An AHP-TOPSIS Integrated Model for QoS-Aware Energy Efficiency in Green Cognitive Radio Networks

A novel concept of green network (GN) has started its functions due to the need of better energy efficiency (EE) of cognitive radio systems. This is a new design that involves devices used to improve a network to make it more energy efficient, which will result a reduction in the amount of CO emitted by telecommunications infrastructures participating in the greenhouse effect.

The spectrum in cognitive radio (CR) is divided between primary users (PU) and secondary users (SU) that will optimize its use [1]. It can be said that the CR enables intelligent communication by adapting its transmission parameters according to the variations of the environment in which it operates. this provides good quality of service (QoS) and better energy efficiency.

OFDM (Orthogonal Frequency Division Multiplexing) is defined as an inviting modulation and transmission technique for CR because of its flexibility in spectrum shaping, its good spectrum management and its ability to better analyze the spectral activities [2]. OFDM is widely used in the literature. However, these studies have been limited to optimizing the transmission rate of SUs, limiting the interference introduced in PUs to predefined thresholds [2, 3].

Recently, research has increased on optimizing EE. Wang et al. [4] have improved the EE of the CR system by relying on OFDM exposed to the interference requirements of PUs and the distinct data rates of SUs. an algorithm was formulated by the authors to improve the EE of OFDM-based CR communication systems with sub-channel inconsistencies [5]. This optimization led them to certify a minimum required throughput and a specific power budget for the SU. Thangaraj and Aruna [6] have worked on energy-efficient resource allocation for an OFDM-based heterogeneous cognitive radio network (HCRN) in an imperfect spectrum sensing scenario with excellent QoS.

On the other hand, other research has focused on the optimization of QoS with multi-objective. Saoucha and Benmammar [7] have exploited firefly, bat, and cuckoo search algorithms to optimize the communication quality of the SU. Benmammar et al. [8] suggested applying to cognitive radio a parallel dynamic programming algorithm for multi-objective Pareto optimum maximization. The approach is used to choose the most compromising solution of the Pareto front. which makes the cognitive engine able to achieve perfect feedback results, compared to using metaheuristics and maximizing Qos. The reliability as well as the efficiency of artificial intelligence (AI) techniques is known to optimize the performance of a cognitive radio network (CRN) [9-11].

To allow a wireless multiple access channel Ostovar [12] propose a function to establish the presence of the main user and optimize the energy efficiency of cognitive radio systems for the calculation of the mathematical structure. The average throughput to average power consumption ratio was established to evaluate the network performance under detection constraints. the authors' objective is to minimize energy consumption and maintain detection accuracy by maximizing detection time by managing the constraints of transmission power and throughput.

Wang et al. [13] have treated the problem of average EE maximization for difficult decisions using hybrid spectrum sharing, in contrast a problem could arise, that of the average formulation of the maximum EE generated by the iterative optimization of the detection time and the number of cooperative SUs. This optimization is based on the following elements: the average transmission power of SUs, the peak transmission power, the data rate constraints and the average interference power constraint of PUs.

To manage energy efficiency, Mili et al. [14] have suggested decreasing the power of the secondary transmitter for location and spectrum sharing in fading moments. However, the additional channel state information (CSI) affecting the secondary antenna has also been studied in the context of spectrum sharing through closed form expressions for the relevant transmission power. It should be noted that there was a discussion about imperfect cross CSI containing a licensed band split between these primary and secondary links. Therefore, the distribution of power over the perfect and imperfect measure was minimized.

Khaled et al. [15] have addressed energy and security in a CR-based communication approach. Furthermore, user and network related factors are reconsidered such as the user's battery level and the type of user data as well as the number of unused bands and the level of vulnerability. Given the time complexity of the optimal solution, the authors have also provided a heuristic solution.

In this context, we have already proposed an AI-based solution more precisely on reinforcement learning, which consists in using the Q-learning algorithm which will help cognitive users to find the optimal channel which has low power emission to reduce the energy consumption of these infrastructures and thus satisfy GN [16]. However, this work two limits; on the one hand the complexity of learning algorithms can affect energy consumption. On the other hand, learning poses a scalability problem with a fairly high number of channels.

Our contribution in this paper is to help the SU to select the most appropriate channel for opportunistic use according to the type of application data used during transmission while saving the consumed energy and avoiding interference with PU based on the information received from the spectrum sensing function. To achieve our goal, we proposed an algorithm based on TOPSIS (technique for order preference by similarity to ideal solution) due to its simplicity with moderate complexity/execution time. To optimize TOPSIS and determine the weighting criteria for each objective, analytic hierarchy process (AHP) method was used to assign the weight of each criterion.

The rest of this paper is organized as follows. In part 2, the OFDM-based cognitive radio environment is presented. Part 3 gives a description on our proposed algorithm whose objective is the QoS-aware energy efficiency optimization in OFDM-cognitive radio networks. In part 4, we present the results of the simulation. Finally, section 5 discusses final remarks and suggestions for future work.

TOPSIS was developed by Yoon and Hwang [21], defined as a multi-criteria analysis method for decision support. The multi-criteria decision-making method (MCDM) is the multi-objective optimization technique used to measure alternatives; the primary idea of this method is to elect an action having:

·The smallest distance to the called “ideal” action (positive-ideal solution).

·The greatest distance to the called "anti-ideal" action.

Recent research has proven the effectiveness of the TOPSIS method in many various fields of communication. Loganathan et al. [22] proposed a novel multi-criteria channel decision model is developed by using the Enhanced TOPSIS (E-TOPSIS) which achieves certain desirable characteristics that allow channel decision makers to construct complex decision-making problems. It takes into account several factors of each channel and node, hence the allocation and sharing of channels effectively. In order to access an integrated weighting network in a heterogeneous multi-cognitive wireless network coexistence area, a selection algorithm based on AHP-TOPSIS has been outlined [23].

The different steps of TOPSIS can be expressed as follows [24]:

Step 1: Calculate the normalized decision matrix.

$n_{i j}=\frac{x_{i j}}{\sqrt{\sum_{i=1}^m x_{i j}^2}}$              (7)

Step 2: Calculate the weighted normalized decision matrix.

$v_{i j}=w_i * n_{i j}, j=1 \ldots \ldots m, i=1 \ldots \ldots n$               (8)

where, w_i is the weight of the ith attribute or criterion, and $\sum_{i=1}^n=1$.

Step 3: Determine the positive ideal and negative ideal solution.

$A^{+}=\left\{v_1^{+}, \ldots . v_n^{+}\right\}=\left\{\left(\max _j * v_{i j} \mid i \in I\right),\left(\min _j *\right.\right.\left.\left.v_{i j} \mid i \in I\right),\right\}$               (9)

$A^{-}=\left\{v_1^{-}, \ldots v_n^{-}\right\}=\left\{\left(\max _j * v_{i j} \mid i \in I\right),\left(\min _j *\right.\right.\left.\left.v_{i j} \mid i \in I\right),\right\}$               (10)

Step 4: Calculate the separation measure from the ideal solution.

$d_j^{+}=\sqrt{\left\{\sum_{i=1}^n\left(v_{i j}-v_i^{+}\right)\right\}, j}=1 \ldots \ldots \ldots m$               (11)

$d_j^{-}=\sqrt{\left\{\sum_{i=1}^n\left(v_{i j}-v_i^{-}\right)\right\}}, j=1 \ldots \ldots . . m$               (12)

Step 5: Calculate the relative closeness to the ideal solution.

$R_j=d_j^{-} /\left(d_j^{+}+d_j^{-}\right), j=1, \ldots \ldots m$                (13)

Since $d_j^{-} \geq 0$ and $d_j^{+} \geq 0$, then clearly $R_j \in[0,1]$.

TOPSIS remains dependent on an adequate approach to delineate the weighting of criterion for each type of application data used. However, this concept retains its reliability and certainty in the treatment and evaluation of several attributes and the quantity of alternatives. To assign the weights to the criteria we used the AHP method.

Based on a mathematical concept, AHP is an organized, reliable technique that supports our decisions and deconstructs a complex problem into a multi-level hierarchical constitution of objectives, criteria, sub-criteria and alternatives [25]. AHP can be used to make many complex and unstructured choices containing many attributes [26].

In this section we detail our QoS and EE optimization algorithms based on AHP-TOPSIS method for OFDM-based cognitive radio system.

Our proposal to improve QoS and EE consists first of all of a function (Calculate_weight ()) illustrated by Algorithm 1 and executed offline to optimize the execution time. This function detects the types of data application used and using the AHP method, for each data type and according to its need it makes a comparison between the criteria and calculates the weight of each criterion while trying in all cases to save the transmission power. Therefore, we have for each type of application a vector with the associated weights of each criterion according to its importance.

The second function (Sensing_init ()) presented by Algorithm 2, loads the parameters necessary for the SU to transmit, i.e., the maximum power and the interferences authorized that a SU can transmit (Power_max, Interference_max), the subcarriers number (N), the modulation types and its maximum value. Thereafter, it chooses the vector of the weights used by TOPSIS after having detected the type of the used application; more over it creates a matrix containing the list of subcarriers with all the necessary parameters calculated: the bit rate (Bit_rate_cur), the interference, the error rate present on the subcarriers (Interf_cur, Error_rate_cur) and also the transmit power (Power_cur).

In Algorithm 3 and in line 4, we call the TOPSIS algorithm by passing it as parameters the matrix Q and the vector of the weights, so the algorithm will choose an alternative, among the set of alternatives present in the matrix Q, which has firstly, the shortest distance to the ideal alternative (the best alternative on all criteria), and, which has the greatest distance to the ideal negative alternative (the one which degrades all the criteria) according to the vector of the weights introduced, so that at the end we obtain a vector (V) with values obtained from the evaluation of each subcarrier.

To choose the best solution, the SU will just take the maximum value of the vector V (Algorithm 3, line 5).

Note that we must test before transmitting if the subcarrier is free and if the transmission power does not exceed the maximum threshold and also if the interference generated on the subcarrier does not also exceed the maximum threshold. In the event that one of the conditions previously presented is not respected, we must choose another solution and this time we will choose a value lower than that chosen before. In the case where all the conditions are respected, the SU can transmit on this subcarrier by putting its state in occupied.