1 Introduction

A heterogeneous network is made up of various connected mobile nodes and links of various sorts. These interconnected structures hold a wealth of information that can be utilised to mutually improve nodes and linkages and spread the information from node to the other [1]. The development of heterogeneous cellular networks (Fig. 1) has gained traction in industry and academia, garnering the attention of standardisation groups such as 3GPP LTE and IEEE 802.16j, whose goals include expanding the capacity and coverage of cellular networks [2]. The fast development of tiny cells is causing the cellular network to become random and heterogeneous. The multi-tier heterogeneous network (HetNet) is designed to meet the huge connectivity requirements of growing cellular networks. Cellular networks are typically described by deterministically putting each tier, such as pico, macro and relay nodes (Fig. 2), on a grid that ignores the nodes' spatial unpredictability [3].

Fig. 1
figure 1

Heterogeneous cellular network [4]

Full size image
Fig. 2
figure 2

Types of cells in cellular network

Full size image

1.1 Motivation and challenges

With the constant development of cellular networks to accommodate more and more users and devices, proper cell selection becomes crucial in the provision of optimal network performance, user experience, and resource utilization. However, traditional cell selection methods typically apply either static criteria that seldom coincide with actual performance or heuristic methods that may fail to anticipate the dynamic requirements of the changing user densities, a variety of services provided, and channel conditions that developers of cellular network can face. In such conditions, the modern simulation-based adaptive selection method may be crucial for creating a network that adjusts to the changing conditions of the operating network and provides a dynamic balance for reducing cell load, interference, and processing time and maximizing other KPIs such as the quality of the signal, data transfer speed, and latency.

Moreover, with the expectation to meet the requirements of 5G technologies that will cover the ultra-reliable low-latency communication and massive machine-type communication spectra in addition to the standard mobile broadband services, the need for an adaptive and frequent-simulation-based cell selection method becomes apparent. As the results of such simulation can be utilized to dynamically adjust the cell selection, applying the method will result in the development of prediction and tuning models that can create a much more efficient system of cellular network operation. In this way, the task’s higher motivation lies in the fact that by addressing the current challenges of the environment, it can facilitate the unlocking of the full potential of next-generation cellular network operation.

1.2 Limitation

  • Simulation-based methods require extensive computational resources, limiting their use for real-time or near-real-time cell selection.

  • Large-scale networks due to 5G technologies make simulating these networks virtually impossible due to memory and processing capacity requirements.

  • The accuracy and effectiveness of the model used to represent signal propagation, user locations, mobility, interference, and other network aspects are crucial for successful simulation.

  • Deviations from the right model can lead to erroneous cell choices, negatively impacting the method's use.

  • The simulation cannot respond to critical conditions like natural disasters or unexpected rituals, making cell selection less efficient.

  • The applicability of the simulation-based method is dependent on the characteristics of the contexts, with urban areas showing poor performance in rural areas.

  • The need for expertise in a non-standard field result in additional costs, especially for smaller network operators.

  • The simulation-based method requires significant verification and validation efforts to confirm its realism.

1.3 Novel contribution of research work

The research paper presents an approach that uses simulation to predict the real-time network conditions. The method utilizes machine learning for predictive modeling effective for optimizing key performance indicators, such as signal quality and latency improvements. Furthermore, the article includes the thorough simulation framework that is consistent across different testing environments, minimizing computational expense for real-time implementation and confirmed with real-world data. Overall, the present work and the source it is commenting on, create a basis for the targeted technologies to develop further leading to new approaches and adaptations to such emerging technologies as 6G and networks designed for the internet of things.

As the need for wireless data reaches historic heights, cellular service providers are looking at new ways to meet it. One potential option is the deployment of small cells with lower consumption of power and smaller covering regions. Small cells, which utilise the same spectrum as existing macro cells, allow for greater bandwidth re-usage, potentially leading to higher user rates. The coverage area for small cells is greatly decreased because their signals are weaker and the incremental benefit of placing each small cell is limited [5].

The process through which it will decide which cell(s) will offer its service to which mobile station is known as cell selection. Optimization in this process is a critical step toward making the most of both present and the future of the cellular networks [6]. Cell selection (Fig. 3) refers to the method that allows each mobile station to receive services. It is vital to optimise the procedure in order to maximise the current and future performance of cellular networks [7]. Cell selection is the procedure of deciding which cell will serve each mobile station (Fig. 4). The purpose is to identify a high-profit subgroup of clients who can be completely pleased by the proposed study [8]. Cell selection has a significant impact on the overall system’s effectiveness. The overall performance of the cellular network can be determined by how effectively the network uses Cell Selection.

Fig. 3
figure 3

Cell selection [13]

Full size image
Fig. 4
figure 4

Cell selection process [9]

Full size image

2 Related work

Make use of the mechanism All-or-Nothing Demand Maximization (AoNDM) will determine the assignment of mobile nodes to base stations in a service-oriented cellular network environment, where many base stations are simultaneously satisfied. It is feasible to make the most profit from clients who are completely satisfied. The study’s findings suggest that a theoretical method based on algorithms can manage capacity, demands, and interference, resulting in a better solution for cell planning challenges and greater scalability [8]. This is a key step toward more targeted 4G network planning. The downlink rate distribution under an extended cell-selection model, which specifically differentiates among long-term channel impacts including such shadowing and path-loss, and small-scale impacts such as fading. The authors offered an equivalent explanation of such a generic cell selection model and demonstrated that the impact of shadowing may be explored similarly by correctly scaling transmit powers. Using this comparable interpretation, research of the influence of shadowing on load balancing was conducted, and it was discovered that in some regimes, shadowing automatically balances load across several tiers, reducing the requirement for artificial cell selection bias [10].

A comprehensive examination of a fresh method for cell selection in 4th generation cellular networks. The suggested mechanism, in contrast to the existing cell selection protocol, is worldwide, provides a performance guarantee, and incorporates many of the predicted 4G technologies [6]. The major enabling technology for meeting the high data rate needs of future generations of wireless networks is dense deployment of tiny cells. In this situation, heterogeneous broadband solutions will be employed to connect small cells to the main network, and accessibility and backhaul must be optimised together to make the best use of available resources [11]. The suggested study substitutes a strategy that optimises network ergodic capacity for the traditional SINR-based association criterion. As a result, it modelled the links between cell load, backhaul restrictions, resource allocation, and ergodic capacity analytically [12].

In order to balance the workload among heterogeneous cells and improve resource usage and consumer fulfilment in terms of both data rate and EMF exposure, a novel cell association paradigm for heterogeneous cellular networks (HetNets) is proposed. Two heuristic strategies are demonstrated to solve the cell selection system’s General Assignment Problem (GAP) with a low level of complexity. The study’s findings show that the suggested substitutes significantly outperform legacy association schemes [15]. Below given Table 1 shows the comparison between Traditional Methods and Improved Simulation Based Method with various aspects.

Table 1 Comparison between traditional methods and improved simulation based method
Full size table

Examples of contexts in which the usage of mobile codes is particularly practical include autonomous decentralised systems. For instance, given their capacity to decentralise processing, adapt to system autonomy, allow for flexible management of installed code, and support user interaction, mobile codes appear to be very advantageous when designing highly scalable distributed systems in a large, heterogeneous, and multi-organizational distributed environment [14]. Asynchronous Transmission Mode (ATM) networks have been chosen for usage in Broadband Integrated Service Digital Networks (B-ISDN) because they can accommodate a variety of services, including phone, data, video, etc. a queuing model for ATM networks that takes into account three different forms of traffic, such as voice, data, and video. We examine a discrete time single-server (GI/1/1) queuing system that has three infinitely large priority queues. The waiting time distribution for each class of packets is explicitly derived [15] the idea of mobile ad hoc networking (MANET) and highlights some of its potential applications in the future. The study also discusses two of the technological difficulties that MANET presents, namely Geocaching and QoS [16]. In many ad hoc sensor network scenarios, security is a top concern. A security strategy must include the detection of malicious nodes. The suggested method employs fuzzy logic to detect node attacks and other harmful activity. The proposed study will both identify the network assault and offer a solution to shorten the network execution time. In order to secure Mobile Adhoc Networks, the project's goal is to provide security. The suggested work employs the AODV algorithm [17]. Depending on the kind of session the mobile nodes are having, it employs a home agent to choose the best network. It chooses the best route inside the chosen ideal network using a route selection algorithm. In the Present work, two distinct position-based ant colony routing algorithms are suggested for mobile ad-hoc networks. Before a session begins, each routing algorithm chooses the best possible route [18].

Cell selection, which comprises spotting and decoding the Primary Synchronization Signal (PSS) and the Secondary Synchronization Signal, is the first step in the strategy (SSS). The most important control parameters in the Master Information Block (MIB), such as system bandwidth, are then extracted from the Physical Broadcast Channel (PBCH), enabling the configuration and operation of the other channels in the cell [19]. The other features of the cell's configuration can be discovered by listening to the unencrypted System Information Blocks (SIB), which a passive radio sniffer can intercept. In order to send and receive user traffic at this point, the UE establishes a genuine network connection using a randomised access procedure and a NAS (Non-Access Stratum) process [16].

Factors affecting cell selection: Now, here it is necessary to discuss about the various factors which will affect the Cell Selection. It includes Mobility pattern and Call arrival pattern.

Mobility pattern: Mobility pattern of the user is the most important factor in determining the cell selection. Think of a general condition in which user will change the location and select new cells during the business hours or working hours if the user is in that kind of profession. But if the user is doing a job in which he/she has to work at single place then in that situation the change in location as well as cell selection is comparatively less in numbers. Various models for Mobility patterns are given below:

  1. 1.

    Memory less (Random Walk) Movement Model [20].

  2. 2.

    Markovian Movement Model [21].

  3. 3.

    Shortest Distance Model [22].

  4. 4.

    Gauss Markov Model [23].

  5. 5.

    Activity based model [24].

  6. 6.

    Mobility Trace [25].

  7. 7.

    Fluid-Flow Model [26].

Call arrival pattern: Time is the important factor to describe the Call Arrival Pattern or rate at which the user will receive the calls. User always gets more number of call during working hours as compared to non-working hours. Various models for Call Arrival Pattern are given below:

  1. 1.

    Poisson Model [27]

  2. 2.

    Call Arrival Trace [28]

3 Proposed methodology

3.1 Components of proposed methodology

Various components used in the proposed methodology are shown in Table 2.

Table 2 Various components of proposed methodology
Full size table

3.2 Proposed algorithm

The overall process developed in this work involves the following steps: the gathering of initial network parameters, setting up a simulation framework, collection of real-time data, training ML algorithms, and predicting the short-term network conditions. The dynamic cell selection algorithm runs, assessing the signal quality, load balancing, mobility consideration, and energy efficiency, making decisions and adjusting the cell selection strategies based on that. Monitor of relevant KPIs is conducted in real-time and used as a basis for running the algorithm as well as making adjustments. Proposed algorithm has been shown in Fig. 5. The algorithm is run first in the simulation and then in the real-time, and validation is conducted through comparison of the simulation results with the real mobile network data. The algorithm can be stopped based on the specified termination conditions.

Fig. 5
figure 5

Proposed algorithm

Full size image

Here in this paper a novel Cell Selection method has been proposed in which the data has been collected through simulation model by considering various situations in which various models are covered like (1) Walking mobility, (2) Driving Mobility (3) Mobility with Low Congestion (4) Mobility with Medium Congestion and (5) Mobility with High Congestion. Here the data collected based on simulation carried out through Qualnet® Software and results are used to do the performance analysis. The major points that are required to be taken care here are (1) No. of nodes, (2) Simulation time (3) Call between no. of cell phones.

4 Simulation

In this paper the simulation has been carried out through Qualnet \(\circledR\) Software to test various possibilities of environment in which the Cell Selection can be applicable to the Cellular network.

For simulation purpose the base environment taken is given in Fig. 5.

Below Fig. 6 shows the Base Diagram used for simulation purpose. It contains total 8 cellular nodes, one base station and one cloud network. The same network environment has been used through-out the whole simulation process.

Fig. 6
figure 6

Base diagram

Full size image

4.1 Simulation setup

Following Table 3 shows details about the simulation setup created through Qualnet software. Here in the above setup total 8 cellular nodes are taken for simulation purpose. Each node is connected with one base station and all the base stations are connected with a cloud network.

Table 3 Parameters used for simulation
Full size table

4.2 Simulation scenarios

Various simulation scenarios are created based on the above parameters for simulating the real time situation. Here various conditions are taken into consideration for making the simulation environment to be like real time. As of now simulation carried out for 300 s (i.e. 5 min). It can be extended for more time also. Total 8 users are taken into consideration with different type of mobility patterns i.e. Low, Medium and High.

In the below Fig. 7 it shows a scenario in which there is No Congestion as well as very low mobility among nodes. There is GSM call going on in between node 7 and node 11. The call has been started after 10 s and the call duration is 300 s. Here only two MS are busy i.e. MS-7 and MS-11. MS-7 moves 2898 m in 300 s while MS-11 moves 3019 m in 300 s.

Fig. 7
figure 7

No congestion low mobility

Full size image

In the below Fig. 8 it shows a scenario in which there is Moderate Congestion as well as moderate mobility among nodes. There is GSM call going on in between node 7 to node 11 and node 14 to node 15. The call has been started after 10 s and the call duration is 300 s. Here total four MS are busy i.e. MS-7, MS-11, MS-14 and MS-15. MS-7 moves 5439 m in 300 s. MS-11 moves 5296 m in 300 s. MS-14 moves 4887 m in 300 s. MS-15 moves 5270 m in 300 s.

Fig. 8
figure 8

Moderate congestion moderate mobility

Full size image

In the below Fig. 9 it shows a scenario in which there is High Congestion and high mobility among nodes. There is GSM call going on in between node 7 to node 11, node 14 to node 15 and node 8 to node 12 respectively. The call has been started after 10 s and the call duration is 300 s. Here total six MSs are busy i.e. MS-7, MS-11, MS-14, MS-15, MS-8 and MS-12. MS-7 moves 9071 m in 300 s. MS-11 moves 8778 m in 300 s. MS-14 moves 7713 m in 300 s. MS-15 moves 7753 m in 300 s. MS-8 moves 7375 m in 300 s. MS-12 moves 7213 m in 300 s.

Fig. 9
figure 9

High congestion high mobility

Full size image

In above Figs. 7, 8 and 9 shows various scenarios. Here in these figures green line shows there is a GSM call going on between two nodes and blue line shows there is a communication going on for cell selection/cell re-selection between given node and the base station. Red flags shows the path for the mobile device to move from one place to another place. The area covered during the simulation is 1500 m and the square area is taken for the same purpose.

The lines could also represent handover decisions where a mobile device switches from one base station to another. This process happens when a device moves out of the coverage area of one base station or when the signal quality improves by connecting to a different base station.

5 Performance analysis

Through the above simulation process following performance analysis can be derived and summarised in a below given Tables 4, 5 and 6:

  • Total Attempts and Failures:

    • Total Number of Cell Selection Attempts:

      The total number of attempts across all nodes is 4 + 3 + 0 + 5 + 3 + 0 + 4 + 4 = 23 attempts

    • Total number of cell selection failures:

      The total number of failures across all nodes is 1 + 0 + 0 + 1 + 0 + 0 + 1 + 1 = 4 failures.

    • Total number of cell re-selection attempts:

      The total number of re-selection attempts is 4 + 2 + 0 + 3 + 2 + 0 + 2 + 3 = 16 attempts.

    • Success rates:

      Cell selection success rate:

      The success rate can be calculated as:

      $${\text{Sucess}}\;{\text{rate}} = \frac{{{\text{Total}}\;{\text{attempts}} - {\text{total}}\;{\text{failures}}}}{{{\text{Total}}\;{\text{attempts}}}} \times 100 = \frac{23 - 4}{{23}} \times 100 = 82.$$

      This indicates that around 82.61% of the cell selection attempts were successful.

    • Overall network performance: With a success rate of 82.61%, the network is performing decently well, but improvements can be made to reduce the failure rate.

    • Problematic nodes: Nodes 7, 14, and 15 might be in areas with poor coverage or signal interference, as they show both selection failures and re-selection attempts.

    • Efficient nodes: Nodes 8 and 12 show no failures, indicating that their areas have good coverage and stable connections.

  • Total attempts and failures:

    • Total number of cell selection attempts:

      The total number of attempts across all nodes is 4 + 0 + 0 + 5 + 0 + 0 + 3 + 4 = 16 attempts

    • Total number of cell selection failures:

      The total number of failures across all nodes is 1 + 0 + 0 + 0 + 0 + 0 + 1 + 0 = 2 failures.

    • Total number of cell re-selection attempts:

      The total number of re-selection attempts is 2 + 0 + 0 + 3 + 0 + 0 + 1 + 2 = 8 attempts.

    • Success rates:

      Cell selection success rate:

      The success rate can be calculated as:

      $${\text{Sucess}}\;{\text{rate}} = \frac{{{\text{Total}}\;{\text{attempts}} - {\text{total}}\;{\text{failures}}}}{{{\text{Total}}\;{\text{attempts}}}} \times 100 = \frac{16 - 2}{{16}} \times 100 = 87.$$

      The success rate has improved compared to the previous data (from 82.61 to 87.5%), indicating better network performance or fewer issues related to cell selection.

    • Overall network performance: The improved success rate of 87.5% shows that the network performance is relatively strong, with fewer cell selection issues compared to the previous data.

    • Stable nodes: Nodes 11 and 15 show the highest number of successful attempts without any selection failures, although Node 11 has had the highest re-selection attempts, which could indicate challenges during the session.

    • Areas for investigation: Node 7 and Node 14 are worth further investigation due to their failure rates, which suggest weaker signals or more challenging environmental conditions.

  • Total attempts and failures:

    • Total number of cell selection attempts:

      The total number of attempts across all nodes is 4 + 3 + 0 + 5 + 3 + 0 + 4 + 4 = 23 attempts

    • Total number of cell selection failures:

      The total number of failures across all nodes is 1 + 0 + 0 + 1 + 0 + 0 + 1 + 1 = 4 failures.

    • Total number of cell re-selection attempts:

      The total number of re-selection attempts is 4 + 2 + 0 + 3 + 2 + 0 + 2 + 3 = 16 attempts.

    • Success rates:

      Cell selection success rate:

      The success rate can be calculated as:

      $${\text{Sucess}}\;{\text{rate}} = \frac{{{\text{Total}}\;{\text{attempts}} - {\text{total}}\;{\text{failures}}}}{{{\text{Total}}\;{\text{attempts}}}} \times 100 = \frac{23 - 4}{{23}} \times 100 = 82.$$

      This indicates that around 82.61% of the cell selection attempts were successful.

    • Overall network performance: With a success rate of 82.61%, the network is performing reasonably well, but improvements can be made to reduce the failure rate.

    • Stable nodes: Node 8 and Node 12 show no failures and successful connection attempts, suggesting they are in well-covered or low-interference areas.

    • Problematic nodes: Node 7, Node 14, and Node 15 are areas of concern, as they experienced both failures and multiple re-selection attempts, which could be due to signal issues or frequent handovers.

Table 4 Performance analysis for no congestion and low mobility
Full size table
Table 5 Performance analysis for moderate congestion and moderate mobility
Full size table
Table 6 Performance analysis for high congestion and high mobility
Full size table

Above given are some of the situations used for simulation purpose. At the time of actual simulation total 9 scenarios are considered but here only 3 samples are given for understanding purpose. Total 9 scenarios covers different types of congension and mobility patterns by the users.

The use of an improved simulation-based method for cell selection in cellular networks enables better performance, efficient resource management, and higher scalability. It ensures a cost-effective, adaptable, and future-proof solution that benefits both users and network operators by improving quality of service, reducing energy consumption, and handling complex real-world network environments.

6 Conclusion and future scope

The paper describes a novel simulation-based method for cell selection in cellular networks. This approach helps in improving the performance of the network by constantly adjusting to the real-time conditions dynamically. This technique has a predictive nature integrated which can help in predicting future conditions, thereby making the adjustment beforehand. It helps in enhancing the quality of signals received, minimizing congestion in the network and improving the throughput and latency of KPIs. The simulation of the various kinds of environments by the simulation framework provides a strong basis for the study of the cell selection strategy across various situations. The advantages of this method waste less networks, improved performance, waste less load imbalances, and better user experience. However, the disadvantages include computational complexity, requirement of precise predictive models, and non-scaling peculiarity for extremely dense networks. Future studies can focus on bettering the scalability and computational efficiency of the approach, better machine learning models, and usage of hybrid methods to improve the cell selection for adapting method nature for dynamic conditions.