Distance‐based cell range extension and almost blank sub‐frame for load balancing and interference mitigation in 5G and beyond heterogeneous networks

The goal of deploying LTE‐Advanced heterogeneous networks is to meet the growing need for Quality of Service (QoS), high data rates, and more coverage. Load balancing is one of the biggest problems, especially when User Equipment (UE) uses received signal strength to connect to different transmission power network tiers. Spectrum for small cells has not been used to its full potential. While all the cells are using the same frequency spectrum and the small cells overlap the Macro Cells, preventing interference is another big problem. The proposed Cell Range Extension (CRE) and Almost Blank Sub‐frame (ABS) provides a solution to overcome these challenges. The proposed ABS in this paper is novel because it is a partial muting one rather than the full muting ABS in recent research studies. In addition, the cell association method is by distance rather than conventional Reference Signal Received Power (RSRP). CRE‐ABS has been applied simultaneously to solve both load balancing and interference. Simulation results show that the proposed scheme achieves substantial throughput and Jain's Fairness Index (JFI).


F I G U R E 1
Interference in a heterogeneous network.
with the operation of other small cells in the network.As a result, load imbalance can occur.This issue has been identified in previous research. 3oad imbalance in a heterogeneous network refers to an uneven distribution of traffic load among the different types of cells in the network.Heterogeneous networks are characterized by a combination of different types of cells, including Macro Cells and other small cells, including relay nodes, which have different coverage areas, capacities, and capabilities.Load imbalance occurs when some cells in the network are overloaded with traffic, while others are under utilized.This can result in poor network performance, reduced capacity, and lower Quality of Service (QoS) for users.For example, if a Macro Cell is overloaded with traffic, users in the cell may experience slow data rates, dropped calls, and poor voice quality.On the other hand, if a small cell is under utilized, the network resources are not efficiently utilized, and the operator may incur unnecessary costs. 4Load balancing in HetNets can pose several challenges, including: 1. Dynamic traffic patterns: Traffic patterns in heterogeneous networks are highly dynamic, and traffic loads can vary significantly over time and across different cells.This makes it challenging to predict and balance the traffic load across the network, as the load balancing algorithms need to adapt quickly to changes in traffic patterns.2. Uneven cell deployment: In heterogeneous networks, cells are deployed in a non-uniform manner, with some cells having higher capacity than others.This makes it difficult to balance the traffic load across the network, as some cells may be under utilized while others may be overloaded.3. Complexity: Heterogeneous networks are complex, and load balancing in such networks requires sophisticated algorithms and techniques.This complexity can make designing, deploying, and optimizing load balancing algorithms in heterogeneous networks challenging.4. Quality of Service (QoS): Load balancing in HetNets needs to take into account QoS requirements for different applications and services.Different applications and services have different QoS requirements, and load balancing algorithms need to ensure that these requirements are met while balancing the traffic load. 5. User mobility: In a heterogeneous network, users may move across different types of cells, and the traffic load may shift from one cell to another.This can result in a load imbalance, as the network needs to dynamically balance the traffic load across the cells.6. Interference: Interference between different cells in the network can impact the traffic load and make it challenging to balance the load.Several factors, including adjacent channel interference, co-channel interference, and inter-cell interference, can add up to the total interference of the network.
To address these issues and restore network balance, Cell Range Extension (CRE), also known as Cell-Specific Offset (CSO), was developed.CRE involves adding CSO to the RSRP of small cells, allowing for increased coverage area and support for more users.However, interference may still be present in the expanded cell areas due to the high power of Macro Cells. 5Figure 2 depicts the CRE region.
However, Inter-Cell Interference (ICI) can occur due to the presence of co-channels and multi-cells.This interference can have negative impacts on both user and cell throughput, especially for edge users.Figure 1 shows how the primary Macro Cell can cause interference for users of Low-Power Cells (LPCs), with interference being represented by the dotted line.Interference refers to the effect of unwanted signals from other cells on the desired signal received by a UE in a particular cell.In LTE-A HetNets, different types of cells are deployed, including Macro Cells, small cells, and relay nodes, which can result in both intra-cell interference (between cells of the same type) and inter-cell interference (between cells of different types). 6These cells use different frequency bands, transmit powers, and antenna configurations, which can cause interference between cells.Several types of interference that can occur in LTE-A heterogeneous networks: 1. Intra-cell interference: Intra-cell interference occurs when the UE receives multiple signals from the same cell.This can happen when the user is at the cell edge or when there are reflections or other multipath effects.2. Inter-cell interference: Inter-cell interference occurs when the UE receives unwanted signals from other cells.This can happen when the UE is close to the boundary between two cells, and the signals from the two cells interfere with each other.3. Co-channel interference: Co-channel interference occurs when cells that use the same frequency band interfere with each other.This can happen when the cells are close to each other, and the signals from the cells overlap.4. Adjacent channel interference: Such kind of interference occurs when cells that use adjacent frequency bands interfere with each other.This can happen when the signals from one cell spill over into adjacent frequency bands used by other cells.
Interference in LTE-A heterogeneous networks can degrade network performance and reduce the QoS for users.Several techniques can be used to mitigate interference, including interference coordination, interference cancellation, interference avoidance, and interference suppression.These techniques enable the network to optimize available resources, reduce interference, and provide a high-quality user experience. 7,8he Time-Domain Inter-Cell Interference Coordination (TD-ICIC) techniques were initially introduced by the 3GPP, in version 10.This approach aims to minimize the interference caused by the use of multiple channels and decrease the power transmitted by the Macro Cell in specific sub-frames, allowing small cells to utilize those sub-frames more effectively. 9The TD-ICIC technique involves lowering the transmission power of certain sub-frames in the Macro Cell's resource blocks to a minimum in the Almost Blank Sub-frames (ABS) of the Common Resource Elements.This is done to minimize interference to UE in neighboring cells, thereby enhancing the performance of cell edge users. 10,11The ABS sub-frames are designed to mute the Macro Cell in specific sub-frames, prioritizing other cells to use protected sub-frames and reducing interference generated by the Macro Cell to neighboring cells. 12Consequently, during ABS sub-frames, certain users may be associated with the protected sub-frames of other cells instead of the Macro Cell, improving the cell edge users' throughput until the Macro Cell remains silent or has reduced power. 13Typically, ABS sub-frames are reserved for common signals such as control, page, and broadcast signals in the Macro Cells. 12By utilizing ABS sub-frames in the small cells and Macro Cells as well, the proposed approach aims to address the issue of inter-cell interference in diverse networks and simultaneously solve load balancing and interference challenges by employing CRE-ABS together in an innovative manner.Overall, this technique has the potential to enhance the performance and efficiency of cellular networks significantly.As a result, this paper proposes an innovative plan to solve the load balancing and interference simultaneously by employing CRE-ABS together.
There are several ways in which the proposed work in this paper differs from the works of other authors, making it unique: 1.The work solves load balancing and interference mitigation simultaneously.2. Distance-based cell association is applied rather than conventional RSRP-based or SINR-based cell association for the CRE scheme.3. The Almost blank sub-frame (ABS) algorithm proposed is a sub-muted one rather than a full-muted one, which is most often discussed in related works.
The paper is structured as follows: Section 2 covers related works, Section 3 presents the system model, Section 4 describes the proposed method, Section 5 presents the results and discussion, and Section 6 offers a conclusion.

RELATED WORKS
Researchers are interested in exploring the potential benefits of utilizing Cell Range Expansion (CRE) to enhance the performance of heterogeneous networks.Several studies, such as References 13 and 14, have proposed modifying the CRE area according to the Signal-to-Interference-plus-Noise Ratio (SINR), while others have suggested expanding the region based on the analysis and updates of the Coordinated Scheduler Operation (CSO), as presented in Reference 14.The aforementioned studies, such as Reference 13 and 14, are representative of this area of research.In order to mitigate the interference arising from multiple cells in heterogeneous networks, 3GPP introduced the concept of TD-ICIC utilizing Almost Blank Sub-frames (ABS), which is detailed in Reference 9. ABS is a technique that prevents Macro Cells from transmitting during specific sub-frames reserved for cell-edge users who use the same frequency.A scheduler similar to the one proposed in Reference 15 is necessary to manage and coordinate these sub-frames effectively.The researchers used a decentralized resource management system that employs dynamic fractional frequency reuse (FFR) to coordinate inter-cell interference caused by users and prioritize edge users.The implementation of ABS, as demonstrated in Reference 16, is expected to improve the throughput of low-power cells, especially at the cell-edge.In general, the researchers aim to investigate CRE and TD-ICIC techniques, such as ABS and FFR, to enhance the performance of heterogeneous networks and address interference challenges in multi-cell environments.
In their study, Reference 17 combined the TD-ICIC scheme with an extended version of CRE, which divided the CRE region into inner and outer parts to optimize the performance of Macro Cells and low-power cells by allocating the entire bandwidth between them.To maintain performance balance throughout the heterogeneous network, the authors introduced almost blank sub-frames and Reduced Power Sub-frames (RPS).This method, known as Further enhanced ICIC (FeICIC), was recommended by 3GPP for release 11 LTE, as stated in Reference 18.The LP-ABS technique reduces the power of certain sub-frames for specific nodes, rather than completely blanking out the sub-frames.This approach has been implemented by Reference 19 and prioritizes user equipment located in the center of Macro Cells.As reported in Reference 20, using this technique can reduce power to 30 dBm, which can lead to improved performance for edge users in Macro Cells.However, the power level value can vary due to different factors, such as the ratio of ABS power to CRE power, the value of the extended region in CRE power, and the channel model.Despite these factors, the researchers consistently used a power level value of 30 dBm, although this may not always yield optimal results, as noted in Reference 21.The authors of Reference 22 introduced an adaptive ICIC approach based on Fractional Frequency Reuse (FFR) to address the impact of power level in LP-ABS and the extended region of CRE on network performance.Furthermore, they recommended a Dynamic FFR (DFFR) technique for small cells to self organize within the network and reduce interference.The objective of this approach was to specifically mitigate interference in the network.
Elsherif and colleagues 23 propose a method to improve the QoS in a heterogeneous network by dynamically adjusting the power provided by ABS (Almost Blank Sub-frame) technology.They specifically apply this method only to multimedia communications in traffic mode.The CRE's coverage area can be adjusted using Cell-Selection Offsets (CSOs), as mentioned in Reference 24, in order to enhance the HetNet performance.The authors of Reference 7 have proposed an innovative concept where both the CRE region and ABS can be dynamically modified based on the traffic conditions of each cell in the network. 2 In addition, researchers have explored various techniques to enhance the user experience in 5G systems, such as interference coordination in massive MIMO (Multiple-Input Multiple-Output) networks during the transition from 4G to 5G. 4 Other strategies, such as coordinated small cell deployment for downlink, coordinated muting, on-demand power boosting, and ICIC (Inter-Cell Interference Coordination) of pilot sequence configuration for cell-edge users, have been recommended by the authors in Reference 3 to mitigate inter-cell interference.
Additionally, other studies have also proposed adaptive control of CRE.For example, Qi and Wang introduced an adaptive bias based on the Resource Block Utilization Ratio (RBUR) in their research. 9RBUR is a metric that measures the load conditions of each cell.The technique involves an initial Cell-Selection Offset (CSO) value and a cutoff value for RBUR.If the RBUR value of a Macro Cell is found to be greater than the cutoff value, the cell is considered overloaded, and the CSO is increased accordingly.This approach aims to improve the performance of the network by adjusting the CRE according to the load conditions of the cells.Other studies, such as References 5 and 6, have also proposed adaptive control of CRE.In Reference 25, an adaptive bias based on RURB was proposed to potentially increase overall capacity, while Reference 11 proposed a separate bias value for each UE instead of a common CSO value for all UEs in a cell, achieved through the Q-learning algorithm.However, this technique requires a significant amount of convergence time.On the other hand, the authors of Reference 12 proposed a solution called Simple Decentralized Adaptive Cell Association (SDACA) that has a fast convergence time and utilizes feedback from individual UEs and broadcast messages from macro eNodeBs.To improve user throughput at the cell edge while maintaining average user throughput, the authors of Reference 26 proposed an adaptive control strategy.However, many research studies fail to include realistic resource scheduling methods that allocate resources to individual users in macro and small cells.Additionally, some studies suggest using an optimal CSO value to maximize total throughput, which can result in high SINR and go against the objective of achieving load balance in CRE.Instead, maintaining equilibrium can offer reasonable average user throughput in the Macro Cell, even if some users operate intentionally at a lower SNR.
In Reference 27, the authors proposed a new approach to updating the CSO value by utilizing Round-Robin (RR) resource scheduling.However, this method only considers fairness among users but not the quality of the radio link for each user.On the other hand, the Proportional Fair (PF) method is preferred as it takes both fairness and individual radio link quality into account and leads to more efficient use of radio resources.Hence, a new adaptive algorithm is proposed in this study, which updates the CSO value using PF resource scheduling.The algorithm aims to achieve an optimal state for the CSO by minimizing the difference in average user throughput of the network based on PF scheduling.
Haroon et al. 28 have suggested the utilization of soft frequency reuse in combination with power control factor to address interference issues in heterogeneous networks.This proposed approach involves transmitting at varying power levels for the internal areas of cells compared to the cell edge areas.The study also takes into account the distribution of SBSs in both uniform and non-uniform patterns around the MBS.By employing stochastic geometry, the researchers analyze the impact of the soft frequency reuse scheme on the proposed model 29 also proposes a new cell deployment technique to enhance user coverage and minimize interference simultaneously.However, studies 28,29 have the limitation of selecting the power control factor precisely, which is critical to minimize the interference.The authors in Reference 30 proposed a technique using Poisson Hole Process to improve the coverage and lower MBS interference.However, the paper considers the uplink rather than the downlink, which is generally the main concern for user satisfaction.Reference 29 also proposes a new cell deployment technique to enhance user coverage and minimize interference simultaneously.
Table 1 provides a qualitative comparison summarizing the different proposed techniques for enhancing the performance of heterogeneous networks mentioned above, along with our own proposed work (CRE-ABS (partial)).The comparison is based on the key focus, methodology, and limitations of each technique.The aforementioned research works indicate that the introduction of CRE (Cell Range Extension) and power rate modifications in ABS (Almost Blank

Load balancing and interference mitigation
Dynamically adjusting bias for CRE and ABS (muting) power Trial and error method to choose the bias and muting ratio Sub-frames) and LP-ABS (Low Power-ABS) can improve the efficiency of cell throughput as well as the user throughput, especially for UEs located at cell-edges.As a result, we suggest implementing an adaptable power ABS system that can be adjusted based on the distance between the base station and the user.This adjustment will lead to changes in SINR values, which will prioritize the edge users in the scheduler.

SYSTEM MODEL
This study focuses on a downlink multi-tier LTE-A network, as depicted in Figure 4 where B is the bandwidth of a sub-channel or physical resource block and is the SINR between BS s and user u.In Equation (2), P s is stated as the power transmitted by base station s, G su is the gain of downlink channel between base station s and user u, P AWGN is the additive white Gaussian noise power.P i is the power transmitted by the interfering cells and G iu is their gain.Since cell association is considered to be carried out on a larger time scale, G su is considered to have been averaged within the association period, and overall physical resource blocks in the whole channel spectrum, that is, frequency-selective fading and fast fading are averaged out.Therefore, G su is constant regardless of the dynamic channel variations within the association period, and the SINR between base station s and user u for each sub-channel is similar.A similar type of model has been considered in Reference 21.

PROPOSED WORK
To achieve cell splitting gain and extend coverage, it is important to establish appropriate cell selection criteria that address the power disparity between the PBS and MBS.Choosing the suitable serving cell for user equipment is crucial for achieving optimal UE performance and system development.Traditionally, the cell selection process was based on the maximum RSRP.However, in scenarios where both MBS and PBS are present, this method is inadequate due to the high transmission power of the MBS, which leads to an imbalanced system load and under utilized PBSs.To address this issue, a cell selection technique based on CRE, as shown in Figure 2, has been adopted to expand the coverage area of the PBS and enhance cell splitting gain.Moreover, solutions to interference issues are necessary to fully utilize the potential of UE performance after cell selection and to further develop the system.The CRE bias is a value that is added to the RSRP of UEs during cell selection, specifically to encourage connections to PBSs with weak received signal strength.The selection criteria for the serving cell is maxRSRP + bias, where bias>0 dB is assigned to PBSs while MBSs are assigned bias=0 dB.This means that the base station with the highest received power in the downlink is not necessarily the one selected.The purpose is to extend the coverage range of PBSs with weak received signal strength.However, this can lead to downlink interference, especially when high bias values are used due to the strong broadcast signals of MBSs.UEs that are offloaded in this scenario may experience low SINR, which worsens as bias values increase, as shown in Figure 2. Therefore, interference mitigation strategies must be implemented prior to implementing CRE to achieve the desired cell splitting gain.
The preceding section emphasized the need to address the interference caused by CRE in the downlink of MBS-PBS deployment.To mitigate this interference, LTE-Advanced uses virtually blank sub-frames called ABS.ABS limits transmission on particular sub-frames of certain cells to reduce interference with other cells.Even though ABS is muted, nearly muted sub-frames transmit Common Reference Signals (CRS), whose direction and amount are based on the type of ABS type utilized.The ABS can be configured for normal operation or "Multimedia Broadcast Multicast Service Single Frequency Network (MBSFN)" operation.MBSFN ABS is used in LTE's implementation of the Multimedia Broadcasting Multicast Service (MBMS).Since MBSFN ABS does not use CRS in the Physical Downlink Shared Data transmission Channel (PDSCH) where data traffic is transmitted, it experiences significantly less interference than regular ABS.To protect UEs scheduled by PBS on the same sub-frames, it is necessary to mute MBS sub-frames 1 and 9, as shown in Figure 3. Algorithm 1 presents the proposed methodology briefly to summarize the whole proposed work.
Consequently, in the present study, the application of CRE with a fixed bias of 15 dB is adopted.The selection of 15 dB bias is determined based on multiple experiments with various bias values.It is found that this particular bias value yields the optimal results for the intended system model.Additionally, the ABS muting ratio is set at 10 dB using a similar trial and error method.Notably, rather than implementing full muting, a partial muting approach is employed for the ABS.In terms of cell association, the distance-based method is utilized, and the distribution of resource blocks is equal among all users linked to each base station.

RESULTS AND DISCUSSION
This study evaluates the effectiveness of time domain muting in the downlink of MBS-PBS with CRE deployment by using a multi-cell system level simulator in Matlab software.The simulation follows the LTE specifications specified in 3GPP, as discussed in the system model section.The experimental scenario involves three PBSs utilizing CRE bias to offload MBS traffic, as shown in Figure 4. UEs are placed in a way that mimics a practical situation where some of the total numbers are randomly distributed in the PBS center, while the others are uniformly distributed between the PBS edges and the MBS.PBSs are generated in each sector of the MBSs.To mimic user mobility, the users are randomly moved to different coordinates in different iterations, and the outcomes are presented as the average of 50 iterations.The simulation is conducted with a constant number of users in a sequence of simulation runs, and the parameters used in simulations are listed in Table 2. Figure 5 presents a comparison of the user throughput achieved by the proposed scheme and other state-of-the-art schemes for 10 randomly selected users, including cell-edge and normal users from both Pico Cells and Macro Cells.The purpose of this comparison is to evaluate the user satisfaction of the proposed approach.The proposed method, which combines CRE with a 15 dB bias and ABS with 10 dB muting, generally outperforms the other schemes, as shown in Figure 5.However, in some cases, the proposed approach did not perform as well due to the random distribution of users and the limited availability of resource blocks.Nevertheless, in most instances, the proposed method demonstrated superior performance.To further validate this claim, the sum throughput of the network was calculated for all schemes and depicted in Figure 6.The sum throughput was evaluated over 50 iterations, with each iteration involving random changes in user positions to simulate different practical scenarios, such as varying path loss and fading.It is evident from Figure 6 that the proposed method outperformed the other schemes in all 50 iterations, providing further evidence that the proposed approach effectively enhances network performance.
The study includes three figures (Figures 7,8, and 9) that compare the performance of the proposed approach to existing methods.Figure 7 depicts a Cumulative Density Function (CDF) graph that shows the network performance of all the methods.The proposed method outperforms the other methods in terms of network performance.Figure 8 presents a comparison of the average throughput for 20 to 100 users, and as the number of users increases, the proposed method continues to outperform the state-of-the-art methods.Upon observation, the trends displayed in all simulated methods are strikingly consistent, implying a significant level of reliability in the new method's simulation.The fluctuations across a number of users are due to random user movement on different iterations, but it has been observed during the experiment that the fluctuation varies on different runs (have different values for different runs, sometimes the peak is at 20, 60 or 80).However, the trend remains almost similar.This is because the user movement is different on all of the runs, and   therefore, the association to the base station also varies, giving different average throughput values.It is different on all of the runs, and therefore, the association to the base station also varies giving different average throughput values.Additionally, Figure 9 demonstrates the effectiveness of the load balancing of the proposed approach using Jain's Fairness Index (JFI). 31The proposed method CRE+bias(15dB)+ABS(10dB) shows better load balancing than the other methods, even as the number of users increases.This shows that a 10 dB muting rather than a full muting of the sub-frame is more effective when we compare them with each other.As we know, the higher the Jain's Fairness index, the better it is for a system, and it also means that the system has a better load balance.The full ABS muting gives better results than the CRE graphs, which represents the importance of implementing ABS to get an optimized load balance.While the partial ABS, referred to as ABS (10dB), gives an even more balanced system; hence, the JFI indexes are higher for it.

CONCLUSION
This study presents a hybrid scheme that utilizes CRE and ABS to address load balancing and interference issues in a multi-cell network.The ABS partially mutes the MBS to mitigate interference, while CRE uses a fixed bias of 15 dB that was selected based on trial and error experiments.Similarly, ABS muting of 10 dB was also determined by trial and error method.Cell association is based on distance rather than conventional methods such as RSRP or SINR.The proposed system model includes three Pico Cells overlaying one Macro Cell, as simulating scenarios with more than three small cells is computationally prohibitive.The simulation outcomes reveal that the proposed strategy, which combines CRE with partial ABS muting, outperforms other existing schemes in terms of user throughput, sum throughput, and average throughput.Additionally, Jain's Fairness Index (JFI) was utilized to assess load balancing, and the proposed approach performed better than other schemes.The proposed methodology also exhibits better performance than the other schemes for varying user numbers between 20 to 100.
It should be noted that there are limitations to this study.One such limitation is that the allocation of users is done randomly in each iteration to simulate user mobility, whereas, in reality, user movement is more deterministic.However, given the small area being simulated and the objectives being evaluated, we believe the random model does not significantly affect the results.It should also be acknowledged that the trial and error method was used to determine the bias for CRE and the muting value for ABS, which may not be the optimal solution.Future work can aim to address this limitation by utilizing optimization algorithms to determine the best values for the parameters.Future work can also include simulating the mobility of the users more precisely as well as multi-objective solutions of load balancing, interference mitigation, bandwidth enhancement, and radio resource management to better serve the end users.Overall, the proposed CRE-ABS scheme with partial muting has shown promising results in improving user throughput, network sum throughput, and load balancing.Further research can build upon these findings to enhance cellular network performance.

F I G U R E 3
Almost blank sub-frame (ABS) in heterogeneous network.

F I G U R E 4
System model.

F I G U R E 5
User throughput with 10 users.F I G U R E 6Comparison of sum throughput of all BSs over 50 iterations.

F
I G U R E 7 CDF evaluation for the proposed scheme compared to other recent schemes.F I G U R E 8 Comparison of average throughput for all the schemes.F I G U R E 9 Evaluation of load balancing by JFI.
Qualitative comparison of techniques for enhancing heterogeneous networks.
TA B L E 1 . The network consists of one Macro Base Station (MBS) and three Pico Base Stations (PBSs) that are randomly located at the edges of the MBS and serve multiple users.The network uses Orthogonal Frequency Domain Multiple Access (OFDMA) with a frequency reuse factor of one.The entire network is modeled according to 3GPP TS 24.526 version 16.5.0Release16,andthespecificationsusedfor the model are outlined in Table2.As mentioned in Table2, there are three Pico Cells within one Macro Cell where each Pico Cell has a radius of 50 m, and the radius of Macro Cell is 500 m.The spectrum considered is the 20 MHz one with 100 sub-channels; hence, each channel bandwidth is 180 KHz, where the remaining is for the control signal as per the 3GPP specification.All other relevant parameters used for simulating the model are described in Table2similarly.S and U represent the sets of base stations and UEs, respectively.The available number of sub-channels in the HetNet is represented by K, and full physical resource block reuse has been permitted in the HetNet.b su is referred to as the association indicator where b Parameters used in the simulation.
su = 1 if UE u associates with BS s, else b su = 0.The throughput or data rate of UE u obtained on a sub-channel by BS s is modeled as Shannon's capacity:R su = Blog(1 + Γ su )(1)TA B L E 2