of sensor on different criteria It is easy to notice that when the number of sensors increases, the sum of energy depleted sensors and the traveling cost rise proportionally The reason for those increments is that when the scale of the network grows, the generated data from sensors will be enlarged Therefore, more duties will be imposed on the sensors, which causes the gain of the energy consumption rate of each sensor As a matter of course, there will be more sensors at risk of depleted energy, and the MC must travel more but can not support all of them on time However, the increment of in the traveling cost of HFLGA and INMA is quite negligible Furthermore, the reductions in the traveling cost of HFLGA and INMA are also much better than other algorithms Particularly, in the INMA, the charging decision is based on the information from the sensors that sent the charging requests So that, if the charging request pool is empty, the MC will not need to waste its time for traveling For HFLGA, because the MC only visits a part of the sensor set, the traveling cost decreases compared to the GACS and the HPSOGA As a result, the MC can arrive at sensors in danger of dying earlier and can spend more energy for charging Due to the above features of HFLGA, its performance is much more efficient than other algorithms, shown in Fig 4.2 Specifically, with the network size n D 250, HFLGA reduces the failure sensor nodes by 10:5; 17:6; and 23:6 (sensors) compared to GACS, INMA, and HPSOGA, respectively 4.2.4 Impact of the data packet transmitting rate of sensors In general, a sensor can be found in four different states: idle state, sensing state, communicating state, and exhausted state Particularly, an idle sensor will wait for others’ data to analyze, summarize, and then broadcast them to sensors or the base station in the communicating state On the other hand, a sensor will collect the environment information in the sensing state Lastly, if a sensor reaches the minimum energy capacity, it will be exhausted and will become a failure sensor node It is easy to see that a sensor mostly con40 sumes energy in its sensing state and communicating state Moreover, these states strongly depend on the data transmitting rate of sensors The more data transmitting rate is, the more energy will be drained off In this experiment, the number of sensors is fixed at 200, while we vary the data packet transmitting rate from to 12 (kbps) As shown in Fig 4.3, the data transmitting rate growth makes a higher number of dead sensor nodes in all algorithms As mentioned before, this is reasonable because the higher the data transmitting rate, the higher the energy consumption, and shorter lifetimes that all sensors have Number of dead sensors 50 40 30 20 10 Data transmitting rate (kbps) 10 12 Figure 4.3: Impact of the data transmitting rate of sensors to the number of dead sensors It is easy to observe that our proposed algorithm HFLGA outperforms all of the other algorithms at every level of data transmitting rate Concretely, in case the data transmitting rate is 12 kbps, HFLGA reduces the failure sensors by 8:8; 14:1; and 21:1 (sensors) compared to GACS, INMA, and HPSOGA, respectively One more striking point is the performance gaps between HFLGA and the other algorithms gradually increase when raising the data transmitting rate This is because the lifetime of sensors is decreased dramatically due to the increase of the data transmitting rate and thus, the MC must travel to guarantee sensors alive The outstanding features of the HFLGA impacting to the total traveling cost can satisfy the need to visit sooner of risky sensors 4.2.5 Impact of the charging rate of the MC Finally, this part will determine the impact of the charging rate of the MC to the number of dead sensors A MC with a sufficient charging rate can save a lot of time for transfer energy Thus, the waiting time of sensors should be reduced and more sensors can be saved This experiment fixed the number of sensors and the data transmitting rate of sensors at 200 (sensors) and (kbps), respectively Then, the charging rate of the MC will be varied from W to 12 W Fig 4.4 presents the results of all considered algorithms As observed in this figure, the number of depleted sensor nodes decreases with the rise in the charging rate of the MC As previously mentioned, because when the MC’s charging rate goes up, the charging time is less than before Therefore, the MC can travel earlier to other sensors before they are exhausted Another notable point is that the number of depleted sensors of the HFLGA is always lower than all the algorithms under consideration This is 41 probably due to the fact that the HFLGA has a better charging path than other algorithms 55 50 Number of dead sensors 45 40 35 30 25 20 15 10 Charging rate (Watts) 10 Figure 4.4: Impact of the charging rate of the MC to the number of dead sensors Specifically, the number of depleted sensors by HFLGA, GACS, INMA, and HPSOGA subside from 33:2, 40:0, 45:0, and 52:0 down to 23:9, 28:1, 36:0, and 49:0 (sensors), respectively Through all of the experiments, we can conclude that the HFLGA has a better performance comparing to GACS, INMA, and HPSOGA on both reducing the number of depleted sensors and saving in traveling cost 42 CONCLUSION AND FUTURE WORKS Conclusion This thesis proposed a partial charging scheme to minimize the exhausted energy sensors in WRSNs and simultaneously guarantee the traveling cost of the MC be minimized under the limited energy constraints Moreover, this thesis presented an algorithm based on the hybrid fuzzy logic inference scheme and genetic algorithm to solve the investigated problem The experimental results on various simulation scenarios indicate that the proposed algorithm significantly reduces the number of depleted nodes and minimizes the traveling cost compared to existing works However, some shortcomings still exist in the thesis The performance of the proposed algorithm is fairly good in large-scale networks But if the sensor network is denser, the number of dead sensor nodes will increase dramatically Furthermore, the charging path and charging time optimization are quite discrete and independent These two phases are weakly related and can not support each other Future works are considered to solve these limitations of the thesis Future works Due to the conclusion of the thesis, some future works should be conducted The charging path and charging time optimization should be combined to an unified algorithm, such as the bi-level approach With an unified method, the charging path and charging time will be strongly connected to each other I should also consider a more general scenario to evaluate the performance of proposed algorithm in the future, such as considering the network until one of the sensors is dead The model can be modified to multi-point charging model which is more appropriate to the dense sensor networks Moreover, the model with multi mobile chargers would be interesting to consider Furthermore, the quality of a charging scheme firmly depends on the charging path of the MC A future approach should explore more charging path of the MC so that we can find more quality charging schemes 43 PUBLICATION Parts of this thesis have been submitted and accepted as papers of 2021 IEEE Congress on Evolutionary Computation (CEC) and Applied Intelligence Journal (2021) Tran Thi Huong, Nguyen Ngoc Bao, Ngo Minh Hai, Huynh Thi Thanh Binh, et al.Effective partial charging scheme for minimizing the energy depletion and charging cost in wireless rechargeable sensor networks In 2021 IEEE Congress on Evolutionary Computation (CEC), pages 217–224 IEEE, 2021 Tran Thi Huong, Le Van Cuong, Ngo Minh Hai, Nguyen Phi Le, Le Trong Vinh,and Huynh Thi Thanh Binh A bi-level optimized charging algorithm for energy depletion avoidance in wireless rechargeable sensor networks Applied Intelligence,pages 1–23, 2021 44 BIBLIOGRAPHY [1] Tran Thi Huong, Nguyen Ngoc Bao, Ngo Minh Hai, Huynh Thi Thanh Binh, et al Effective partial charging scheme for minimizing the energy depletion and charging cost in wireless rechargeable sensor networks In 2021 IEEE Congress on Evolutionary Computation (CEC), pages 217–224 IEEE, 2021 [2] Ian F Akyildiz, Weilian Su, Yogesh Sankarasubramaniam, and Erdal Cayirci Wireless sensor networks: a survey Computer networks, 38(4):393–422, 2002 [3] Moshe Masonta, Yoram Haddad, Luca De Nardis, Adrian Kliks, and Oliver Holland Energy efficiency in future wireless networks: Cognitive radio standardization requirements In 2012 IEEE 17th International Workshop on Computer Aided Modeling and Design of Communication Links and Networks (CAMAD), pages 31–35 IEEE, 2012 [4] Nitin Goyal, Mayank Dave, and Anil K Verma Data aggregation in underwater wireless sensor network: Recent approaches and issues Journal of King Saud UniversityComputer and Information Sciences, 31(3):275–286, 2019 [5] He Ba, Ilker Demirkol, and Wendi Heinzelman Passive wake-up radios: From devices to applications Ad hoc networks, 11(8):2605–2621, 2013 [6] Nguyen Thi Tam, Huynh Thi Thanh Binh, Dinh Anh Dung, Phan Ngoc Lan, Bo Yuan, Xin Yao, et al A hybrid clustering and evolutionary approach for wireless underground sensor network lifetime maximization Information Sciences, 504: 372–393, 2019 [7] Amar Kaswan, Abhinav Tomar, and Prasanta K Jana An efficient scheduling scheme for mobile charger in on-demand wireless rechargeable sensor networks Journal of Network and Computer Applications, 114:123–134, 2018 [8] Yong Feng, Nianbo Liu, Feng Wang, Qian Qian, and Xiuqi Li Starvation avoidance mobile energy replenishment for wireless rechargeable sensor networks In 2016 IEEE International Conference on Communications (ICC), pages 1–6 IEEE, 2016 [9] Jinqi Zhu, Yong Feng, Ming Liu, Guihai Chen, and Yongxin Huang Adaptive online mobile charging for node failure avoidance in wireless rechargeable sensor networks Computer Communications, 126:28–37, 2018 [10] Yi Shi, Liguang Xie, Y Thomas Hou, and Hanif D Sherali On renewable sensor networks with wireless energy transfer In 2011 Proceedings IEEE INFOCOM, pages 1350–1358 IEEE, 2011 [11] Liguang Xie, Yi Shi, Y Thomas Hou, and Hanif D Sherali Making sensor networks immortal: An energy-renewal approach with wireless power transfer IEEE/ACM Transactions on networking, 20(6):1748–1761, 2012 45 [12] Zengwei Lyu, Zhenchun Wei, Jie Pan, Hua Chen, Chengkai Xia, Jianghong Han, and Lei Shi Periodic charging planning for a mobile wce in wireless rechargeable sensor networks based on hybrid pso and ga algorithm Applied Soft Computing, 75: 388–403, 2019 [13] Lingkun Fu, Liang He, Peng Cheng, Yu Gu, Jianping Pan, and Jiming Chen Esync: Energy synchronized mobile charging in rechargeable wireless sensor networks IEEE Transactions on vehicular technology, 65(9):7415–7431, 2015 [14] Tran Thi Huong, Huynh Thi Thanh Binh, Phi Le Nguyen, Doan Cao Thanh Long, Vuong Dinh An, et al Optimizing charging locations and charging time for energy depletion avoidance in wireless rechargeable sensor networks In 2020 IEEE Congress on Evolutionary Computation (CEC), pages 1–8 IEEE, 2020 [15] Xiao Lu, Ping Wang, Dusit Niyato, Dong In Kim, and Zhu Han Wireless charging technologies: Fundamentals, standards, and network applications IEEE communications surveys & tutorials, 18(2):1413–1452, 2015 [16] Lingkun Fu, Peng Cheng, Yu Gu, Jiming Chen, and Tian He Optimal charging in wireless rechargeable sensor networks IEEE Transactions on Vehicular Technology, 65(1):278–291, 2015 [17] Tran Thi Huong, Phi Le Nguyen, Huynh Thi Thanh Binh, Kien Nguyen, Ngo Minh Hai, et al Genetic algorithm-based periodic charging scheme for energy depletion avoidance in wrsns In 2020 IEEE Wireless Communications and Networking Conference (WCNC), pages 1–6 IEEE, 2020 [18] Zadeh Lotfi et al Fuzzy sets Information and control, 8(3):338–353, 1965 [19] Abhinav Tomar, Lalatendu Muduli, and Prasanta K Jana A fuzzy logic-based ondemand charging algorithm for wireless rechargeable sensor networks with multiple chargers IEEE Transactions on Mobile Computing, 2020 [20] Dr Nguyen Dinh Thuc Artificial Intelligence - Evolutionary Programming Education Publishing House, 2009 [21] David B Fogel Evolutionary programming: An introduction and some current directions Statistics and Computing, 4(2):113–129, 1994 [22] Thomas Bäck, Frank Hoffmeister, and Hans-Paul Schwefel A survey of evolution strategies In Proceedings of the fourth international conference on genetic algorithms Citeseer, 1991 [23] Huynh Thi Thanh Binh, Pham Dinh Thanh, and Ta Bao Thang New approach to solving the clustered shortest-path tree problem based on reducing the search space of evolutionary algorithm Knowledge-Based Systems, 180:12–25, 2019 [24] Tran Thi Huong, Le Van Cuong, Ngo Minh Hai, Nguyen Phi Le, Le Trong Vinh, and Huynh Thi Thanh Binh A bi-level optimized charging algorithm for energy depletion avoidance in wireless rechargeable sensor networks Applied Intelligence, pages 1–23, 2021 46 [25] Lawrence V Snyder and Mark S Daskin A random-key genetic algorithm for the generalized traveling salesman problem European journal of operational research, 174 (1):38–53, 2006 [26] Kalyanmoy Deb, Ram Bhushan Agrawal, et al Simulated binary crossover for continuous search space Complex systems, 9(2):115–148, 1995 [27] Kalyanmoy Deb and Debayan Deb Analysing mutation schemes for real-parameter genetic algorithms International Journal of Artificial Intelligence and Soft Computing, 4(1):1–28, 2014 [28] Kavitesh Kumar Bali, Yew-Soon Ong, Abhishek Gupta, and Puay Siew Tan Multifactorial evolutionary algorithm with online transfer parameter estimation: Mfea-ii IEEE Transactions on Evolutionary Computation, 24(1):69–83, 2019 [29] Aki Sorsa, Riikka Peltokangas, and Kauko Leiviska Real-coded genetic algorithms and nonlinear parameter identification In 2008 4th International IEEE Conference Intelligent Systems, volume 2, pages 10–42 IEEE, 2008 [30] Rogaia Mhemed, Nauman Aslam, William Phillips, and Frank Comeau An energy efficient fuzzy logic cluster formation protocol in wireless sensor networks Procedia Computer Science, 10:255–262, 2012 47 ... Minh Hải Đề tài luận văn (Tiếng Việt): Áp dụng giải thuật tiến hóa giải toán tối thiểu số lượng cảm biến cạn kiệt lượng mạng cảm biến sạc không dây Đề tài luận văn (Tiếng Anh): Evolutionary algorithms... description” (trang 18) • Giải thích rõ cách tính tham số ti (trang 19) • Vẽ lại hình để kí hiệu hiển thị rõ ràng (trang 39 - 42) • Sửa số lỗi tả (trang 20) • Hiệu chỉnh lại cách mơ hình hóa tốn (trang... mơ tả rõ ký hiệu luận văn (trang 23) • Thêm số hướng nghiên cứu tương lai vào phần Kết luận (trang 43) • Sửa Chương 5: Kết luận thành phần không đánh số chương (trang 43) Hanoi, ngày Giáo viên
Ngày đăng: 02/06/2022, 17:18
Xem thêm:
