Capacity and Coverage Dimensioning for 5G Standalone Mixed-Cell Architecture: An Impact of Using Existing 4G Infrastructure

Abstract

With the increasing demand for expected data volume daily, current telecommunications infrastructure can not meet requirements without using enhanced technologies adopted by 5G and beyond networks. Due to their diverse features, 5G technologies and services will be phenomenal in the coming days. Proper planning procedures are to be adopted to provide cost-effective and quality telecommunication services. In this paper, we planned 5G network deployment in two frequency ranges, 3.5 GHz and 28 GHz, using a mixed cell structure. We used metaheuristic approaches such as Grey Wolf Optimization (GWO), Sparrow Search Algorithm (SSA), Whale Optimization Algorithm (WOA), Marine Predator Algorithm (MPA), Particle Swarm Optimization (PSO), and Ant Lion Optimization (ALO) for optimizing the locations of remote radio units. The comparative analysis of metaheuristic algorithms shows that the proposed network is efficient in providing an average data rate of 50 Mbps, can meet the coverage requirements of at least 98%, and meets quality-of-service requirements. We carried out the case study for an urban area and another suburban area of Kathmandu Valley, Nepal. We analyzed the outcomes of 5G greenfield deployment and 5G deployment using existing 4G infrastructure. Deploying 5G networks using existing 4G infrastructure, resources can be saved up to 33.7% and 54.2% in urban and suburban areas, respectively.

1. Introduction

There were about 4.9 billion active Internet users in 2023 and 5.45 billion up to July 2024 worldwide, i.e., 62% and 67.1% of the world’s total population, respectively [1,2]. The usage trend of data volume is increasing rapidly in the world. The total volume of data created, copied, and consumed in 2010 and 2020 was about 2 and 64.2 zettabytes, respectively. It is estimated that it will be 181 zettabytes in 2025. After the COVID-19 pandemic, the surge increase in data usage is due to online work, learning, and entertainment work carried out by people from home [3]. The demand for higher data rates is also increasing in the world due to the usage of Over-the-Top (OTT) applications, the connectivity of the Internet of Things (IoT), and the usage of innovative applications such as intelligent transportation [4], innovative medical treatment [5], and smart metering [6]. 4G LTE wireless technology is insufficient to provide the current demand for data rate and low-latency reliable services [7]. Critical applications such as e-health and other emergency services need reliable random access techniques [8]. As the demand for higher data rates in dense areas and the connectivity of IoT devices is increasing day by day, the 4G network is not sufficient due to its higher latency and lack of intelligent processing despite 4G technology having better features as compared to previous generations [9]. 5G wireless communication is designed to address the limitations of previous generations. It has key features such as a data speed of gigabit per second (Gbps), massive connectivity, ultra-low latency of 1 ms [10], massive network capacity, increased availability, and a more uniform user experience to more users. 5G enables a new kind of network to connect virtually everyone and everything together, including machines, objects, and devices. Higher performance and improved efficiency empower new user experiences and connect new industries [11]. In 5G, these diverse requirements are fulfilled by allowing flexible frame structure and multiple numerologies or different Sub-Carrier Spacing (SCS) based on the nature of services, with a wide range of frequencies compared to previous generations, from 0.5 GHz to 100 GHz [12,13]. Cognitive Radio (CR) is used in 5G and offers channel assignment by intelligently detecting ideal and busy channels for the optimum usage of spectrum resources using efficient spectrum sensing methods [14,15]. 5G network utilizing Software-Defined Networking (SDN) allows dynamic bandwidth allocation, improves latency, and ensures optimal data flow [16]. Massive Multiple Input Multiple Output (mMIMO) and beam-forming technology are keys to enhancing throughput, coverage, and connectivity density of IoT devices in 5G networks [17,18].

5G has numerous advantages and diversified features. However, planning appropriately to exploit these features at a lower cost is essential. Capacity and coverage contrast with each other as operating frequency increases. Capacity in this paper refers to the total throughput of Remote Radio Units (RRUs), and capacity dimensioning refers to allocating data speed of at least 50 Mbps to each user connected. Capacity increases with an increase in frequency but decreases in the coverage area. We need to plan an adaptive approach to balance both the constraints [19]. Transitioning to 5G at once is challenging and costly [20], and thus, it requires a huge amount of investment due to more resource requirements such as more BTS due to small area coverage using millimeter wave [21]. In 5G New Radio (NR), different frequency spectrum bands are categorized into two groups: (a) Frequency Range 1 (FR1) (410 MHz–7.125 GHz), which includes sub-6 GHz mid-band frequencies), and (b) Frequency Range 2 (FR2), which includes higher-frequency bands, i.e., millimeter bands (24 GHz–100 GHz). FR1 frequency bands have more comprehensive coverage but limited data rate capabilities and are suitable for rural areas. In contrast, millimeter bands can deliver higher data rates of the order of multi-Gbps but have limited coverage [22] and are suitable for highly dense urban areas. Due to the small coverage, there is a need for more RRUs to be installed, which will be costlier and require higher CAPEX, IMPEX, and OPEX. A trade-off between capacity and coverage can be achieved using both millimeter-band and mid-band frequencies, such as the sub-6 GHz band. Proper planning, preliminary preparations, and choice of suitable frequency bands based on the population density are required to meet data rate requirements, comprehensive coverage, and QoS requirements.

In Nepal, 5G has not yet been deployed. There is already considerable investment in deploying 4G infrastructure. 5G greenfield deployment will incur a large amount of CAPEX, and there is a significant amount of savings if we optimize the 5G BTS, keeping the existing 4G BTS in the same positions and ensuring the required QoS. In this paper, we study 5G greenfield deployment optimization and analyze the impact of 5G SA deployment leveraging the existing 4G infrastructure. We propose a novel mixed cell approach to deploy both macro-RRU (RRU_M) and micro-RRU (RRU_m) appropriately based on the demand of data requirement and population distribution for the 5G network.

The main objective is to find the optimum placement of RRU_M and RRU_m to meet data rate requirements and ensure all users’ coverage. Due to the smaller area coverage of RRU_m using millimeter wave, there is a need for a massive number of RRU_m, resulting in excessive investment then strategically positioning RRUs in areas with high demand while minimizing overlap, can reduce cost. We first calculate the link budget for 3.5 GHz and 28 GHz simultaneously and throughput for both frequency bands. We initially implemented a random distribution of RRU_m and RRU_M based on the calculated link budget. We implemented PSO [23], GWO [24], WOA [25], MPA [26], SSA [27], and ALO [28] and carried out a comparative study for optimum placement of RRU_m and RRU_M. We optimized 5G RRUs placement without relocating the existing 4G RRUs using metaheuristic algorithms for 5G SA deployment. Then, we analyzed the impact of using the existing 4G infrastructure with 5G greenfield deployment. The proposed approach ensures 98% coverage and an average data rate of 50 Mbps. To achieve the capacity and coverage requirement, we summarize our contributions as follows:

• We performed appropriate system parameter definition and user service requirement for frequency bands FR1 (3.5 GHz) for macro-RRU (RRU_M), and FR2 (28 GHz) for micro-RRU (RRU_m).
• Next, we performed dimensioning for cell capacity and cell coverage requirements for the required QoS, determining the preliminary required number of RRU_M and RRU_m.
• We simulated two different use-case scenarios for 5G SA greenfield deployment (Case I) and 5G deployment using existing 4G infrastructure (Case II), and carried out a comparative analysis.

The rest of this paper is organized as follows. Section 2 presents the background and related work. Section 3 describes the system model formulation. Section 4 discusses the experimental setup and evaluation. Section 5 discusses the outcome of the experimental results, while Section 6 concludes the paper.

2. Background and Related Work

Metaheuristic approaches are problem-independent, hit-and-trial-based population search processes to intelligently obtain the optimal and best solutions. Metaheuristic approaches can be distinguished as local search, based on population or random search and constructive [29] or breaking the problem into different sub-problems [30], and are used for multi-point global optimization. PSO is a technique inspired by the coordinated movements of a bird flock. While a bird flies and searches randomly for food, for instance, all birds in the flock can share their discovery and help the flock obtain the best hunt [31]. The SSA mimics the foraging and anti-predation behavior of the sparrow population. The algorithm formulates the search for better food sources by the sparrow population as an optimization algorithm [32]. The WOA mimics the social behavior of humpback whales, drawing inspiration from their unique bubble-net hunting strategy. This foraging behavior involves whales hunting krill or small fish schools near the surface by creating distinctive bubbles along a circular or “9”-shaped path [25]. The GWO simulates the leadership hierarchy and hunting behavior of grey wolves in nature categorizing grey wolves into four types, alpha, beta, delta, and omega, to represent the leadership structure [24]. The MPA mimics the foraging behaviors of marine organisms such as sharks, toucans, sunfish, and swordfish while searching for food and follows the foraging strategy of Lévy and Brownian movements with an optimal encounter rate policy in marine ecosystems [26]. ALO mimics the hunting mechanism of antlions in nature where ants explore their prey by selecting antlions, random walks of ants, and adapting to shrinking boundaries of antlions [28].

Different scholars have presented innovative ideas about wireless network cellular planning. In [33], base station deployment optimization techniques are investigated in interested areas using Simulated Annealing (SA) for heterogeneous cells in LTE networks. An efficient, optimized cellular planning approach is proposed using the stochastic optimization method in [34] and under demand uncertainty in [35]. A greedy-based algorithm maximizes power-saving indicators of heterogeneous networks in [36]. A comparative study between the Poisson point process and the Poisson hole process is carried out in [37] for pico base stations in a two-tier heterogeneous network, concluding energy efficiency improvement can be achieved based on appropriately selecting the transmit power of pico base stations. Multi-objective optimization algorithms are proposed in [38] considering capacity, coverage, and cost considering interference as a significant constraint, adding local search and decomposition strategy to the genetic algorithm for 4G heterogeneous network and finding the most feasible solutions for Base Transceiver Station (BTS) planning with faster computations. In [39], cellular network planning is carried out for a micro-macro-relay combination scenario considering coverage, capacity, and cost scenario and reusing existing sites in mixed cell structure and concluded that the approach could reduce the cost of Mbps per square km up to 20 times. In [40], the concept of smart grid integration, passive cooling is proposed for central offices, and smart sleep modes for energy efficient green 5G network and beyond are proposed and carried out techno-economic analysis for the migration of existing networks. A joint optimized procedure is proposed for cell planning and fiber back-haul designing problems in [41]. A non-dominated sorting genetic algorithm is proposed to deploy a minimum number of base stations using wired and wireless back-haul systems satisfying cell capacity and coverage and a proposed cost-minimizing approach in a 5G network.

Different researchers carried out techno-economic and sensitivity analyses for cost-effective deployment of 5G networks. Different approaches, algorithms, and models are suggested and developed to minimize the CAPEX, IMPEX, and OPEX to enhance the QoS of 5G networks. Open mobile edge cloud vision was introduced in the white paper [42] and focuses on the impact of SDN and Network Functions Virtualization (NFV) on telecommunication networks. Technical challenges have been investigated while migration and security policy issues are based on the IEEE SDN Initiative. A techno-economic analysis is performed in [43] to improve indoor coverage by using a small cell network, and this approach is economically viable and managed by a third-party micro-operator as a neutral host. Priority-based parameters such as performance, business, and sub-criteria such as low latency, high reliability, and high data rate underperformance, privacy, and security under acceptance and SDN and NFV under technology are identified in [44]. This results from road-mapping activities of technologies, financial, regulatory, and standardization issues, and surveys of pairwise comparison based on the fuzzy Analytical Hierarchy Process (AHP) method among experts in the CHARISHMA project of the EU.

A techno-economic analysis of the 5G network is accomplished in [45] for the urban area of South London using macro, micro, and hotspots together with 802.11ac access points for the indoor and outdoor environments. Cost modeling incorporating the cost of base stations, transport, spectrum, operation, and maintenance has been carried out at different frequencies. The cost of a 5G network in urban areas increased significantly compared to 4G LTE, but capacity and coverage improved significantly. Using the frequency of a 700 MHz macro cell, blanket coverage can be provided but cannot offer 100 Mbps capacity. Using a 3.5 GHz micro cell, the outdoor capacity becomes significant but relatively lowers indoor capacity. Using a millimeter wave in a 5G network, the capacity can be increased to a hundred or thousand times higher with the integration of the 820.11ac access point but at the expense of higher cost. The demand-driven and supply-driven analysis is conducted in [46] for the different users experiencing demand of 30, 100, and 300 Mbps using the Netherlands’ use case scenario. Providing higher speed in all areas is not cost-effective, and targeted small-cell deployment is cost-effective in dense areas. The first 30% most densely populated area covers 5% of investment, the next 40% population of the suburban area accounts for 20% of investment, and the remaining 30% rural population needs 75% of total investment. An improvement of 40% is possible by utilizing a spectrum integration strategy through supply-driven analysis.

The existing literature lacks a study on the proper usage of existing 4G infrastructure and its impact while deploying 5G networks. We propose deploying 5G networks with or without using existing infrastructure and subsequent impact. We are motivated by the revision carried out in this section to design an optimized mixed cell architecture to provide the required data rate and meet coverage constraints.

3. System Model and Problem Formulations

In this section, a proposed system model is defined and proposed as an optimization problem formulation. 5G has diversified features such as higher speed, massive connectivity, and very high reliable service, and to meet these requirements, 5G introduces new technologies such as SDN, NFV, network slicing, MIMO, millimeter band, and Heterogeneous Network (HetNet). Due to the use of millimeter waves, a higher data rate can be achieved. Still, it has a smaller coverage area, and there is a need for a more significant number of millimeter wave BTS, resulting in higher costs. In highly dense areas, it is advantageous to deploy the millimeter band, but if there is an uneven distribution of population, the coverage becomes an issue, and the millimeter wave will not be cost-effective; there is a need to use mid-band frequencies to have broader coverage.

In this paper, a mixed cell architecture of macro-RRU using sub-band 3.5 GHZ and micro-RRU using 28 GHZ is implemented to provide the required data rate and ensure full coverage. Figure 1 shows the proposed mixed cell structure, where RRU_m are placed in dense areas, and RRU_M are placed in sparsely populated areas.

We calculated the link budget for 3.5 GHz and 28 GHz, considering different losses encountered while propagating the 5G signal. We then computed the Maximum Allowable Path Loss (MAPL) and the corresponding maximum radius for coverage, maintaining QoS parameters for uplink and downlink. Capacity dimensioning was performed based on the demographic distribution and users’ required data rate of at least 50 Mbps. The capacity and coverage dimensioning was carried out in 5G greenfield and 5G brownfield, i.e., using existing 4G infrastructure. The minimum number of RRUs after coverage and capacity dimensioning are fixed, and their best placement is optimized using metaheuristic algorithms. The best-optimized location of RRUs among them is selected, maintaining the requirement of both an average data rate of 50 Mbps and coverage of at least 98%. If the criteria are unmet due to coverage holes, the RRUs are added iteratively. The comparative analysis of 5G greenfield and brownfield is performed, and mixed deployment of both FR1 and FR2 is implemented if it is best result-oriented. Carrier Aggregation (CA) is carried out to minimize the number of RRUs without degrading QoS. Figure 2 shows the proposed framework of 5G network RRU placement optimization, demonstrating the thoroughness of our approach.

Table 1. Parameter notations and description.

Notations	Descriptions	Notations	Descriptions
UMa	Urban macro cells	UMi	Urban micro cells
$R_{{RRU}_M}$	Cell radius of macro-RRU	$R_{{RRU}_m}$	Cell radius of micro-RRU
$MAP{L}_{{RRU}_M}$	MAPL for macro-RRU	$MAP{L}_{{RRU}_m}$	MAPL for micro-RRU
$N_{Mcov}$	Number of RRUM for coverage	$N_{mcov}$	Number of RRUm for coverage
$N_{Mcap}$	Number of RRUM for capacity	$N_{mcap}$	Number of RRUm for capacity
$P{L}_{UMa}$	Path loss model for UMa	$P{L}_{UMi}$	Path loss model for UMi
$N_{{RRU}_M}$	Number of users served by macro-RRU	$N_{{RRU}_m}$	Number of users served by micro-RRU
${NUM}_{{RRU}_M}$	Total number of set of macro-RRU	${NUM}_{{RRU}_m}$	Total number of set of micro-RRU
${RRU}_M(x^i,y^i)$	Location of $i^{th}$ macro-RRU	${RRU}_m(x^i,y^i)$	Location of $i^{th}$ micro-RRU
UE($x^i,y^i$)	Location of $i^{th}$ UE	${NU}_{total}$	Total number of users to be served

shows notations and corresponding descriptions. Based on the third generation partnership project (3GPP TR-38.901) specifications [47], we carry out link budget calculations for both the frequencies FR1 (3.5 GHz) and FR2 (28 GHz), respectively.

3.1. Link Budget Analysis

Link budget calculation is performed for both FR1 (3.5 GHz) and FR2 (28 GHz) separately. Maximum radius and MAPL for both frequencies are calculated based on (3GPP TR-38.901) specifications [47]. The placement is based on the Cartesian plane coordinates RRU ($x^i,y^i$). RRU_m and RRU_M are initially placed randomly, and the minimum number of RRU_m and RRU_M is calculated as per link budget analysis to satisfy the capacity and coverage constraints. The user equipment (UE) location is denoted as UE ($x^i,y^i$). The population distribution is considered in two scenarios.

Scenario I (Urban): The population is distributed heavily in urban areas, 20,000 per square kilometer [48]. An area of 4 km² was taken for simulation. A case study of the area of Baneshwor and Gaushala in Kathmandu was chosen.

Scenario II (Suburban): The suburban area with a population density of about 7500 per square kilometer was chosen for study. An area of 16 km² of Nepaltaar and Tokha in Kathmandu was considered as a case study and simulation analysis [49,50].

RRU_m and RRU_M have $N_{sector}$ numbers of sector antenna. All the UEs have the same parameters, such as transmit power, cable loss, temperature, interference margin, bandwidth, and height. It is for clarity and uniformity, as different parameters can also be considered for different UE types. The Effective Isotropic Radiating Power (EIRP) for RRU_M, RRU_m, and UE is calculated using the formula listed in Equation (1).

$$EIRP=TXP+AG-CL$$

(1)

where TXP = Transmit Power, AG = Antenna Gain, and CL = Cable Loss.

The Thermal Noise (TN) for RRU_M, RRU_m, and UE is calculated [51] using Equation (2).

$$TN\left[\mathrm{dBm}\right]=10\times {log}_{10}(T\times K\times BW)$$

(2)

where T = Temperature in Kelvin, K = Boltzmann’s constant, and BW = Bandwidth in Hz.

Now, we calculate the radio link budget for each micro cell and macro cell in each downlink and uplink side. We calculate MAPL from RRU_m to UEs and RRU_M to UEs for downlink and vice versa for uplink. MAPL is the total energy loss of transmitted signal from the transmitter in both uplink and downlink transmission, maintaining adequate signal quality [52]. The MAPL for uplink and downlink is calculated using Equation (3).

$$MAPL=EIRP+UEg-PL-FL-BL-IM-RIL-SFM-NF-DT-RBL$$

(3)

where UEg = UE Gain, PL = Penetration Loss, FL = Foliage Loss, BL = Body Block Loss, IM = Interference Margin, RIL = Rain/Ice Loss, SFM = Slow Fading Margin, NF = Noise Figure, DT = Demodulation Threshold SNR, and RBL = 10 ${log}_{10}$(SCQ). Subcarrier Quantity (SCQ) is formulated as the product of subcarrier per resource block and number of resource blocks as specified by 3GPP TS-38.101-2 [53]. For the 100 MHz and 200 MHz bandwidths, SCQ is calculated as $12\times 273=3276$ and $12\times 264=3168$, respectively [54].

Table 2. RRU_M, RRU_m, and UE specifications [55,56].

Parameters	RRUM	RRUm	UE
Frequency Band	n78	n257	–
Frequency	3.5 GHz	28 GHz	–
Frequency Range	FR1	FR2 (mmWave)	–
Duplex Mode	TDD	TDD	–
Channel BW (MHz)	5, 10, 20, 40, 80, 100	50, 100, 200, 400	–
UL-DL Range (MHz)	3300–3800	26,500–29,500	–
Height (m)	23	8	1.5
Bandwidth (MHz)	100	200	10

shows the specification of RRU_m, RRU_M, and UE, and

Table 3. RRU_M and RRU_m downlink and uplink link budget parameters.

		RRUM (3.5 GHz)		RRUm (28 GHz)
Parameters	=	Uplink	Downlink	Uplink	Downlink
Transmit Power (dBm)	A	-	49	-	36
Subcarrier Spacing (KHz/MHZ)	SCS/BW	30/100	30/100	60/200	60/200
Transmit Antenna Gain (dBi)	B	11	11	21	21
Transmitter Cable Loss (dB)	C	2	2	0	0
Penetration Loss (dB)	D	28	28	9.5	9.5
Foliage Loss (dB)	E	8	8	7.74	7.74
Body Block Loss (dB)	F	4	4	8	8
Interference Margin (dB)	G	6	6	2	1
Rain/Ice Margin (dB)	H	0	0	3	3
Slow Fading Margin (dB)	I	6	3	4	2
Temperature (Kelvin)	K	290	290	290	290
Thermal Noise (dBm)	L	−129.206	−129.206	−126.195	−126.195
Noise figure UE (dB)	M	7	7	8	8
Demodulation threshold SNR (dB)	N	−1	−1	−1	−1

lists the values for calculation of MAPL.

3.2. Capacity and Coverage Dimensioning

We carry out coverage and capacity dimensioning based on the link budget profile. For coverage dimensioning, we calculate the maximum radius that the given specification allows. For that, we calculate the path loss as specified by 3GPP TR-38.901, and then we calculate the area covered by one RRU_m and RRU_M. After calculating the area covered by each RRU_m and RRU_M, we calculate the required number of RRU_m and RRU_M that ensures coverage of all UEs. No UEs in that area are in a dead zone. Next, we carry out the capacity dimensioning based on the link budget analysis for both uplink and downlink and place RRU_m and RRU_M properly.

3.2.1. Coverage Dimensioning

The path loss model to carry out the minimum radius of macro-RRU is based on 3GPP TR-38.901 specifications [47], where the path loss model for the Urban Macro (UMa) cell is given as Equation (4).

$${PL}^{UMa}\left[\mathrm{dB}\right]=max({PL}_{(UMa-LOS)},{PL\prime }_{UMa-NLOS})\mathrm{for}10 \mathrm{m}\le d_{2D}\le 5 \mathrm{km}$$

(4)

where

$${P{L}^{\prime }}_{(UMa-NLOS)}=13.54+39.08{log}_{10}\left(d_{3D}\right)+20{log}_{10}\left(f_c\right)-0.6(h_{UE}-1.5)$$

(5)

and

$${PL}_{(UMa-LOS)}=\left\{\begin{array}{cc}28+22{log}_{10}\left(d_{3D}\right)+20{log}_{10}\left({f_c}^{UMa}\right)\hfill & \mathrm{for}10 \mathrm{m}\le d_{2D}\le d_{BP}^{\prime }\hfill \\ 28+40{log}_{10}\left(d_{3D}\right)+20{log}_{10}\left({f_c}^{UMa}\right)-\hfill \\ 9{log}_{10}({\left(d_{BP}^{\prime }\right)}^2+{(h_{BS}-h_{UE})}^2)\hfill & \mathrm{for}{d}_{BP}^{\prime }\le d_{2D}\le 5 \mathrm{km}\hfill \end{array}\right.$$

(6)

The minimum radius for UMa-LOS is calculated using,

$${Radius}_{(UMa-LOS)}={10}^{({MAPL}_{UMa-LOS}-28-20{log}_{10}\left(f_c^{UMa}\right))/22}$$

(7)

where $d_{3D}$ is the 3D distance between transmitter, UE, $d_{2D}$ is the 2D distance between the transmitter and UE, $f_c$ is the operating center frequency in GHZ, and $d_{BP}^{\prime }$ is the breakpoint distance [57] and is given as $d_{BP}^{\prime }$ = ${4h_{BS}}^{\prime }\times {h_{UE}}^{\prime }\times f_c\times {10}^{9/c}$. The 3D distance between the transmitter and UE is calculated using the relation,

$$d_{3D}=\sqrt{d_{2D}^2+{(h_{BS}-h_{UE})}^2}$$

(8)

We carry out link budget analysis using Equation (3) and find MAPL for macro-RRU for uplink and downlink as ${MAPL}_{uplink}=101.05$ dB and ${MAPL}_{downlink}=98.05$ dBn respectively. The 3D cell radius calculated for uplink and downlink using Equation (7) is 669.829 m and 489.326 m, respectively. Then, the 2D cell radius is calculated for uplink and downlink using Equation (8) and is 669.828 m and 488.853 m, respectively. Similarly, the path loss for Urban Micro (UMi) cell based on 3GPP TR-38.901 specifications [47] is given as,

$${PL}^{UMi}\left[\mathrm{dB}\right]=max({PL}_{(UMi-LOS)},{P{L}^{\prime }}_{UMi-NLOS})\mathrm{for}10 \mathrm{m}\le d_{2D}\le 5 \mathrm{km}$$

(9)

where

$${PL}_{UMi-LOS}\left[\mathrm{dB}\right]=\left\{\begin{array}{cc}32.4+21{log}_{10}\left(d_{3D}\right)+20{log}_{10}\left({f_c}^{UMi}\right)\hfill & \mathrm{for}10 \mathrm{m}\le d_{2D}\le d_{BP}^{\prime }\hfill \\ 32.4+40{log}_{10}\left(d_{3D}\right)+20{log}_{10}\left({f_c}^{UMi}\right)-\hfill \\ 9.5{log}_{10}({\left(d_{BP}^{\prime }\right)}^2+{(h_{BS}-h_{UE})}^2))\hfill & \mathrm{for}{d}_{BP}^{\prime }\le d_{2D}\le 5 \mathrm{km}\hfill \end{array}\right.$$

(10)

and

$${P{L}^{\prime }}_{UMi-NLOS}\left[\mathrm{dB}\right]=22.4+35.3{log}_{10}\left(d_{3D}\right)+21.3{log}_{10}\left(f_c^{UMi}\right)-0.3((h_{UE}-1.5)$$

(11)

The minimum radius for UMi-LOS is calculated using,

$${Radius}_{UMi-LOS}={10}^{({MAPL}_{UMi-LOS}-32.4-20{log}_{10}\left(f_c^{UMa}\right))/21}$$

(12)

After calculating $MAP{L}_{{RRU}_m}$ using Equation (3), we calculated minimum 3D radius using Equation (12) and 2D radius using Equation (8) and the 2D cell radius for UMi for micro-RRU is 146.90 m. The calculated values for MAPL, 2D cell radius, 3D radius, and minimum required number of both micro and macro-RRUs in urban and suburban areas for both uplink and downlink are included in

Table 4. RRU_M and RRU_m requirements calculations.

	RRUM (3.5 GHz)		RRUm (28 GHz)
Parameters	Uplink	Downlink	Uplink	Downlink
MAPL (dB)	101.052	98.052	109.947	106.947
$d_{BP}^{\prime }$ (m)	513.333	513.333	1306.666	1306.666
$d_{3D}$ (m)	669.829	489.326	206.304	148.467
$d_{2D}$ (m)	669.828	488.853	206.299	146.902
Total urban area (km2)	4	4	4	4
Total Suburban area (km2)	16	16	16	16
Coverage dimensioning
Minimum RRU (Urban)	2.83 (≈3)	5.32 (≈6)	29.91 (≈30)	59.00 (=59)
Minimum RRU (Suburban)	11.35 (≈12)	21.31 (≈22)	119.66 (≈120)	236.00 (=236)
Capacity dimensioning
Minimum RRU (Urban)	69.44 (≈70)	138.88 (≈139)	34.72 (≈35)	69.44 (≈70)
Minimum RRU (Suburban)	17.36 (≈18)	34.72 (≈35)	8.68 (≈9)	17.36 (≈18)

for both micro-RRU and macro-RRU.

The surface area for each cell i, $A_{Mcell}^i$ is expressed using the formula, $A_{Mcell}^i$ = $\pi {\left({RRU}_m^i\right)}^2$.

After calculating minimum cell radius and surface area for each cell, the total number of RRU_m and RRU_M is calculated as,

$$N_{Mcov}={\displaystyle \frac{A_{total}}{A_{Mcell}}}$$

(13)

and

$$N_{mcov}={\displaystyle \frac{A_{total}}{A_{mcell}}}$$

(14)

where $A_{total}$ is the total case-studied area, and $A_{mcell}$ and $A_{Mcell}$ are areas covered by each RRU_m and RRU_M per sector, respectively.

3.2.2. Capacity Dimensioning

For capacity planning, first of all, the total number of population (UEs), ${NU}_{total}$ of the total serving area $A_{total}$, is estimated. It is based on the population density of the serving area. We assume that total throughput is equally shared between the UEs of particular cells, ensuring the required data rate. Let $T_{mmax}$, $N_{sec}$ and ${DR}_{UE}$ be the total throughput of an RRU_m, number of sector antenna of RRU_m and data rate required per UE then, the number of UEs to be served by an RRU_m, $N_{{RRU}_m}$ is expressed as,

$$N_{{RRU}_m}={\displaystyle \frac{T_{mmax}\times N_{sec}}{{DR}_{UE}}}$$

(15)

The throughput of RRUs is calculated using Equation (16), The throughput calculation for RRU_m and RRU_M are based on the 3Gpp 38.306 specification [58] is expressed as,

$$T_{max}\left(Mbps\right)={10}^{-6}{\sum }_{j=1}^J(v_{Layers}^{\left(j\right)}\times Q_m^{\left(j\right)}\times f^{\left(j\right)}\times R_{\left(max\right)}\times {\displaystyle \frac{N_{PRB}^{BW\left(j\right),\mu }}{T_s^{\mu }}}\times (1-{OH}^j))$$

(16)

where J is the number of aggregated component carriers, $v_{Layers}^{\left(j\right)}$ is the maximum number of supported layers; $Q_m^{\left(j\right)}$ is the maximum supported modulation order; $f^{\left(j\right)}$ is the scaling factor; $\mu$ is the numerology as defined in TS 38.211; $T_s^{\mu }$ is the average OFDM symbol duration in a subframe for numerology, i.e., $T_s^{\mu }={\displaystyle \frac{{10}^{-3}}{14\times 2^{\mu }}}$; ${\displaystyle \frac{N_{PRB}^{BW\left(j\right);\mu }}{T_s^{\mu }}}$ is the maximum RB allocation, where BW(j) is the UE supported maximum bandwidth and ${OH}^j$ is the overhead; and $R_{max}$ is the target coding rate.

The number of users to be served by an RRU_M to ensure the required data rate (required data rate per UE for RRU_M is the same as that of RRU_m is also calculated using the same formula with relevant parameters,

$$N_{{RRU}_M}={\displaystyle \frac{T_{Mmax}\times N_{sec}}{{DR}_{UE}}}$$

(17)

The total throughput of an RRU_m or RRU_M in one sector antenna in terms of bandwidth and spectral efficiency is expressed as $T_{max}$ = Bandwidth × Spectral Efficiency. The values of $N_{Mcap}$ and $N_{mcap}$ are calculated by dividing the total number of population in the case study area by $N_{{RRU}_M}$ and $N_{{RRU}_m}$, respectively. Now, we calculate the approximate initial number of RRU_m and RRU_M using,

$${NUM}_{{RRU}_m}=max(N_{mcov},N_{mcap})$$

(18)

and

$${NUM}_{{RRU}_M}=max(N_{Mcov},N_{Mcap})$$

(19)

Finally, the total number of RRU_m and RRU_M together is calculated as,

$${NUM}_{total}={NUM}_{{RRU}_m}+{NUM}_{{RRU}_M}$$

(20)

3.2.3. Optimization Problem and Constraints

This paper aims to optimize the location of RRU_m and RRU_M with their minimum number for both dense urban and suburban areas, fulfilling the required data rate and ensuring coverage. We assume all users are connected with at least one of RRU_m or RRU_M, and we also assume that all users have the same demand of equal data rate. Initially, we distribute randomly generated RRU_M and RRU_m for 5G greenfield. If we distribute enough RRU_m and RRU_M, we can achieve the required data rate and have good coverage, but we need to spend more cost due to more resources and will be costly services. So, we use optimization techniques to locate RRU_m and RRU_M properly using metaheuristic algorithms fulfilling the required constraints. Now, we define constraints and decision variables and set and formulate objective function and optimization problems.

Binary Constraints: Binary variables are used for sensing where there is the connection of UE to RRU_m or RRU_M or not. To address the connectivity with RRU_m or RRU_M, ${\xi }_{RRU}$$(x^i,y^i)$ is used to check whether the UE located at $(x^i,y^i)$ is connected to any RRU or not. i.e.,

$${\xi }_{RRU}(x^i,y^i)=\left\{\begin{array}{cc}1\hfill & \text{if there is coverage from any RRU}\hfill \\ 0\hfill & \text{otherwise}\hfill \end{array}\right.$$

(21)

But any UE can be connected with multiple RRU_m or RRU_M. But if there is no connectivity, then there is a coverage issue.

Coverage Constraints: Coverage constraints are ensured for all interest areas when the following equation satisfies,

$$\begin{array}{c}\hfill {\displaystyle \frac{{\sum }_{i=1}^{{NUM}_{{RRU}_M}}{\xi }_{{RRU}_M}(x^i,y^i)+{\sum }_{i=1}^{{NUM}_{{RRU}_m}}{\xi }_{{RRU}_m}(x^i,y^i)}{{NU}_{total}}}\ge \alpha {NU}_{total}\end{array}$$

(22)

where $\alpha$ is the tolerance factor for relaxing coverage constraints.

Capacity Constraints: For capacity constraints $\beta$, the parameters are to be considered: the sector of the antenna and ${NUM}_{total}$. Capacity constraints should satisfy the condition,

$$\begin{array}{c}\hfill {\displaystyle \frac{{\sum }_{j=1}^{{NUM}_{{RRU}_M}}{\sum }_{i=1}^{N_{{RRU}_M}}{\xi }_{{RRU}_M^j}(x^i,y^i)+{\sum }_{k=1}^{{NUM}_{{RRU}_m}}{\sum }_{l=1}^{N_{{RRU}_m}}{\xi }_{{RRU}_m^k}(x^l,y^l)}{{NU}_{total}}}\ge \beta {NU}_{total}\end{array}$$

(23)

where $\beta$ is the tolerance factor for relaxing capacity constraints.

Optimization Problem: The main aim of optimization problem is to minimize the number of RRU_m and RRU_M ensuring required data rate and coverage. The optimization problem (OP) is defined as,

$${OP:}_{{\rho }_j^i\in \{0,1\}}^{minimize},(x_j^i,y_j^i)\sum _{i\in \{{RRU}_M,{RRU}_m\}}\sum _{i=1}^{{NUM}_{total}}{\rho }_j^i$$

(24)

subject to the condition in Equations (22) and (23). Since the optimization problem is an NP-hard problem and computationally complex, we carry out optimization using PSO, GWO, SSA, WOA, ALO, and MPA simultaneously for comparative analysis and select the best-optimized locations of RRU_m and RRU_M.

3.3. Metaheuristic-Approach-Based Algorithms

This paper used PSO, SSA, WOA, MPA, GWO, and ALO simultaneously to optimize placement. Comparative studies for performance evaluation with detailed notations and algorithms are illustrated in PSO [23], SSA [27], WOA [25], MPA [26], GWO [24], and ALO [28].

Table 5. Parameters for metaheuristic algorithms.

Algorithms	Parameters
GWO	$\overrightarrow{a}$ ranges from 2 to 0 and $r_1$, $r_2$ are random number generation [0, 1] for alpha, beta, and gamma value and position update of search agents.
SSA	Alarm value (R2 ∈ [0, 1]) and Safety Threshold (ST ∈ [0.5, 1.0])
WOA	$\overrightarrow{a}$ ranges from 2 to 0, $r_1$, $r_2$ are random number generation [0,1], a probability of choosing between shrinking encircling or the spiral model = 0.5 for position update of whales
MPA	Fish Aggregating Devices (FADs) = 0.2, low-velocity ratio (Lévy strategy) = 0.1, unit velocity ratio (Brownian strategy) = 1, high-velocity ratio (Brownian or Lévy strategy) ≥ 10
PSO	Inertial range = [0.1 1.1], swarm size = 100, self-adjustment weight = 1.49, social weight = 1.49
ALO	For sliding ants towards antlion, weightage (w = 2 (for t > 0.1T), 3 (for t > 0.5T), 4 (for t > 0.75T), 5 (for t > 0.9T), 6 (for t > 0.95T) for current iterations t in maximum iterations T. Stochastic function for random walk r(t) = 0 for r (t > 0.5) = 1 and r (t ≤ 0.5) = 0, and t is random number range [0 1]

shows algorithm-specific parameters and assumptions.

4. Experimental Setup and Evaluation

In this section, we evaluate the RAN’s performance in two user distribution scenarios: urban and suburban. We also compare 5G deployment using existing 4G infrastructure and without it.

4.1. Experimental Setup

We used MATLAB 2024a and the MATLAB toolbox to implement the defined algorithms running on a PC having 12th Gen Intel(R) Core(TM) i7-1270P 2.20 GHz, 16.0 GB RAM, 64-bit operating system, x64-based processor, and Windows 11 Pro. We propose to conduct the experiments on two scenarios.

Scenario I: We select an urban geographic area of 2 km × 2 km (4 km²) to serve 10,000 UEs for simulation. In an urban area, we optimize the micro-RRU and macro-RRU locations for both the cases of 5G greenfield and 5G SA using existing 4G infrastructure. Figure 3 and Figure 4 show both use cases for Scenario I.

Scenario II: We choose an urban area of 4 km × 4 km (16 km²) with a population density of about 5000 per km². The optimization is carried out for both 5G greenfield and 5G with existing 4G infrastructure. The performance evaluation is carried out for both cases in this scenario. Figure 5 and Figure 6 show both cases of the initial deployment scenario with population distribution for Scenario II.

Scenario I (Urban initial deployment scenario)

Figure 3. Case I: Urban 5G greenfield.

Figure 3. Case I: Urban 5G greenfield.

Figure 4. Case II: Urban 5G with existing 4G.

Figure 4. Case II: Urban 5G with existing 4G.

Scenario II, (Suburban initial deployment scenario)

Figure 5. Case I: Suburban 5G greenfield.

Figure 5. Case I: Suburban 5G greenfield.

Figure 6. Case II: Suburban 5G with existing 4G.

Figure 6. Case II: Suburban 5G with existing 4G.

After coverage dimensioning, we calculated the total throughput of RRU_m and RRU_M using Equation (16) and using parameter coding rate (948/1024), a bandwidth of 100 MHz for RRU_M, and using the parameter J = 1, $v_{Layers}^{\left(j\right)}$ = 1, $Q_m^{\left(j\right)}$ = 6 (64QAM), $f^{\left(j\right)}$ =1, $\mu$ = 2 (60 KHz SCS), ${OH}^j$ = 0.14 and $R_{max}$ = 948/1024 and with spectral efficiency of 12.63 bits/s/Hz, the total throughput per sector is 1246.73 Mbps. Similarly, for RRU_m, taking bandwidth of 200 MHz and SCS = 120 KHz and with a spectral efficiency of 12.467 bits/s/Hz, the total throughput per sector is 2.493 Gbps.

After coverage dimensioning, we chose the appropriate values for the radius of RRUs and the number of RRUs for simulation. Table 4 shows the link budget outcomes and the number of RRUs required after coverage and capacity dimensioning. The minimum number of RRU_m for the urban area after coverage dimensioning is maximum (uplink, downlink), i.e., maximum (30, 59) is 59. A maximum of the two values is chosen since 59 RRU_m meet the requirement of coverage for both uplink and download and 30 RRU_m does so for uplink only. Similarly, for the minimum number of RRU_M after coverage dimensioning is 6 i.e., maximum (3, 6). The number of RRUs after coverage dimensioning satisfies the coverage constraints but does not satisfy capacity constraints. So the minimum number of RRU_m and RRU_M for the urban area after capacity dimensioning is 139 and 70, respectively. Using 139 RRU_M is a waste of resources, and we chose 70 initial numbers of RRU_m in an urban area (Scenario I). For Scenario II, i.e., suburban area, the minimum number of RRU_m and RRU_M required to meet coverage constraints are 236 and 22, respectively. To meet capacity constraints, the requirements of RRU_m and RRU_M are 18 and 35, respectively. To meet both constraints, we chose the number of RRU_M, which is 35 because choosing RRU_m in the suburban area requires 236, which is a waste of resources in such a population distribution.

4.2. Area Selection for Case Study

For the case study, the urban region of Kathmandu metropolitan city, Nepal, around New Baneshwor, Old Baneshwor, and Gaushala region bounded by the longitude, latitude points (${85}^{\circ }{19}^{\prime }{42}^{″}$ E, ${27}^{\circ }{41}^{\prime }{24}^{″}$ N), (${85}^{\circ }{19}^{\prime }{42}^{″}$ E, ${27}^{\circ }{42}^{\prime }{29}^{″}$ N), (${85}^{\circ }{20}^{\prime }{55}^{″}$ E, ${27}^{\circ }{41}^{\prime }{24}^{″}$ N) and (${85}^{\circ }{20}^{\prime }{55}^{″}$ E, ${27}^{\circ }{42}^{\prime }{29}^{″}$ N) is selected. The total area is 2 km × 2 km (4 km²). For Scenario II, the suburban region around Tarakeshwor and Tokha municipalities, Nepal, has an area of 4 km × 4 km (16 km²), is selected for the case study. The area is bounded by the longitude latitude points (${85}^{\circ }{17}^{\prime }{33}^{″}$ E, ${27}^{\circ }{44}^{\prime }{52}^{″}$ N), (${85}^{\circ }{17}^{\prime }{33}^{″}$ E, ${27}^{\circ }{47}^{\prime }{02}^{″}$ N), (${85}^{\circ }{20}^{\prime }{55}^{″}$ E, ${27}^{\circ }{44}^{\prime }{52}^{″}$ N), and (${85}^{\circ }{20}^{\prime }{55}^{″}$ E, ${27}^{\circ }{47}^{\prime }{02}^{″}$ N). In both scenarios, the performance evaluation of both cases i.e., Case I (5G greenfield) and Case II (5G using 4G infrastructure) is carried out and analyzed.

4.3. Data Visualization and Collection

The data required for simulation, such as the location of existing 4G sites and cost price based on the global tender in Nepal, are collected from real data from one of Nepal’s telecommunication service providers. The assumptions are made for simulation suitable for the real case study field.

5. Results and Analysis

We present the result of optimized locations of RRUs for both urban (Scenario I) and suburban (Scenario II).

5.1. Urban (Scenario I)

This subsection presents the simulation and case study implementation outcome of 5G greenfield and 5G deployment using existing 4G infrastructure.

5.1.1. Urban 5G Greenfield (Scenario I, Case I)

To fulfill the coverage and capacity constraints, we calculated the estimated RRU_m in the urban area as 70. Figure 3 shows a randomly generated population of 10,000 with the random initial deployment of 70 RRU_m for simulation in 2 km × 2 km, i.e., 4 km² urban area. We implemented the metaheuristic algorithms PSO, GWO, SSA, WOA, ALO, and MPA simultaneously for proper placement of RRU_m using initial parameters as lower bound = 0, number of search agents = 30, and maximum number of iterations = 3000, upper bound = 2 and the number of RRUs = 70. Figure 7 shows the convergence of all algorithms and Figure 8 shows time execution for all algorithms. Figure 9 shows that MPA performed best to obtain the best-optimized locations among all algorithms, and Figure 10 shows the best-optimized locations and their coverage with hexagonal cell coverage.

5.1.2. Urban 5G Brownfield (Scenario I, Case II)

We implemented 5G deployment using existing 4G infrastructure, i.e., we implemented 5G SA RRUs along with all existing 4G BTS sites and optimized new RRUs, keeping the 4G sites fixed. We implemented optimization using metaheuristic algorithms like the 5G greenfield. Due to the fixed 5G RRUs we implemented in existing 4G sites, the optimization needs five more micro-RRUs to obtain the same coverage as the 5G greenfield. Figure 4 shows initial random 5G deployment with population, keeping the existing 4G sites fixed for the urban area (Scenario I, case II). Then we implemented the metaheuristics algorithms simultaneously.

Figure 11 shows convergence, and Figure 12 shows the execution time of all algorithms. Figure 13 and Figure 14 show the optimized locations by MPA and ALO, respectively. Figure 15 shows that MPA performed best and then ALO to obtain the best-optimized locations among all algorithms. Figure 16 shows the best-optimized locations and their coverage with hexagonal cell.

5.1.3. Mixed Cell Structure

However, we achieved a coverage percentage of 95.77% using MPA algorithms, but still, there is no access for 4.23% of users due to coverage holes so far. We have introduced a two-tier RAN implementation using both micro-RRUs and macro-RRUs to have full coverage. We implemented the algorithms for macro-RRUs to obtain full coverage and Figure 17 shows the best coverage obtained so far by MPA and is 99.41%. Figure 18 shows the best-optimized location of macro-RRUs and coverage in hexagonal cell.

The trade-off between coverage-enriched macro-RRUs and capacity-enriched micro-RRUs is optimized by calculating coverage and capacity dimensioning, and 66 micro-RRUs and 8 macro-RRUs are implemented to have full coverage and capacity. Figure 19 shows the optimized deployment of RRUs, and Figure 20 shows the optimized deployment of RRUs in the satellite map of mixed cell deployment for greenfield.

For urban brownfields, we achieved a coverage percentage of 95.09% using MPA algorithms, but still, there is no access for 4.91% of users due to coverage holes. We have introduced a two-tier RAN implementation using both micro-RRUs and macro-RRUs for full coverage as in Case I. We implemented the algorithms for macro-RRUs to obtain full coverage and Figure 17 shows the best coverage obtained so far by MPA and is 99.41%. Figure 18 shows the best-optimized location of macro-RRUs and coverage in hexagonal cells. The trade-off between coverage-enriched macro-RRUs and capacity-enriched micro-RRUs is optimized by calculating coverage and capacity dimensioning, and 71 micro-RRUs and eight macro-RRUs are implemented to have full coverage and capacity. Thus, using the existing infrastructure of 4G, we can save the cost of 31 sites. Figure 21 shows the optimized deployment of RRUs, Figure 22 shows the optimized deployment of RRUs in the satellite map of mixed cell deployment for 5G deployment using existing 4G infrastructure.

5.1.4. Carrier Aggregation

Implementing carrier aggregation between 3.5 GHz and 28 GHz can further reduce the number of sites. For carrier aggregation, the macro-RRUs are randomly moved to appropriate micro-RRUs iteratively without compromising coverage, capacity constraints, or QOS. For Scenario I and Scenario II, two extra macro-RRU sites are needed to have at least a coverage of 98%. A total of 10 macro-RRUs are needed, but the total number of RRUs sites is reduced by eight.

5.2. Suburban (Scenario II)

We optimized the best locations of RRUs using metaheuristic approaches for suburban areas. We designed a radio access network to serve 2500 users in 4 km × 4 km i.e., 16 km² area. The minimum number of macro-RRUs we need, as illustrated in Table 4, is min(max(12, 22), max(30, 60)), i.e., 22. Using 60 micro-RRUs is not feasible in suburban areas because 22 macro-RRUs are sufficient for full coverage. For capacity fulfillment requirements in suburban areas, we need max(max(18, 35), max(9, 18)), i.e., 35, because 18 micro-RRUs are sufficient for capacity constraints but not for coverage constraints. To meet both requirements, the maximum (coverage, capacity), i.e., 35, is selected for simulation for both 5G greenfield and 5G with existing 4G infrastructure for suburban areas.

5.2.1. Suburban 5G Greenfield (Scenario II, Case I)

We simulated randomly distributed macro-RRUs as shown in Figure 5 using metaheuristic algorithms GWO, SSA, WOA, MPA, ALO, and PSO. We assumed initial parameters as lower bound = 0, number of search agents = 30, and maximum number of iterations = 3000, upper bound = 4 and the number of RRUs = 35. Figure 23 shows the convergence rate of all algorithms and Figure 24 shows their respective optimized coverage as 99.44%, 98.96%, 99.88%, 96.6%, 99.56%, and 99.84% thus MPA performing the best. Figure 25 shows the best-optimized placement of RRUs with their hexagonal coverage cell. Figure 26 shows the field implementation of RRUs in the case studied area.

5.2.2. Suburban 5G Brownfield (Scenario II, Case II)

We simulated randomly distributed macro-RRUs with the fixed existing 4G sites, also considered 5G sites. The newly added RRUs were optimized during the simulation, keeping existing 4G sites fixed. Figure 6 shows the initial deployment scenario where 4G sites are fixed, and 5G sites are randomly generated. Figure 27 shows the convergence of all used algorithms and Figure 28 shows the corresponding optimized coverage. MPA performed best compared to all other algorithms in terms of coverage. Figure 29 shows the best-optimized RRUs with Fixed 4G sites and population. Figure 30 shows the field implementation of RRUs in the case studied area.

5.3. Impact of Using Existing 4G Infrastructure

5G is being rolled out massively worldwide, and most of the countries have invested in deploying 4G infrastructure. Nepal rolled out 4G on 1 January 2017 and invested about 142 million dollars to deploy 4G nationwide. Deploying 5G will incur a huge amount of investment again, and there will be a longer time to repay the investment. Using existing 4G infrastructure such as towers, backbone infrastructure (microwave/optical fiber), and civil infrastructure reduces the initial investment substantially. This paper analyzed the impact of existing 4G infrastructure for deployment of 5G networks. We have carried out simulations for two scenarios, i.e., urban and suburban areas, taking two use cases separately for each scenario. We calculated the requirement of RRUs with or without using existing 4G infrastructure as listed in

Table 6. RRU requirements for 5G optimization.

	Urban		Suburban
Components	5G Greenfield	5G with 4G sites	5G Greenfield	5G with 4G sites
Initial requirement (coverage and Capacity Dimensioning)
Number of New RRU required (28 GHz)	70	40	×	×
Number of New RRU required (3.5 GHz)	×	×	35	16
Number of Existing 4G sites	×	30	×	19
Final requirement (After Optimization)
Number of New RRU required (28 GHz)	66	41	×	×
Number of New RRU required (3.5 GHz)	8	8	35	16
Number of Existing 4G sites	×	30	×	19
Total number of new sites required	74	49	35	16
Total length of Civil work for ducting and trenching (m)	14,800	9800	17,500	8000
Number of new Backbone sites required (Microwave/Optical Fiber) sites	74	49	35	16
Total Backhauling Fiber length (m)	16,280	10,780	18,550	8480
Final requirement (After Carrier Aggregation)
Number of New RRU required (28 GHz)	66	41	×	×
Number of Carrier aggregated sites (28 GHz + 3.5 GHz)	10	10	×	×
Total number of new sites required	66	41	35	16
Total length of Civil work for ducting and trenching (m)	13,200	8200	17,500	8000
Number of new Backbone sites required (Microwave/Optical Fiber) sites	66	41	35	16
Total Backhauling Fiber length (m)	14,520	9020	18,550	8480

. We conducted a financial analysis calculating the cost of components such as backhaul optical fiber, civil work for ducting and trenching to lay optical fiber backhaul, and microwaves for backhaul connections and

Table 7. Cost Estimation (All prices are in US dollars).

		5G Greenfield		5G with 4G Sites
Components	Unit Cost	Quantity	Total Cost	Quantity	Total Cost
Urban
Civil work (m)	20	14,800	296,000	9800	196,000
Fiber length (m)	1	16,280	16,280	10,780	10,780
Sub Total			312,280		206,780
Microwave sites	4000	74	296,000	49	196,000
Total			608,280		402,780
Final requirement (After CA)
Civil work (m)	20	13,200	264,000	8200	164,000
Fiber length (m)	1	14,520	14,520	9020	9020
Sub Total			278,520		173,020
Microwave sites	4000	66	264,000	41	164,000
Total			542,520		337,020
Sub-Urban
Civil work (m)	20	17,500	350,000	8000	160,000
Fiber length (m)	1	18,550	18,550	8480	8480
Sub Total			368,550		168,440
Microwave sites	4000	35	140,000	16	64,000
Total			508,550		232,480

shows the cost calculation of the resources. Both backhaul and optical fiber installation can be used as redundancy for backhaul, and reliability can be ensured even if a disaster happens.

Urban Area (Scenario I): We optimized and calculated the required number of RRUs to deploy greenfield 5G networks without using existing infrastructure that need not compromise anything that needs 66 micro-RRUs and eight macro-RRUs to serve 10,000 users in a 2 km × 2 km area. In another simulation, we need 41 macro-RRUs and eight macro-RRUs if we use the existing 4G infrastructure. Table 6 lists the comparative calculated values of requirements of RRUs and other resources. After carrier aggregation, the number of sites can be reduced further, and the initial investment can be reduced. Table 6 and Table 7 illustrate that using Existing 4G infrastructure, about 33.7% of resources and their cost can be saved.

Suburban Area (Scenario II): In suburban areas, using millimeter wave for 5G deployment is not beneficial due to its smaller coverage range, and we used only macro-RRUs for simulation. Simulation results show that to serve 2500 users in a 4 km × 4 km suburban area, the required number of macro-RRUs is 35 if we deploy 5G greenfield networks. However, using the existing 4G infrastructure, there are already 19 4G sites, and the required number of new macro-RRUs sites is 16. Using existing sites, up to 54.2% of resources and their cost can be saved in suburban areas as illustrated in Table 6 and Table 7.

6. Conclusions

We studied network planning and techniques for heterogeneous or mixed cell architecture to deliver the required data rate. We used the appropriate link budget calculation process to deploy a minimum number of RRU_m and RRU_M. We used a metaheuristic approach to perform simulations using GWO, SSA, WOA, MPA, PSO, and ALO for comparative studies. MPA performed in the best-optimized location, as results show. Two 5G frequencies, FR1 (3.5 GHz) and FR1 (28 GHz) were evaluated for the link budget. Combining the frequencies gave better results for capacity and coverage, and complements each other in urban areas. In suburban areas, using millimeter-wave was not cost-effective, and only macro-RRUs were used to maintain the required average data rate. Using both frequencies, the average data rate was achieved 50 Mbps at any time, with excellent coverage of at least 98%. For 5G deployment, using proper optimization methods and existing 4G infrastructure, such as backhaul and civil infrastructure, the resources can be minimized up to 33.7% and 54.2% in urban and suburban areas, respectively. Aggregating two carrier frequencies 3.5 GHz and 28 GHz in urban areas, resources can be minimized further up to 41.42%. The deployment solution is suited for urban areas and outdoor scenarios. Further research will focus on modeling actual fading, indoor scenarios, and rural and hard-to-reach areas without adequate fiber infrastructure. Artificial intelligence and machine learning techniques will be suitable for additional research while designing telecommunication infrastructure in diverse territories.