www.ijird.com July, 2021 Vol10 Issue 7 ISSN 2278 – 0211 (Online) Enhancing Cell Coverage and Capacity in LTE Heterogenous Network Using Intelligent Base Station Ifesinachi E.O. Engineer, Department of Electronic, Electronic Development Institute (ELDI), Awka, Nigeria Idigo V. E. Professor, Department of Electronic and Computer Engineering, Nnamdi Azikiwe University Awka, Nigeria Ohaneme C.O. Professor, Department of Electronic and Computer Engineering, Nnamdi Azikiwe University Awka, Nigeria Obioma P.C. Engineer, Department of Electronic and Computer Engineering, Nnamdi Azikiwe University Awka, Nigeria Abstract: The high data traffic demand by mobile users has tremendously increased in recent times, thus creating the need to further improve the capacity and coverage of already existing systems. Recent advancements in self-organizing and self driving Base stations are being exploited to bridge this gap. Although the paradigm have already evolved in 2G, 3G and 4G, the automation is realized by predefined policies, rather than interact with the environment to make smart decisions. In this paper, the eNB’s of LTE networks have been designed in such a way that makes them self-aware, selfadaptable and intelligent using the Ant Colony Optimization algorithm. The ACOA was used to minimize the path trailed by the moving eNB’s employing the SON capability of LTE systems. Three different scenarios were considered; results showed that in the first scenario, which is considered to be an ideal case, the optimized system improved the system throughput by 19.7% after 150s. The packet loss for this scenario was also reduced 38% at 150s. The call blocking probability outperformed the conventional system by 45.9% as at 120s and by 22.2% when the time was 150s. The second scenario considered analyzed the system performance when one cell is overloaded while a nearby cell is idle. Simulation results showed that the ACOA resulted to about 820% utilization of the idle cell, while each UE achieved an improved throughput compared to the conventional system. Results also showed that the throughput improvement provided by the ACOA in one of the eNB’s was about 27Mbps when compared to the conventional system, and 55Mbps in another eNB when all the cells were loaded to congestion. Keywords: Het Net, LTE, mobile base station, ant colony, dynamic base station 1. Introduction The fourth generation mobile communication system (4G) network also known as the Long Term Evolution (LTE) cellular network system was developed as an advancement to the former third generation (3G) network that falls within the premise of the third generation project partnership long time evolution 3GPP design by the 3GPP LTE.Due to the high demand by mobile subscribers, high data traffic has tremendously increased in recent times, thereby creating need to search for approaches to further improve the already-known channel capacity and coverage of the already existing systems. For example, in the LTE network systems, the capacity of a network have been improved by (a) improving the spectral efficiency with advanced modulation and coding techniques, (b) improving interference reduction conditions within a cell by using advanced antenna technology (example beam forming), (c) increasing the bandwidth of the radio communication channel, (d) increasing the number of links between a transmitter and a receiver through the use of spatial multiplexing technology (example., MIMO), and (e) adopting aggressive frequency reuse, that is, reusing the same radio resources several times over a given area by deploying more base stations. The last technique of deploying more base stations has been shown to provide significant network capacity gains since the evolution of cellular networks. Against these improvements, the viability of such improvements may be hampered due to the limitations on the hardware implementation and channel conditions, as well as, the increase in the system complexity due to the use of these advanced techniques. As stated before, in order to increase capacity, cells have to be densely deployed. Beyond the cost implication of these cells, the number of parameters to be configured in these base stations also increases. Also, with the operational complexity in LTE systems, it is challenging and time-consuming for mobile operators to operate and configure the network all by humans. Hence, this is making it necessary to increase INTERNATIONAL JOURNAL OF INNOVATIVE RESEARCH & DEVELOPMENT DOI No. : 10.24940/ijird/2021/v10/i7/JUL21027 Page 39 www.ijird.com July, 2021 Vol10 Issue 7 automation to wireless mobile networks. orks. Currently, Cur many of the autonomous functionalities alities an and automation mechanisms are based on a Self-Organised Network rk (SON). (SON This is a network concept designed to o autonomously autonom configure, optimise and heal LTE HetNets by limiting human uman inte intervention [1]. This was first introduced in n LTE Release Rel 8 in order to manage mobile cell coverage. This concept relies elies on ar artificial intelligence (AI) to analyse the network twork par parameters network and take actions independently [2].Although the self-operation-network self (SON) paradigm have already lready evolved evo in 2G, 3G and 4G, the automation is realized by predefined ed policie policies, rather than interact with the environment ment to make m smart decisions. It is promising to apply AI to the network rk so as to t make it smarter, consequently making the SON coordination to configure, optimize and heal the network more ore efficiently. efficie Artificial intelligence is the technique ue that makes machines to solve problems with intelligence like human an beings beings. With artificial intelligence, the machinee mimics human minds and ‘learns’ from the environment, then solve problems roblems by maximizing the success probability. Swarm In Intelligence is an AI scheme which is based on observations of collective behaviour in biological organisms like ants and b bees, including division of labour, counting larvae, building, cooperative operative transport, etc. Self-organization protocol ol via swarm swa intelligence is a useful solution which has already demonstrated strated advantages ad in terms of building an intelligent, ligent, sel self-organizing router which tackles the traditional router problems, ms, as we well as balancing the problems of coverage in wireles wireless communications [3]. The Ant Colony Optimisation (ACO) technique nique is one o example of swarm intelligence inspired red algorit algorithm that is widely used by researchers to solve problems in LTE E systems. systems This work makes use of the ACO algorithm thm to ma make the BTSs self-organized based on pheromone density. The ant nt agent will w be spread throughout the network to o find ove overcrowded and idle or less crowded BTSs and in turn move the dense BTSs BT loads towards the less dense BTSs. 2. Review of Related Works Many different research models odels have hav been proposed to enhance capacity and nd coverage covera in LTE with the aim to permit the LTE’s attributes to evaluate ate their needs n directly or through the use of equipment pment suc such as the coordinate multipoint transmission / reception (CoMP) P) technique techni to enhance the equity of traffic in the networ twork.[4] proposed an efficient utilization approach for both lower and highe higher frequency bands by separating the frequency quency between be wide (macro cells) and local (micro cells) areas. In this work, their the approach was used to solve the problem m of insufficiency insuf of spectrum in the lower frequency bands. In the study by [5], th they used cognitive base stations in LTE networks etworks to enhance the functionality of femtocell base stations. [6] worked ed on capa capacity analysis and optimization in heterogeneous geneous n network with adaptive cell range control, here the capacity of users sers in ma macrocells, small cells and range extensionss were analyzed ana in condition with or without cell range extension (CRE). ). [7] pro proposed a scheme that allows mitigation of inter-cell inter interference through fractional self-powered control performed ormed at eeach femtocell user. This study analyzes a scheme with w optimum power value that provides a compromise between en the ser served uplink signal within unwanted interference rference p plus noise ratio to enhance spectral efficiency in terms of throughput. ughput. In the study by [8] on emerging trend on small cell cel technology, capacity and coverage of indoor environment in a heterogeneous heterog network was enhanced by the deployment deploymen of small cell technology like the Metrocell (up to 2km), microcell icrocell (100m (1 to1000m), picocell (50m to 100m) m) and femtocell fe (10m-20m). This according to the study can doubled the capaci capacity of a HetNet without any addition of spectrum pectrum and a infrastructure. [9] used particle swarm optimization (POS) algorithm to conduct dynamic user association by y finding the optimal bias value via considering the network balance index dex (the actual ac load and the predicted load).Minimizing izing the energy e and interferences of mobile users are major concerns of Heterogeneous Heterogen Networks (HetNet). The work by [10] aimed imed at improving the energy of mobile users by minimizing a weighted hted funct function.In this paper, the ACOA is employed to solve the t underlying problem of enhancing cell capacity in LTE systems tems by considering c a densely deployed LTE network twork in which the BTSs‘ coverage overlap and the traffic load fluctuate te over time ti and space. The work enhances the system ystem thr throughput, minimize packet loss, and the load of the system as performanc erformance metrics in order to prove the system viability, iability, w which most reviewed works failed to ascertain. rithm (ACOA) 3. The Ant Colony Optimization Algorithm The ant colony optimization on algorithm algorit (ACO) is an evolutionary meta-heuristic ristic algorithm alg based on a graph representation that has been applied ed success successfully to solve various hard combinatorial optimiz timization problems. The main idea of ACO is to model the problem m as the search s for a minimum cost path in a graph. ph. Artific Artificial ants walk through this graph, looking for good paths. Each ant has a rather simple behaviour so that it will typically pically on only find rather poor-quality paths on its own. Better paths are found und as the emergent result of the global cooperation n among ants in the colony [11]. In 1989 and 1990, in Santa ta Fe Ins Institute, Bonabeau and his colleagues demonstrat emonstrated that similar problems involving the best and shortest path h can be solved more easily with computers creating ting ‘virtu ‘virtual ants’[11]. According to this, virtual ants leave a scent that represents the length of the route on their behind and by this th means the other virtual ants will find the shortest routes. Smell of th the traces of long routes that are not preferred eferred w will disappear gradually by simulating evaporation of the smell trace at a certain speed. This will prevent the deflection lection of the t virtual ants to the long road outside short cut roads.Figure 1 shows the th ants going to their food through a linear ar way. Figure gure 1: The Path of the Ants to Their Food [11] INTERNATIONAL JOURNAL OF INNOVATIVE RESEARCH ESEARCH & DEVELOPMENT DOI No. : 10.24940/ijird/2021/v /v10/i7/JUL21027 Page 40 www.ijird.com July, 2021 Vol10 Issue 7 Figure 2: Path Selection by Ants Coming across an Obstacle [11] Ants help the next ants to choose hoose the path by dropping off stored Pheromoness on the cchosen way. They show the path of their food using a volatile chemical substance s known as pheromone which evaporate evaporates with time. A remote ant generally moves randomly and while le moving, moving a quantity of pheromone is deposited by y the ant on its trail. Other ants will select the path with the highest pheromone eromone intensity which represents the most attractive tractive w way for ants. When an ant follows a path, it strengthens the trail ail by add adding its own pheromone. At the end, thiss results in a collective behaviour of which there is more than one path with ith high p pheromones; the higher the pheromone, the higher is the number of ants that follow it. This instinctive behaviourr explains how they had found the shortest path to their heir food, fo even in the case preexisting path cannot be used. In fact, t, if an obs obstacle is on the path to food, the ant in front ont of this obstacle cannot continue and has to make a choice for a new w way as shown in figure 2. The basic idea of Ant-based based algorithms a is that artificial intelligent agents using a simple communicat mmunication mechanism can produce the solutions ns to many man complex problems. The objective of these algorithms is to enable able the network n to learn and use that experience for future actions. 4. Development of the ACO Algorithm d the optimal optim pilot eNB to help it move to dense areas reas and serve the increased newly The ACOA is applied to find arrived UEs. This scheme tackles thee problems problem of fixed coverage scheme, reducing thee ratio of coverage c holes, proportion of coverage overlaps and probability of blocking bloc new users which enhances the overall verall per performance of the network. Optimization of the new location of the mobi mobile eNB’s is modelled by an ant colony movement ovement ggoing from one location to another looking for food. During thee food search, sea the ants will visit the frequently attended ended cell cells by the mobile UE. At the end of the algorithm, the pheromonee deposit deposited by the ants will accumulate on the cells. ells. As ea each cell will be served by a mobile eNB, the cell having the most st pheromone pherom intensity is considered as the densest st cell of the mobile and its serving eNB will require movement to idle or less de dense eNB’s. This method is also used to predict redict the moving eNB towards the dense cell of mobile users. The less pheromo pheromone quantity cell means less or unused eNB, thus it i will be the future moved eNB. The algorithm to predict the movement ovement of the eNBs will be launched when a mobile bile user enters e a location. It creates an N-entries table called Movementt Table (MT). (M The content of this table will be fed d by the ccache which has history of mobile user‘s movement behaviour in the cas case of period (i.e. work day or holiday) or when hen ther there is an event (i.e. a football match in a stadium cell). The table will ill includ include the most recent entries corresponding to the mobile, mo with the same source eNB and destination eNB. The number ber of entries ent for a mobile user is denoted by N1. If MT is n not full which means N1=0, the system will fill it using N2 entries ies corres corresponding to other mobiles with the same source cell ce and the same period or event which assume that mobile users rs in the ssame cell can have the same direction. The algorithm will ensure that hat eNB’s managing the cells can update their users’ rs’ history movements or directions towards other eNB’s and deduce their eir future locations according to their current location ation and their behaviours. If a new mobile user enters in the system and d the eNB does not have history for it, it can use the he history of other mobile users that have the same mobility profile. In thee case of p public events such as a match in a stadium, m, a new u user will more likely travel from the city, to the stadium. The system then creates a colony of ants whose size is equal to the movement nt table (MT) ( size. Each ant * is associated with the line i of the m movement table. Two fields are used to define the th structure of an ant, the visibility field and the pheromone field. ield. The v visibility field is considered as a vectorƞ of R elem elements corresponding to the number of adjacent cells. Each element ent of this th vector represents the visibility of an adjacent cell c by an ant. This field is activated to reflect the fact that antss prefer an already visited cell when they search forr food. If= If represents the visibility of the cell for ant then: = $%% & (1) Where which represents presents the t adjacent cells, is a value > 0 and ' is the parameter to increase the cell degree of visibility if it was already ady visited visite by the mobile. represents the () entry ntry of the movement table. To keep the trace of the pheromone deposited d by the ants, a we use a vector ! with R elements. Each element of the vector corresponds correspon to the quantity of pheromone in a correspondi rresponding cell. This vector will be # for initialized to ! " (where k represents the cell k and " representss the initial initia time). The search process proceedss in a set of iterations in which each ant moves towards wards an adjacent cell, deposits the quantity Q of pheromone in this celll to encourage encou the other ants to go towards it. It then en return returns towards its nest (i.e., the INTERNATIONAL JOURNAL OF INNOVATIVE RESEARCH ESEARCH & DEVELOPMENT DOI No. : 10.24940/ijird/2021/v /v10/i7/JUL21027 Page 41 www.ijird.com July, 2021 Vol10 Issue 7 current cell) in order to repeat the same proc process. Each ant chooses the future cell according ording to its degree of visibility and the intensity in pheromone of the cell. ll. At any m moment t, the ants chooses to go to the cell K according acc to the probability: + -. / ( 01 -=2 3 04 , (2) 1 4 56 /78-. / ( 0 -=2 / 0 After each iteration time 9 , thee pherom pheromone intensity in the locations is updated according to ants’ contribution and the evaporation rate. For each location k,, we have: h -! 9 0 :; ! 9 # (3) Note that a small value of ; generates slow pheromone dissipation and a high value generates gen a faster dissipation. Notice that n represents the number er of ants that choose location k. The proposed algorithm lgorithm in this work provides the ability to integrate a mobile’s behaviour, iour, the existing e infrastructure and other mobiles’’ behaviou behaviour in the prediction process (mobility prediction based on an antt system).The system). flowchart of the system is as shown in figure 3. From the flow chart, the information obtained by the ACOA from each visited site is sent as a script to the Network ork Interface Inter Collector (NIC) of the eNB, which will first sense the network ork state information i to decide whether (or not) to move the desired eNB based on the traffic load information, cell density, channel conditions c and UE requirements information tion available. availa Figure 3: The Flowchart of the System 5. Simulation Results The simulation was conducted ted using the OPNET simulator. The downlink scenario nario of a two-tier OFDMA network is considered, and comprises of macrocells crocells and an small cells where the former is overlaid id with th the latter. In each macrocell, an eNB is located at the center of hexagonal exagonal coverage c area to afford a basic network coverage. Macro users are randomly distributed in the whole area as shown. wn. A netw network of unit area of 1000m×1000m was simulated; simulated one node acts as a server and is connected to 3 eNb’s with wired ed links aas depicted in figure 4. INTERNATIONAL JOURNAL OF INNOVATIVE RESEARCH ESEARCH & DEVELOPMENT DOI No. : 10.24940/ijird/2021/v /v10/i7/JUL21027 Page 42 www.ijird.com July, 2021 Vol10 Issue 7 Figure 4: Simulation Environment The traffic between the eNB’s ’s and mobile mo nodes transmits packets with a Constant stant Bit Rate R (CBR). The packet size is 512 bytes and the simulation time is 150s. In I table 4, the network parameters are shown. Paramet ameter Value Nodes 8 Speed 10m/s Movemen Movement Random MAC 802.11 Number er of Active Activ users 9 Applicatio Application CBR Simulation mulation ttime 150s Area 1000X1000 m2 eNodeB eB covera coverage area 100 m Routing uting Protocol Prot DSDV Transmitting smitting capacity c 2 Kbps, 4 Kbps Number ber of ant agent 4 Packet ket Size 512bytes Table 1:: Shows the Parameters Used In Configuring the Network for This Experiment The results and analysis off the sim simulation are presented for the comparison on of thr threedifferent scenarios. The system performance is analysedforr the con conventional system and for an optimized d system with the use of the ACO algorithm. Figure 5 illustrates an ideal eal scenar scenario where there is a normal load distribution tion among amon the network, and the UE are considered to be mobile. Figure 5: Simulation Scenario 1 INTERNATIONAL JOURNAL OF INNOVATIVE RESEARCH ESEARCH & DEVELOPMENT DOI No. : 10.24940/ijird/2021/v /v10/i7/JUL21027 Page 43 www.ijird.com July, 2021 Vol10 Issue 7 Figure 6: Throughput ghput Network Performance for Conventional Movement ent and ACO Optimized Movement for Scenario 1 ACOA The scenario represented in n figure 5 is a common one in most real life situations, ns, the sim simulation result in terms of throughput and packet loss is as illustrated strated in figure 6 and 7. From the results, it can be found ound that tha the scenario where the ACOA algorithm was used, use the system significantly outperformed the conventional method. thod. With Within the first 30 to 50 seconds, the throughput hput from both the normal and the optimized algorithm performed fairly ly similar similarly, with the optimized algorithm outperforming orming the th normal technique. It can be seen that with time, the performance ance of the ACOA optimized system increased and became sig significant from 60s seconds with up to 70Mbps throughput. At 150 50 second seconds, the throughput of the ACO algorithm improved by 19.7% when compared to the conventional system. The packets ackets los lost by both systems were also accessed,, so as to determine how well the algorithm performs. The result is as shown in figure 7. Figure 7: Packet Loss Los Ratio in the Network for Both Systems for Scenario nario 1 From the simulation result,, the application applic of the ACOA shows a better reduction tion in the percentage of packet loss as shown in figure 4.4 for all the times es consid considered. Between 15 to 20s, both system performanc rformance was fairly similar, but as the time increased, the ACOA showed ed better results (i.e., gave shorter paths) to the UE dense locations. Comparing the conventional movement and the ACOA o optimized movement, the optimized movement ovement clearly achieved better performance by approximately 39% at 150 seconds se than the normal moving technique. The simulation simu also considered the user equipment experience, and the result is p presented in figure 8. INTERNATIONAL JOURNAL OF INNOVATIVE RESEARCH ESEARCH & DEVELOPMENT DOI No. : 10.24940/ijird/2021/v /v10/i7/JUL21027 Page 44 www.ijird.com July, 2021 Vol10 Issue 7 Figure 8: Throughput put Network Performance of UE for Both Systems in Scena Scenario 1 the optimized design, each UE achieved an improved improve throughput compared to From figure 8, it is obvious that in th the conventional system. For UE1, the he optimi optimized algorithm achieved 14Mbps better than han the normal n algorithm. Similarly for UE5 where both systems had an improve roved performance due to user mobility and d position in the cell, the optimized algorithm achieved 3Mbits/s better than the conventional system. The ACO algorithm m was als also applied to the UE2, UE3 and UE4 that are sparsely distributed ed in the cell; the system employing the ACOA achieved hieved 11Mbps, 11 8Mbps and 10Mbps respectively better than the conventional tional syst system. The call blocking probability is also considered onsidered as shown in figure 9. Figure gure 9: Call Blocking Probability for Scenario 1 th first 30s, the call blocking probability for both systems were fairly close, As can be seen in figure 9,, within the about 0.22 and 0.25 for the optimized ized syste system and conventional system respectively. ly. As the time increased, with UE movement and activity within the cell, the ca call blocking probability remained fairly low. For the system employing the ACOA, the call blocking probability outperfor outperformed the conventional system by 45.9% as at 120 120s and by 22.2% when the time was 150s.These investigations were further fur extended to throughputs achieved at each eNBs e when there’s a major event happening close to a particularr eNB site. site Figure 10: Simulation Scenarios 2 When the density of the cells lls was increased as shown in figure 10, the system m performance perform at each of the eNB’s was monitored. From figure 10, the system at eNB2 was loaded to congestion as shown n and som some UE sparsely distributed INTERNATIONAL JOURNAL OF INNOVATIVE RESEARCH ESEARCH & DEVELOPMENT DOI No. : 10.24940/ijird/2021/v /v10/i7/JUL21027 Page 45 www.ijird.com July, 2021 Vol10 Issue 7 at eNB3. This scenario is likened to when there ther is an event, say a football match and there ere are clusters of people very close to a particular eNB (in this case eNB2), B2), and a close by cell is idle. The result of the performa performance of the two systems is presented in figure 11. Figure re 11: Comparison of Throughput Data When the Enbs Are Moved For Both Systems From figure 11, the results obtained showed that the ACOA helped to utilize the idle base ba at eNB1, thus reducing the load at eNB2. The throughput improveme mprovement provided by the ACOA in eNB2 even with the density d of the cell is about 2.5Mbps when compared to the convention onventional system, and 7Mbps and 72Mbps respectively vely for eNB1 and eNB3. This results to 820% utilization of an idlee cell. Beyond Bey improving the throughput, an added advantag advantage of the optimized system is that the call blocking probability iss greatly rreduced as shown in figure 12 Figure 12: Throughput Network Performance of UE for Both Systems for Scenario 2 From figure 12, it is obviouss that in the th optimized design, each UE achieved an n improve improved throughput compared to the conventional system. For UE1, the he optimi optimized algorithm achieved 25Mbps better than the normal n algorithm. Similarly for UE5 where both systems had an n improve improved performance due to user mobility and d position in the cell, the optimized algorithm achieved 4Mbits/s better than the conventional system. The ACO algorithm m was als also applied to the UE2, UE3 and UE4 that are sparsely distributed ted in the cell; the system employing the ACOA achieved chieved 7Mbps, 7M 12Mbps and 9Mbps respectively better than the conventional tional syst system. The call blocking probability is also considered onsidered as shown in figure 13. INTERNATIONAL JOURNAL OF INNOVATIVE RESEARCH ESEARCH & DEVELOPMENT DOI No. : 10.24940/ijird/2021/v /v10/i7/JUL21027 Page 46 www.ijird.com July, 2021 Vol10 Issue 7 Figure 13: Call Blocking Probability t first 30s, the call blocking probability ty for bot both systems was fairly high, As can be seen in figure 13,, within the about 0.3 and 0.5 for the optimized system and an conventional system respectively. This covers th the early periods were there are still little influx of persons migrating ating into the cell. As the time increased up 150s, the cell is heavily loaded and at this point, the call blocking probability for eNB2 spiked to 0.8, which meant that there was as a high chance for calls being put through in the venue will be blocked. d. For the system employing the ACOA, the call blocking ocking probability pro dropped to about 0.1, making it very difficult for calls to be block blocked. Figure 14: Packet et Lo Loss Ratio in the Network for Both Systems for Scenar enario 3 From the simulation result,, the application applic of the ACOA shows a better reduction tion in the percentage of packet loss as shown in figure 14 for all the times mes consi considered. As at 30s into the simulation, thee optimized optimize system already showed 27.7% reduction in packet loss when en compared compar to the conventional system. From figure gure 15 it its obvious that as the time increased, the ACOA showed better er results (i.e., gave shorter paths) to the UE dense ense loca locations accept at the 120s. Comparing the conventional movement ment and the ACOA optimized movement, the optimized timized movement m clearly achieved better performance especially at the he 150s where w the packet loss reduction was about out 51.4%.A 51.4% third scenario is also considered where all the cells are congested ngested as a shown in figure 15. Figure 15: Simulation Scenario 3 INTERNATIONAL JOURNAL OF INNOVATIVE RESEARCH ESEARCH & DEVELOPMENT DOI No. : 10.24940/ijird/2021/v /v10/i7/JUL21027 Page 47 www.ijird.com July, 2021 Vol10 Issue 7 When the density of the cells was increased as shown in figure 16, the system performance formance at each of the eNB’s was monitored. This scenario is likened to when th there is a rally and everyone is clustered all ll over the sites. Figure ure 16: Call Blocking Probability for Scenario 3 high given traffic and over As can be seen in figure 16,, the call blocking probability for both systems is fairly h loaded nature of the cells. With the optimized system there is a slight reduction in the call blocking block probability. As at 30s, the optimized system showed a 11.11% 11% reduction reduc in the call blocking probability. Though ugh at 90 90s, the conventional system performed better than the optimized ed system by 5%, it can be seem that as the simulation mulation p progressed, the optimized system outperformed the conventional onal system syste by 6.7%. The performance at the various rious eNB’s eNB were also analyzed and shown in figure 18. Figure re 17: 1 Base Station Performances for Scenario 3 From figure 17, the results obtained showed s that the ACOA helped to utilize the he base stations st at full capacity. The throughput improvement provided by the ACO ACOA in eNB1 even with the density of the cell ell is abou about 27Mbps when compared to the conventional system, and 55Mbps Mbps for eeNB2 (which is about 9.7% increase). Beyond eyond improving imp the throughput, an added advantage of the optimized system ystem is that th the packet loss is also reduced as shown n in figure figur 19. Figure ure 1 18: Packet Loss Percentages for Scenario 3 From the simulation result,, the application applic of the ACOA shows a better reduction tion in the percentage of packet loss as shown in Figure 18for all the times considered. con Comparing the conventional movement ovement and the ACOA optimized movement, the optimized movement clearly achieved a better performance by approximately ately 38. 38.2% at 90 seconds than the normal moving technique. INTERNATIONAL JOURNAL OF INNOVATIVE RESEARCH ESEARCH & DEVELOPMENT DOI No. : 10.24940/ijird/2021/v /v10/i7/JUL21027 Page 48 www.ijird.com July, 2021 Vol10 Issue 7 6. Conclusion In this paper, the eNb’s of LTE networks have been designed in such a way that makes them able to be self-aware, self-adaptable and intelligent using the Ant Colony Optimization algorithm. This algorithm was inspired from the biological ant behaviour when in search for food. The ACOA was used to minimize the path trailed by the moving eNb’s employing the SON capability of LTE systems. This technique was deployed to increase the coverage capacity of the system by increasing the system throughput and reducing the packet loss. Three different scenarios were considered, results showed that in the first scenario, which is considered to be an ideal case, the optimized system improved the system throughput by 19.7% after 150s. The packet loss for this scenario was also reduced 38% at 150s, and by applying the antcolony optimization algorithm to the moving BTSs technique, improved the throughput significantly by up to 40 Mbps and reduced packet loss rate in the network by up to 19%. For the system employing the ACOA, the call blocking probability outperformed the conventional system by 45.9% as at 120s and by 22.2% when the time was 150s. In the second scenario, the system at eNB2 was loaded to congestion and some UE sparsely distributed at eNB3. This scenario is likened to when there is an event, say a football match and there are clusters of people very close to a particular eNB (in this case eNB2), and a close by cell is idle, the results obtained showed that the ACOA helped to utilize the idle base at eNB1, thus reducing the load at eNB2. The throughput improvement provided by the ACOA in eNB2 even with the density of the cell is about 2.5Mbps when compared to the conventional system, and 7Mbps and 72Mbps respectively for eNB1 and eNB3. This result to 820% utilization of an idle cell, while each UE achieved an improved throughput compared to the conventional system. For UE1, the optimized algorithm achieved 25Mbps better than the normal algorithm. For the system employing the ACOA, the call blocking probability dropped to about 0.1, making it very difficult for calls to be blocked. A third scenario is also considered where all the cells are congested. When the density of the cells was increased, the system performance at each of the eNB’s was monitored. This scenario is likened to when there is a rally and everyone is clustered all over the sites. The call blocking probability for both systems was fairly high, with the optimized system there was a slight reduction in the call blocking probability. As at 30s, the optimized system showed a 11.11% reduction in the call blocking probability. The throughput improvement provided by the ACOA in eNB1 even with the density of the cell was about 27Mbps when compared to the conventional system, and 55Mbps for eNB2 (which is about 9.7% increase). NS3 is an alternative network simulation tool that could be used in the study of network architectures, technologies and topologies. This software is worth trying out since it provides results comparable to that of OPNET. More scenarios with sophisticated configuration examples under different density settings are suggested to be considered for the future study of this work. With regards to the bio-inspired optimization techniques to move the appropriate BTS and balance the number of mobile users allocated to each BTS, further study about other bio-inspired techniques to build the SON capabilities is recommended. 7. References i. D. Lopez-Perez, I. Guvenc, G. De la Roche, M. Kountouris, T. Q. Quek, and J. Zhang, ‘Enhanced intercell interference coordination challenges in heterogeneous networks,’ IEEE Wireless Communications, vol. 18, pp. 22-30, 2011 ii. Sesia. S, Toufik. I, and Baker .M, LTE - The UMTS Long Term Evolution: From Theory toPractice, John Wiley & Sons, Ltd, Chichester, UK, 2012. iii. K Son, H Kim, Y Yi, B Krishna machari, ‘Base station operation and user Association mechanisms for energy-delay tradeoffs in green cellular Networks’. Selected Areas Commun. IEEE J. 29(8), 1525–1536, 2011Attar, V. Krishnamurthy, and O. N. Gharehshiran, ‘Interference management using cognitive base-stations for UMTS LTE,’IEEE Communications Magazine, vol. 49, pp. 152-159, 2011 iv. Yen-Wei Kuo and Li-Der Chou, ‘Fuzzy-based coverage and capacity scheme in LTE heterogeneous networks’ Journal of the Chinese Institute of Engineers, https://doi.org/10.1080/02533839.2017.1384323, 2017 v. RebenKurda, ‘Heterogeneous networks: Fair power allocation in LTE-A uplink scenarios’https://doi.org/10.1371/journal.pone.0252421, Accessed 16th June 2021 vi. Andrea Tassi, ChadiKhirallah, DejanVukobratovic, Francesco Chiti, John S. Thompson and Romano Fantacci, ‘Resource Allocation Strategies for NetworkCodedVideo Broadcasting Services over LTE-Advanced’ IEEE Transactions onVehicular Technology, pages 1–12, 2014 vii. M. R. Tabany and C. G. Guy, ‘Design and implement delay-aware QoS scheme for 3GPP LTE/LTE-A networks for mixed traffic flow,’ IEEE Symposium on Computers and Communication (ISCC), pp.38-44, 2015 viii. Samira Achki, Layla Aziz, Fatima Gharnati, Abdellah Ait Ouahman, ’User Association Strategy for Energy Efficiency and Interference Mitigation of Heterogeneous Networks’, Advances in Materials Science and Engineering, vol. 2, 2020, Article ID 7018727, https://doi.org/10.1155/2020/7018727 ix. M. Ünal et al‘Optimization of PID Controllers Using AC and GA’, SCI 449, pp. 31–35., 2012 INTERNATIONAL JOURNAL OF INNOVATIVE RESEARCH & DEVELOPMENT DOI No. : 10.24940/ijird/2021/v10/i7/JUL21027 Page 49