Keywords

1 Introduction

The facilities provided by the wireless telecommunication become an integral part of our day-to-day life. In mobile communication, bandwidth distribution amongst the cells is always a crucial problem. The challenge is to propose a scheme for efficient distribution of the most costly resource bandwidth among the cells without disturbing the quality of service. Classically this problem is termed as Call Admission Control (CAC) in cellular network. Inefficient distribution of bandwidth may create congestion, though there are plenty of resources available in other cells at that time. It needs a dynamic distribution scheme that can access the density of nodes in each cell and will be able to change the distribution scheme accordingly.

A new call can be admitted in a cell depending on the available bandwidth at that time instant. Either the call would be allowed or blocked due to the lack of resources. The situation for ongoing call is a little bit different [4]. A node (with running call) may move from one cell to another (handover) and due to the lack of resources the call may be dropped. Generally an ongoing call should get more priority than a new call. The performance of a call admission control scheme depends on the call blocking and call dropping probabilities. Hence the effective distribution of bandwidth may allow more calls, minimizing the number of call dropped and call blocked that leads to a good call admission control scheme.

There are several call admission control mechanisms for bandwidth allocation to users in fair way and to manage the incoming calls in an efficient manner. Several mathematical and Statistical procedures are also applied to handle the situation and to ensure comparatively better allocation [2]. But the problem is of more severe kind when there is sudden influx in number of users within a confined area. That results difficulty to manage the sudden increase in traffic load within the network. Moreover it leads to technical adversity like insufficient bandwidth and high network interference. Thus it remains a challenge to the CAC solution provider to propose a scheme that will be able to perform better than the existing schemes in this kind of unpredictable situation. The most crucial intention of the service providers are to allow as many as users and to ensure the services at least for some priority users in this typical situation. In this paper, an attempt is made to achieve the above mentioned objectives.

2 Related Work

In this section, a brief outline of the dominant existing approaches for call admission control is presented. A few of existing CAC schemes are briefly discussed here to clarify the scenario and the necessity of proposing a new CAC.

First come first serve is the most common and simple solution that can be proposed to solve the problem of call admission control. In this approach, [5] the total bandwidth is divided into segments and each segment can serve only a particular type of call. The main disadvantage here is the wastage of bandwidth due to the static boundary concept. In order to overcome this difficulty the idea of movable boundary scheme [3] is introduced where the segment widths are adjustable. But it can not provide priorities to new call and handoff call separately which is a very important objective in proposing CAC. This could be improved using the concept of guard channel where the new and handoff calls are treated differently. Using the concept of dynamic partitioning scheme [7], some channels are reserved exclusively for voice calls and data calls and the remaining channels are shared for both voice and data calls. A new voice call or handoff voice call is accommodated either in the reserved channels or in shared channels depending on the availability. The wastage of resources in this scheme is obvious as there may be a situation when the bandwidth would be available for data call but not for voice call. In order to overcome this disadvantage of dynamic partitioning scheme the Dual Threshold Bandwidth Reservation (DTBR) [6] was introduced. It has higher network utilization and provides better quality of service. In DTBR, the channels of the cell are divided into regions by thresholds. The system will drop handoff voice call if the channels are unavailable. The concept of reserve channel strategy along with DTBR is proposed to overcome this problem [8]. The combination of DTBR and reserve channel CAC scheme offer much better channel efficiency and lower call-blocking and call-dropping probabilities. Markov chain [8] is used to calculate call dropping and call blocking probabilities in R-DTBR. The quality of services in case of CAC schemes is often measured by the handoff call dropping probability and that is better in [8].

A model has been discussed in [9] to predict the amount of resource demands directly in real time, based on time series analysis. This scheme is working fine in general situations but may not be efficient to handle a sudden influx.

The major objective of a call admission control policy is to decide whether to take a particular call request or not. In order to take such a decision at a particular time, some factors are considered that are helpful to maximize network utilization and minimize rejection ratio. Incorporating the concept of priority may be helpful in this context. Assigning a priority to the calls will guarantee lower call rejection ratio for high priority user. Call requests are classified into categories based on customer type [1]. Here it is assumed that arrival time, call duration and required bandwidth are different for each customer type. Required bandwidth is predicted using statistical method. Let us consider there are M categories. A numbers of thresholds (Ti) are being fixed for each of the categories (\(\mathrm{{C}}_{\mathrm{{i}}}\)). When a call request from Category i (\(\mathrm{{C}}_{\mathrm{{i}}}\)) arrives at the base station and the required bandwidth of the call request is less than the remaining bandwidth of the system and as well as of \(\mathrm{{T}}_{\mathrm{{i}}}\), then only the call request is accepted. Further, if the required bandwidth of the call request is less than the remaining bandwidth of the system but more than T\(_{i}\), then the call request is admitted to the buffer if there is no call request from the \(\mathrm{{C}}_{\mathrm{{i}}}\) already. Otherwise the request is rejected. The solution is not satisfactory for the particular confined area network as any call request cannot be accepted until a channel allotted for the particular category of calls is available. Thus, though there may exist a free channel but cannot be assigned to any call unless the call type is matching with that of the free channel.

3 Proposed Solution

In this section, a new CAC scheme is proposed based on statistical prediction methodologies to ensure essential and less interrupted services to the prioritized users in this typical situation. The handoff calls are not so much concerned in this case as the concentration is mainly on a confined area with a sudden increase in user’s concentration. The objective is to decrease the call blocking probability and to maintain the service as much as possible for the priority users. Users are accommodated based on their classified privilege categories. Users are classified into categories based on some proposed parameters. The high user concentration in a confined area can be handled by allocating proper weight to each of the concerned parameters involved in categorizing. Weights and corresponding levels of parameters are determined by choosing an unbiased sample from the total users. Parameters considered are Duration of communication, Frequency of each communication type, Average cost per minute, Rental paid by the user. Based on the sampled data, each parameter is divided into some levels and each level is assigned with a score. Now a combined score is calculated depending on the score of the proposed parameters for deciding the category of the concerned user. The parameters are selected carefully to reflect the user profile as these are found to have direct relation with loss or profit of a service provider. The user producing higher profit for service provider gains higher priority.

3.1 Duration of Communication

Duration of communication reflects how much amount of time a user make calls. In general, it is simple that higher the duration higher the profit of service provider. Thus duration of call is the most important metric to asses the profile of the user from the view of a service provider. Data on duration of communications of a certain user for past few months is scrutinized and based on that average call duration of a communication is predicted. A few methods are well-known for predicting values. By taking only the seasonal fluctuation under consideration a method of weighted moving average is used. In this method all previous available data is taken into account according to their relevancy that is older the data lesser the contribution in calculation. For example, the Service provider is considered to have 1000 users. User records for previous three months are available. Among which a random sample of 100 is drawn to analyze and categorize users. For each of the 100 user average duration of communication is calculated by Weighted moving average method.

3.2 Frequency of Each Communication Type

In general a particular cell phone can offer limited number of services. Here, higher the data size of the communication, higher the weights in measuring user profile. The types of requests accepted or forwarded by a service provider vary with the type of communication of the user wants as well as the type of plan the user have selected. Initially, types of requests considered are namely Voice call, SMS, Video data transferring, Internet surfing. The score of a user is calculated based on the used amenities in the chosen sample. The transaction type for all users in the sample can be found out and then the scores are given based on the most likely call-type to occur.

3.3 Average Cost Per Minute

The bill amount of the user not always reflects the actual importance of the user. Higher amount of bill does not always imply that the usage is high. It is a common case that a particular user has chosen a plan that is enough cost cutting even for higher amount of data consumption and results no or little profit for service provider. Thus it is crucial to grade a particular user based only on previous bill paid. To maintain uniformity in some means the average call cost is calculated. Thus cost per minute is a measurable amount that gives the average cost per communication. Let us declare variable Costpermin. It is measured by:

$$\begin{aligned} \mathbf{Costpermin } = \mathbf{(avg\ bill\ per\ week)/(avg\ call\ duration\ per\ week). } \end{aligned}$$

3.4 Choice of Rental

The idea of rental is introduced to provide user better value for money. The actual usage pattern is not reflected independently by Average cost per minute. Another parameter considered here is the choice of rental. It is evident that a customer paying lesser rental gives more profit in general from service provider’s perspective. Thus both the components Average cost per minute and choice of rental helps to reflect a user profile correctly.

3.5 Choice of Weights and Final Score Calculation

The weights given to all the four factors considered previously are very important to formulate. The weights are decided According to the importance of the factors in assessing a user profile. Duration of Communication is very important in the context of confined area network with higher user concentration as lower the call duration lower the congestion and higher the call request accepted in a stipulated time. But from service provider’s perspective, it is better to have long duration call costing higher profit. Though in some way shorter call duration will make the situation better, it can’t be ensured that premium users always request for short calls. Thus considering average cost per minute, the privilege users are found (giving priority to the higher call cost). These two factors alone can not reflect the actual scenario. Thus two new parameter introduced but these have lesser important. Often short duration communication may transmit bulk data or the huge rental facilitates the user to send bulk data in nominal cost per minute. Communication type and rental chosen by the user are considered as additional parameters. Duration of communication and Average cost per minute should have same weights. Frequency of each communication type has the next higher weight, whereas the rental paid by the user plays the least significant role in evaluating user’s profile. So the total score of a customer will be \(\mathrm{{W}} = \mathbf{g}_{1}\mathbf{^ *w}_{1}\mathbf{+g}_{2}\mathbf{^*w}_{2}\mathbf{+g}_{3}\mathbf{^*w}_{3}\mathbf{+ g}_{4}\mathbf{^*w}_{4}, \) where \(\mathrm{{i}}{\mathrm{{th}}}\) parameter has the score \(\mathrm{{g}}_{\mathrm{{i}}}\) and a pre determined weight for \(\mathrm{{i}}^{\mathrm{{th}}}\) parameter be \(\mathrm{{w}}_{\mathrm{{i}}}\). Now, whenever a burst of requests are coming simultaneously to the base station, it should content the user ID and the corresponding score of that user. The Base station takes decision by taking into account the score of the user and the current network (channel) occupancy.

Fig. 1
figure 1

Call handling by proposed scheme

Full size image

Let us consider that the score for \(\mathrm{{j}}{\mathrm{{th}}}\) user be \(\mathrm{{w}}_{\mathrm{{j}}}\) .There exists a predefined threshold value say T (Threshold). The value of T at a specified point of time is dependent on the current network utilization. A request will be granted if the score of that user is greater than the threshold value of the network. Otherwise the call will be blocked. The threshold may be at a very low value at earlier point of time. Then all low privilege (low score) user requests can be granted. But with the increasing of load within a network, T is increasing. As a result only the higher privilege user requests are granted. The handling of call request is depicted in Fig. 1. The data for scores can be updated periodically (period may be a month). Each time the database updated the analysis is done on most recent data.

4 Conclusion

Primary objective of proposing a call admission control scheme was to propose a scheme to provide cellular services at high user concentrated confined area network. The confined area network will be any one of the train, bus, stadium, cinema hall and theatre etc, where a huge gathering of people is often present. It is tough for a network service provider to handle a certain influx in confined area network when a huge number of present users are simultaneously requests for communication. It is discussed so far that there is no such existing solution to the particular problem that can minimize the number of blocked users, provide satisfaction to privilege users as well as maintaining a satisfactory level of the quality of service.

Fig. 2
figure 2

Diagram showing comparisons between two schemes when total number of \(\mathrm{{channels}}=40\)

Full size image
Fig. 3
figure 3

Diagram showing comparisons between two schemes when total number of \(\mathrm{{channels}}=80\)

Full size image

In this proposed scheme, it is tried to ensure the minimum services at least to the privilege users. After doing all the data analysis work the final score of each of the 1000 users is known. Combining scores for all the parameters the final score for each of the hundred users is calculated. Further, the scheme discussed in [1] and the proposed schemes are implemented using C. There are some assumptions of the implementation such as, call requests are generated randomly, call duration of the call are also random, the channel width is fixed for both the schemes. The simulation is done for a time period of 1000 s. In the proposed scheme, privilege users get better service than that in [1]. Call request time is within 1000 s and call duration assumed to be not more than 2000 s. The number of user categories in [1] is considered to be 10. Each category has 5 dedicated channels. User Score in case of the proposed solution is assumed to be any integer value from 0 to 7. In both the cases total number of channels are assumed to be 40 and 80. The numbers of requests at a specified time interval for varying channel number are as follows: For \(\mathrm{{channels}}=40\), total number of call requests served by scheme [1] is 175 and that of proposed scheme is 221 (Fig. 2). For \(\mathrm{{channels}}=80\) total number of call requests served by scheme [1] is 282 and that of proposed scheme is 360 (Fig. 3).

Experimental results have shown that proposed scheme can cater more number of requests also. A comparative study is being done between the existing solution [1] and the proposed work. This CAC scheme performs better compared to the existing solutions for this specific environment as it ensures better service to privileged users and that is really an attractive scheme from the service provider point of view.