Multiple CUAV-Enabled mMTC and URLLC Services: Review of Energy Efficiency and Latency Performance

Cognitive unmanned aerial vehicles (CUAVs) play a vital role in next-generation wireless networks as they assist in massive machine-type communication (mMTC) and ultra-reliable low-latency communication (URLLC) services. This study focuses on multiple CUAV-enabled networks wherein CUAVs are paired with each other. We analyze the data rate, energy efficiency, and latency of such networks by applying the finite information block length theory, wherein mMTC and URLLC information use a non-orthogonal multiple access technique. Furthermore, we formulate an optimization problem to maximize the energy efficiency of paired CUAV devices by jointly optimizing the transmission power of the mMTC and URLLC information to satisfy the latency requirement. The numerical results indicate that our proposed multiple-CUAV-enabled scheme enhances the network performance of CUAV devices in terms of energy efficiency and latency better than the existing scheme.

by the 3 rd generation partnership project (3GPP) [5]. Based on 3GPP, the following two connection linkages between a UAV and ground gNB are required. (i) Control link, referred to as a command-and-control (CaC) connection: it controls and regulates UAV flight operation among ground gNBs; and (ii) data link, referred to as an application data connection: it controls and regulates UAV flight operation according to customer requirements between the UAV and ground gNB [5].
Ultra-reliable and low-latency communications (URLLC) systems are needed for CUAVs. URLLC designs must balance low latency and high reliability. Since CUAV needs both high reliability and low latency, it is now important to research wireless networks using finite block length theory (FBLT). However, the traditional infinite block length theory (IBLT) can no longer be used because it does not meet the requirements for latency and reliability. The FBLT scheme is employed in mission-critical applications, such as real-time tracking, rapid transmission, autonomous driving, remote control, and tactile Internet. Contrastingly, the IBLT scheme is employed in sensing tasks, such as remote sensing, remote coverage, surveillance, security, observation, smart monitoring in the agricultural and civil infrastructure sectors, and environmental monitoring [6].
In the past few years, researchers have explored the various applications of UAV networks by applying IBLT and FBLT. Using the FBLT scheme, non-orthogonal multiple access (NOMA)-based UAV-aided communication networks have been proposed [7]. The authors of [8] focused on the control and non-payload links of UAV communication systems. From a UAV-enabled relay network perspective, wherein the UAV serves as a relay, the total decoding error rate minimization [9], joint power allocation and location optimization of the UAV [10], joint transmit power, and block length optimization scheme [11] have been addressed. In addition, UAV-assisted Internet-of-Things (IoT) networks [12] and device-to-device (D2D) networks [13] have been shown to be efficient ways of improving system performance. Moreover, millimeter-wave communication [14] and secure communication [15] in UAV networks have been investigated for lowlatency and high-reliability applications. Spectrum-sharing cognitive radio (CR) network performance has been investigated using the FBLT scheme [16].
The aforementioned studies on UAV communications [7], [8], [9], [10], [11], [12], [13], [14], [15], [16] focused on relay networks, IoT networks, D2D networks, and CR networks where the FBLT was applied. In [7], the authors investigated optimum resource allocation methods for NOMA and relaying systems to maximize UAV throughput while ensuring ground users (GUs) transmission quality in terms of throughput and reliability. In [10], the research studied jointly optimizing the UAV's position and power to decrease the probability of decoding errors while maintaining latency limitations. In [11], the authors proposed an efficient joint blocklength and transmit power optimization technique to optimize the adequate amount of information received by the control center while considering latency and reliability constraints. In [16], the authors optimized the minimal average rate for the secondary UAVassisted IoT network, subject to a probabilistic interference power restriction on the primary network.
Few recent studies have focused on integrating NOMA with UAV systems in a pairing scheme. In [17], the authors investigated the combined NOMA power allocation, user pairing, and UAV deployment for UAV-based wireless systems. For this configuration, the author optimized the user pairing, power distribution, and UAV placement in order to maximize the minimum total rate for each user pair. In [18], the study explored the utilization of a UAV as a pairing user to increase the total capacity through flexible pairing in NOMA. In addition, the performance of flexible pairing was described in terms of total capacity, outage probability, and throughput. In [19], a K-means clustering-based UAV deployment technique was presented to optimize the uplink NOMA service regions. Moreover, a location-based user pairing approach was provided for the multiple UAVs-assisted uplink NOMA. In [20], the study provided an energy-efficient pairing and power allocation approach for UAVs and GUs in a NOMA UAV network. Furthermore, the objective was to decrease the energy consumption of both UAVs and GUs during uplink data transfer and to ensure their needed transmission speeds by optimizing pairing and power.
To the best of the authors' knowledge, there is no recent work on the CUAV-assisted IoT network using the NOMA technique beyond [1]. In addition, the subject of maximizing energy efficiency has not been studied in the available literature [7], [10], [11], [16]. Moreover, the UAV pairing technique in CR environment utilizing the FBLT scheme has not been recently studied [17], [18], [19], [20]. In prior research [1], the CUAV has a massive machine-type communication (mMTC) transceiver for transferring application data, such as smart agriculture, environmental monitoring, and security monitoring. In addition, CUAV is equipped with a URLLC transceiver for transferring CaC data to conduct CUAV flight operations. Herein, one transceiver transmits mMTC and URLLC information by applying the NOMA technique jointly; therefore, this CUAV is hereafter referred to as m/uCUAV. Hence, it is necessary to investigate the m/uCUAV paring scheme using the FBLT because CUAVs provide both mMTC and URLLC services based on customer requirements.
In this study, we investigate a pairing approach for multiple m/uCUAV devices to achieve better performance using the NOMA system. Our contributions to the present study are as follows.
1) We propose multiple m/uCUAV-enabled NOMA technology wherein two m/uCUAV devices in each pair communicate with the gNB at the same bandwidth. 2) Next, we derive the theoretical expressions of data rate, energy efficiency, and propagation latency for a pairing scheme of m/uCUAV devices by considering the FBLT scheme. In addition, we consider the 3GPP-based path-loss model and gNB antenna gain for accurately characterized channels, which is different from our earlier work [1]. 3) Furthermore, we propose an optimization problem to maximize the energy efficiency of paired m/uCUAV devices by jointly optimizing the transmission power of the mMTC and URLLC information while fulfilling the latency requirement. To solve this problem, we divide the problem into two sub-problems with two stages: the pairing scheme and the optimum powers of the mMTC and URLLC information. We obtain the optimum powers of mMTC and URLLC information using the Karush-Kuhn-Tucker (KKT) criteria and Cramer's approach. 4) Finally, we evaluate the performance of the proposed multiple-m/uCUAV-enabled wireless network. Simulation results show that compared to the existing scheme [1], the proposed scheme can significantly increase energy efficiency and reduce latency. In addition, the proposed approach maximizes energy efficiency compared to orthogonal multiple access (OMA), equal power, and channel inversion. The remainder of this paper is organized as follows. Section II introduces the network model, frame structure, spectrum sensing, signal model, and propagation channel model. Section III presents the performance metrics, such as the data rate, energy efficiency, and propagation latency, for the considered network model. Section IV presents the optimization problem formulation for multiple m/uCUAV-enabled wireless networks and subsequently provides a solution to the formulated problem. Section V presents the simulation results and demonstrates the superiority of the proposed solution. Finally, Section VI concludes the study.

A. Network Model
We consider a CUAV-IoT network model for the execution of sensing tasks. One gNB, multiple rotary-wing 1 m/uCUAVs, and one primary transmitter (PT) are deployed, as shown in Fig. 1. First, the m/uCUAVs sense the data/information in the uplink phase; they have a small server that stores the collected information. Subsequently, they transmit the information to the gNB in the downlink phase. This study describes only the downlink phase of the m/uCUAV devices where two types of information are transmitted: URLLC information, such as the 1 A rotary-wing m/uCUAV can hover and fly in any direction, take off and land vertically, and travel slowly enough to be easily controlled and maneuvered [4]. For the sake of simplicity, we assume m/uCUAVs are hovering at different altitudes in the current study. Future considerations include m/uCUAV mobility. CaC operation, and mMTC information, such as the sensing operation. The m/uCUAV and gNB are the secondary transmitter and receiver, respectively. They are each equipped with a single omnidirectional antenna with unitary gain in any direction and a uniform linear array antenna. A set of m/uCUAVs is denoted by K = {i |1, 2, 3, . . . , K }, where K is the total number of m/uCUAVs. All m/uCUAVs were deployed at various altitudes h ui (∀i ∈ K).
As shown in Fig. 2, each m/uCUAV must access one resource block with bandwidth B/K , where B is the available bandwidth for the system. Motivated by [21], [22], we proposed a pair of m/uCUAVs, such as (k−1)-th and k-th m/uCUAVs, owing to the higher bandwidth 2B/K . Therefore, the set of m/uCUAV pairs was denoted by K p = {(k − 1, k )|(1, 2), (3,4), . . . , (K − 1, K )}. Based on the channel gains, gNB 2 determined the pairs in which the k-th m/uCUAV is paired with k−1-th m/uCUAV. Without loss of generality, we considered the channel gains to be arranged as |g k ug | 2 < |g k −1 ug | 2 , where |g k ug | 2 and |g k −1 ug | 2 represent the channel gains from k-th m/uCUAV to gNB and k−1-th m/uCUAV to gNB, respectively. In addition, the corresponding power allocations for URLLC and mMTC information were arranged as:

B. Frame Structure and Spectrum Sensing
Owing to the solution for spectrum scarcity, m/uCUAV utilized the same radio resource blocks to transmit both mMTC and URLLC information; each resource block comprised a single frequency channel and a single time slot. To avoid collisions, we applied the NOMA scheme to mMTC and URLLC information so that m/uCUAV used the same frequency and timeslot in different power allocations. In primary communication, the PT transmits information to the assigned primary users. As a secondary communication, m/uCUAV measures the signal-to-interference-plus-noise ratio (SINR) of the PT. The mathematical term Ω(γ p ) is denoted as [23]: where γ p is the SINR of PT, and γ t is the intended SINR. It should be noted that "1" implies an occupied resource block, and "0" indicates an unoccupied resource block. In secondary communication, m/uCUAV uses unoccupied resource blocks.

C. Signal Model
The NOMA signal transmitted by the (k-1)-th and k-th m/uCUAV can be represented as where x u and x m are the transmitted signals of URLLC and mMTC information, respectively. P u and P m are the transmit powers of the URLLC and mMTC information, respectively.
Because P u is URLLC information, we assume P u > P m . The received signal at the gNB can be expressed as follows where x p is the transmitted signal of the PT; P p is the transmit power of the PT; g pg denotes the channel gain between the PT and gNB; and n g is the additive white Gaussian noise (AWGN) with a mean of zero and variance σ 2 g . Based on the spectrum sensing scenario discussed in Section II-B, two types of SINR are described as follows.
(i) Effectual SINR: when the PT is inactive in communication; and (ii) Intrusion SINR: when the PT is active in communication.
1) Effectual SINR: In such a scenario, there is no interference from the PT. The SINR of the k-th URLLC information can be expressed as follows: where For the transmission of mMTC and URLLC information, the gNB first decoded the k-th URLLC information followed by (k-1)-th URLLC information, k-th mMTC information, and (k-1)-th mMTC information using the successive interference cancellation (SIC) method. The SINR of the (k-1)-th URLLC, k-th and (k-1)-th mMTC information can be expressed as follows: where Furthermore, Θ 1 , Θ 2 , and Θ 3 are SIC errors with values ranging from 0 to 1, where Θ = 0 and Θ = 1 indicate correct and incorrect SIC, respectively.
2) Intrusion SINR: In such a scenario, there is interference from the PT. The SINR of the k-th and (k-1)-th URLLC information are as follows: The SINR of the k-th and (k-1)-th mMTC information can be expressed as follows:

D. Propagation Channel Model
Each propagation link comprises the path-loss, fast fading, and antenna gain of the gNB. Therefore, each channel gain is calculated as [23] g xy r xy = A ag (φ, θ) where x and y are transmitter and receiver, respectively; r xy is the two-dimensional (2D) and three-dimensional (3D) distance represented as r 2D ui , respectively;g l is the lineof-sight (LoS) link with |g l | = 1;g n is the non-LoS (NLoS) link considering the randomly scattered link with zero mean and unit variance; κ o is the Rician factor; and A ag is the antenna gain of the gNB.
In this study, we adopted two path-loss models based on the 3GPP channel model: the U2G (m/uCUAV to gNB) and P2G (PT to gNB) channels. In addition, we briefly discussed the gNB antenna gain based on the 3GPP model. 1) U2G Channel: Each m/uCUAV has an LoS and NLoS link with varying probabilities depending on the m/uCUAV and gNB locations. The LoS probability for an m/uCUAV hovering between m/uCUAV and gNB is given by [5] where r o is measured in meters. The values of p o and r o can be represented as p o = 4300 log 10 (h ui ) − 3800 and r o = max(460 log 10 (h ui ) − 700, 18), respectively. In [5], [24], the average path-loss model was calculated as follows where the path-loss for the LoS and NLoS links is given by Pl l xy = 28 + 22 log 10 (r 3D xy ) + 20 log 10 (f ) and Pl n xy = −17.5 + (46 − 7 log 10 (h ui )) log 10 (r 3D xy ) + 20 log 10 (40πf /3), respectively; and f is the carrier frequency.
2) P2G Channel: By considering the property of the P2G channel, the path-loss factor is given as [25] PL xy = max Pl l xy , Pl n xy , where Pl l xy = 28 + 22log 10 (r 3D xy ) + 20log 10 (f ), ; h b is the average building altitude; and h t and h r are the altitudes of the transmitter and receiver, respectively.
3) gNB Antenna Gain: As stated in [26], the gNB antenna gain was calculated as the antenna element gain and beamforming gain (array factor). The gain of each antenna element is given by: Authorized licensed use limited to the terms of the applicable license agreement with IEEE. Restrictions apply. where φ and θ are the azimuth and elevation angles, respectively, between the main beam direction of gNB and m/uCUAV; G m = 8 dBi is the maximum directional gain of the antenna elements; φ 3dB = 65 o and θ 3dB = 65 o are the 3 dB beam-widths of the horizontal and vertical patterns, respectively; A f = 30 dB is the front-back ratio; A s = 30 dB is the side-lobe level limit; and A e,h (φ) and A e,v (θ) are the antenna radiation patterns in the horizontal and vertical directions, respectively.
For beamforming [27], the total gNB antenna gain from (16) can be expressed as where is the antenna array gain of the gNB; and N a is the number of antennas.

E. User Plane Latency
User plane (UP) latency is a significant challenge for the interaction and monitoring of CUAV with a secure and reliable approach in the CUAV-IoT network. The UP latency, denoted by t upl is the time required to transfer information from the transmitter to the receiver. As determined by [1], it consists of eight parts.
where t ss represents the time required for spectrum sensing and determines any unoccupied resource block over the unlicensed spectrum; t pp is the pre-processing time necessary for the transfer of information, such as channel requests, scheduling grants, and queuing delays; t en and t de denote the encoding and decoding of the information, respectively; t t denotes the time required for transmitting a single fragment of information; and t prop denotes the propagation time of information between the transceiver pair. In (18), we mention two times t t + t prop due to the transmission and re-transmission. Because the UP latency budget has been proposed as one millisecond (ms), conferring beyond 5G, one transmission time interval (TTI) is equivalent to 0.125 ms, based on (18).

III. PERFORMANCE METRICS
This section presents a theoretical analysis of the data rate, energy efficiency, and propagation latency for the considered network model.

A. Data Rate
Based on the spectrum sensing scenario discussed in Section II-B, the following two types of data rates exist (i) Effectual data rate: when PT is inactive in communication, and (ii) Intrusion data rate: when PT is active in communication. The theoretical expressions for the data rate of paired m/uCUAVs are presented below.

1) Effectual Data Rate:
Considering the URLLC information, the total effectual data rate of K p paired m/uCUAVs for Ω = 0 is given by [1], [28] and t s denote the paired bandwidth, channel block length, the probability of decoding error, transmission time, and sensing time, respectively; Q −1 (·) is the inverse of the Q-function; and p r (Ω = 0)(1 − p f ) denotes the perfect detection probability, where p f represents the probability of a false alarm.
Similarly, for the decoding of mMTC information, the total effectual data rate of K p paired m/uCUAVs for Ω = 0 is given as [1] where 2) Intrusion Data Rate: Considering the intrusion from the PT, we determined the intrusion data rate of the mMTC and URLLLC information. Thus, the total intrusion data rate of the URLLC information of K p paired m/uCUAVs for Ω = 1 is given by where Moreover, the total intrusion data rate of the mMTC information of K p paired m/uCUAVs for Ω = 1 is given by where

B. Energy Efficiency
The energy efficiency of the considered network is a crucial characteristic because it is directly related to the network lifetime of the m/uCUAV operation. As mentioned in [1], the theoretical expressions of energy efficiency at the (k−1, k)-th paired m/uCUAV (∀(k − 1, k ) ∈ K p ) for the effectual and intrusion cases are given by where P c indicates the power consumption for spectrum sensing, information processing, encoding, and decoding. Furthermore, P hov indicates the propulsion power consumption during the hovering of m/uCUAV and can be expressed r, η, and W denote the profile drag coefficient, air density (kg/m 3 ), rotor solidity, rotor disc area (m 2 ), blade angular velocity (rad/s), rotor radius (m), incremental correctional factor of induced power, and weight of m/uCUAV (N), respectively. For simplicity, we considered the same power P c and P hov for all m/uCUAV devices.

C. Propagation Latency
Based on the discussion in Section II-E, the TTI value is 0.125 ms. Therefore, the propagation latency is derived as: where T c = 0.125 ms.
On the basis of (19) and (21), the propagation latency at the (k−1, k)-th paired m/uCUAV (∀(k − 1, k ) ∈ K p ) for the effectual and intrusion cases is defined as where N b denotes the number of bits to be transmitted.

IV. ENERGY EFFICIENCY MAXIMIZATION
To enhance the energy efficiency of our proposed model, we formulated an optimization problem and subsequently solved it for the effectual and intrusion cases.

A. Effectual Case
In the effectual case, the optimization problem hereafter referred to as problem IV-A, of energy efficiency maximization was formulated by jointly optimizing the transmit power in terms of the total power budget and SINR constraints such that the propagation latency and data rate requirements were satisfied. The optimization problem of the K p paired m/uCUAVs can be expressed mathematically as where P tk = P k + P k −1 = 2P t represents the entire power budget (Simply assume that each m/uCUAV has the same power budget). υ k u and υ k m represent the lower limits of the SINR for URLLC and mMTC information, respectively, for the k-th pair of m/uCUAVs. υ k −1 u and υ k −1 m represent the lower limits of the SINR for URLLC and mMTC information, respectively, for (k-1)-th pair of m/uCUAVs. Constraint (28a) refers to the energy efficiency of paired m/uCUAV for the effectual case provided in (23). Constraint (28b) describes the overall power limitation for a paired m/uCUAV. Constraints (28c) and (28d) describe the SINR necessities for the k-th and (k-1)-th pairs of m/uCUAVs to achieve the URLLC latency requirements provided in (4) and (5). Constraints (28e) and (28f) describe the SINR necessities for the k-th and (k-1)-th pairs of m/uCUAVs to achieve the minimum specified data rates provided in (6) and (7). Constraint (28g) is used to ensure that constraints (28c) and (28d) are greater than constraints (28e) and (28f). To solve this problem, we rearrange constraints (28c), (28d), (28e), and (28f) and rewrite problem (28) as It is observed that the constraints of problem (29) are linear functions. However, it is still challenging to solve the problem (29) because its objective function is nonconvex with respect to P u and P m . Two issues observed in the problem were m/uCUAV pairing and transmit power allocation. Consequently, we divided the problem into two sub-problems to solve it: (a) m/uCUAV pairing, and (b) transmission power optimization. Fig. 3 provides an overview of the entire procedure for solving the optimization problem. We employ Algorithm 1 in the initial step to identify the m/uCUAV pairing. The optimal transmit power is then determined in the second step by executing the Algorithm 2.
1) Pairing Scheme: Based on [29], [30], three types of pairing schemes were proposed: (i) high-high channel gain of two m/uCUAVs; (ii) high-low channel gain of two m/uCUAVs; and (iii) uniform channel gain of two m/uCUAVs. In [29], the usefulness of sum capacity maximization in a NOMA system was demonstrated using a uniform m/uCUAV pairing technique. However, the uniform channel gain of two m/uCUAVs was more complex for practical implementation because the high-gain m/uCUAV was paired with the mid-gain m/uCUAV. Another issue was that the synchronization occurred over a long distance between two m/uCUAVs. For the high-low channel gain, the capacity (data rate) gain of the pair decreased. This was due to the fact that reducing the power of a highgain m/uCUAV also reduced its capacity. In contrast, when the power of a low-gain m/uCUAV was raised proportionally, its capacity gain was smaller than the capacity loss of a high-gain m/uCUAV. Similarly, with uniform channel gain, synchronization is a significant issue over long distances between two m/uCUAVs. To address the aforementioned drawbacks of highlow and uniform channel gains, we used a high-high channel gain of two m/uCUAVs. The whole process of the m/uCUAV pairing method is outlined in Algorithm 1.
2) Optimizing Transmit Power: Based on the above discussion in problem (29), the following theorem provides the optimal solution for the transmit powers.
To obtain the optimum transmit powers, we calculate the minimum SINR requirements for URLLC and mMTC information using (A.3), (A.4), (A.5), and (A.6). We obtain some mathematical expressions in (A.11), (A.12), (A.13), and (A.14) by applying the KKT condition to the optimization problem in (29). From (A.11) -(A.14), the coefficient matrix (C o ), transmit power matrix (V), and constant matrix (C) for Cramer's rule are found. In the end, we find the optimal solutions using Cramer's rule. From the aforementioned subject, the proposed transmit power allocation algorithm is summarized in Algorithm 2.

B. Intrusion Case
Following the effectual case, we formulated the optimization problem of the K p paired m/uCUAVs, hereafter referred to as problem IV-B, as follows Algorithm 2 Transmit Power Allocation 1. Input: Given the simulation parameter according to Table I; 2. All m/uCUAVs will be paired and found channel gain (g p ) determined using Algorithm 1; 3. for all (k − 1, k ) ∈ K p do 4. Determine the SINR lower limits for URLLC and mMTC information if It is observed that the optimization problem (30) is similar to the problem (28). Constraint (30a) refers to the energy efficiency of paired m/uCUAV for the intrusion case provided in (24). Constraint (30b) describes the overall power limitation for a paired m/uCUAV. Constraints (30c) and (30d) describe the SINR necessities for the k-th and (k-1)-th pairs of m/uCUAVs to achieve the URLLC latency requirements provided in (8) and (9). Constraints (30e) and (30f) describe the SINR necessities for the k-th and (k-1)-th pairs of m/uCUAVs to achieve the minimum specified data rates provided in (10) and (11). Constraint (30g) is used to ensure that constraints (30c) and (30d) are greater than constraints (30e) and (30f). In (30c), (30d), (30e), and (30f), P p |g pg | 2 is included in the SINR provided in (8), (9), (10), and (11), respectively. Because P p |g pg | 2 is constant, the proof can be determined by Theorem 1. Therefore, the same optimal transmit powers (P k * u , P ) are obtained using both the effectual and intrusion cases. The method to maximize energy efficiency for effectual and intrusion scenarios with propagation latency is outlined in Algorithm 3.

C. Implementation of Algorithms
m/uCUAV pairing is planned based on channel gain, characterized by a small-scale path-loss. In Algorithm 1, the m/uCUAV pairing scheme is presented. The optimal transmit power was determined using KKT conditions and Cramer's rule. In Algorithm 2, the transmit power allocation is presented. All the above-mentioned procedures were performed using ground gNB. Each m/uCUAV was required to send local information (i.e., locations and maximum average Algorithm 3 Energy Efficiency Maximization of Pairing Algorithm for Effectual and Intrusion Cases 1. Input: Given the simulation parameter according to Table I, SumEE E = 0, SumEE I = 0, Sumt propE = 0, and Sumt propI = 0, etc; 2. All m/uCUAVs will be paired and found channel gain (g p ) determined using Algorithm 1; 3. for all (k − 1, k ) ∈ K p do 4. Determine the transmit powers:

D. Complexity of Algorithms
The following items were considered for calculating the system complexity: (i) m/uCUAV pairing scheme, and (ii) optimal solution of transmit power. Hence, the complexity of the proposed model lies primarily in Algorithms 1 and 2. The complexity of Algorithm 1 involved in solving optimization problem is O (KA c1 A s1 A p1 ), where A c1 , A s1 , and A p1 represent the arithmetic operations of m/uCUAV channel gain, sorting of all m/uCUAV channel gains, and the pairing of m/uCUAVs, respectively. The complexity of Algorithm 2 involved in solving optimization problem is O (K p A l2 A k 2 A c2 ), where A l2 , A k 2 , and A c2 represent the arithmetic operations for determining the SINR lower limits, the KKT condition, and Cramer's rule, respectively. Consequently, the total complexity of Algorithms 1 and 2 is

V. PERFORMANCE EVALUATION
In this section, we evaluated the performance of the proposed CUAV-IoT network for the mMTC and URLLC services. As shown in Fig. 4, we assumed an enclosed area of 300 × 600 × 100 m 3 ; herein, PT and gNB were positioned at (10,500,20) and (0,0,15), respectively. We consider that ten m/uCUAVs were randomly generated from the same area. For the considered network, g ug channel gain was computed using (12), (13), (14), (16), and (17); and g pg channel gain was computed using (12) and (15). Moreover, the simulation parameters are provided in Table I. In the flowing graphs, average effectual energy efficiency and average effectual propagation latency are defined as the mean values of energy efficiency and propagation latency for the effectual case (PT is inactive). Similarly, the average intrusion energy efficiency and average intrusion propagation latency are defined as the mean energy efficiency and propagation latency values for the intrusion case (PT is active). Fig. 5 shows that the average energy efficiency initially increases due to the CUAV altitude (h ui ) increment of approximately up to 120 m. However, when the CUAV altitude exceeded 120 m, the energy efficiency decreased with increasing altitude. The highs and lows of energy efficiency can be seen for m/uCUAV because of the LoS probability and path-loss of the link. For the effectual case, the optimal altitude was 120 m, and the maximum energy efficiency was 8289.54 bits/J and 12733.1 bits/J for m/uCUAV 3 [1] and the proposed m/uCUAV 4 pair, respectively. For the intrusion case, the optimal altitudes were 100 m and 120 m for m/uCUAV [1] and the proposed m/uCUAV, respectively. In addition, the maximum energy efficiency was 471.20 bits/J and 1313.13 bits/J for m/uCUAV [1] and the proposed m/uCUAV, respectively. Hence, the proposed scheme improved by 53.60% for effectual and 178.68% for intrusion cases.  Fig. 6 shows the average propagation latency for the effectual and intrusion cases. For the effectual case, the propagation latency was constant at 0.118 ms for m/uCUAV 5 [1] and 0.074 ms for the proposed m/uCUAV. 6 However, in the inference case, the propagation latency decreased for an altitude increment of approximately up to 120 m and then increased with increasing altitude. From this analysis, it can be inferred that the propagation latency for m/uCUAV [1] was 2.03 ms, whereas the propagation latency for the proposed paired m/uCUAV was 0.713 ms at an altitude of 100 m. Hence, applying the proposed scheme, we observed a 184.71% reduction in the intrusion case. Fig. 7 presents that the average energy efficiency decreases for both the effectual and intrusion cases as the horizontal distance increases (r 2D ug ). For the effectual case, the energy efficiency was 8290.43 bits/J and 12733.2 bits/J for  m/uCUAV [1] and the proposed paired m/uCUAV at 200 m, respectively. Contrastingly, the energy efficiency was 656.33 bits/J and 1384.66 bits/J for m/uCUAV [1] and the proposed paired m/uCUAV at the same distance in the intrusion case, respectively. Fig. 8 illustrates the average propagation latencies for the effectual and intrusion cases. The curve's tendency in Fig. 8 is similar to that of Fig. 6. In the intrusion case, the propagation latency increased with increasing distance. We analyzed the propagation latency at 200 m and found that the propagation latency was 1.47 ms and 0.678 ms for m/uCUAV [1] and the proposed paired m/uCUAV, respectively. Hence, the propagation latency decreased by 116.81% for the intrusion case when the proposed scheme was applied. Fig. 9 shows the average intrusion propagation latency owing to various transmitted bits. From the observations in Figs. 6 and 8, it can be inferred that the propagation latency was higher than 0.125 ms for the intrusion case. Because of the intrusion case, it was possible to reduce the propagation latency to less than 0.125 ms by increasing the entire power budget and decreasing the number of transmitted bits.  As shown in Fig. 9, the propagation latency increased for both schemes with the number of bits transmitted. The propagation latency was much lower than 0.125 ms owing to the low number of bits transmitted.

B. Compared to the OMA Scheme
We compare our proposed NOMA pairing scheme to the OMA scheme to demonstrate its usefulness. This scheme was identical to the proposed design except that it employed the OMA scheme during transmission rather than the NOMA scheme. The bandwidth was equally partitioned and allocated to each m/uCUAV device under the OMA system. Moreover, each device was assigned an orthogonal bandwidth allotment, ensuring that m/uCUAV devices did not interfere with one another. Fig. 10 shows the average energy efficiency of the OMA and NOMA schemes as a function of altitude. As shown, it can be seen that the energy efficiency of all systems improved with altitude and eventually decreased with altitude. Observations revealed that the proposed NOMA pairing scheme consistently achieved superior performance compared to the OMA scheme. Regarding effectual energy efficiency, the proposed  scheme improved performance by about 81.33% compared to the OMA scenario at 100 m. Regarding intrusion energy efficiency, the proposed scheme improved performance by around 399% compared to the OMA scenario at 100 m. Fig. 11 shows the average propagation latency of the OMA and NOMA schemes as a function of altitude. Observations revealed that the proposed NOMA pairing scheme regularly outperformed the OMA scheme. Regarding effectual propagation latency, the proposed scheme reduced performance by about 595.89% compared to the OMA scenario at 100 m. Regarding intrusion propagation latency, the proposed scheme reduced performance by around 613.18% compared to the OMA scenario at 100 m.

C. Compared to Benchmark Schemes
The proposed pairing method is evaluated against two reference schemes: (i) the equal power and (ii) channel inversion approaches to determine its efficacy. The two reference frameworks are compared below. Equal power: Given that the m/uCUAV does not know the channel gains, assigning equal power to mMTC and URLLC information is appropriate. Channel inversion: Channel inversion is a kind of power regulation in which the transmitted power is inversely proportional to the channel quality. Therefore, the received signal power at the receiver is constant. Fig. 12 shows the performance of proposed pairing schemes compared to two reference schemes as the number of SIC coefficients varies. As depicted in Fig. 12, the proposed pairing scheme considerably outperformed the equal power and channel inversion methods as the number of SIC coefficients increased. It can be seen that when the SIC coefficient increased, the average energy efficiency performance decreased. This was because interference terms associated with SIC operation led to a decline in system energy efficiency performance. In general, the SINR decreased as the SIC coefficient rose. The proposed pairing scheme improved performance in terms of effectual energy efficiency by 26.73% and 25.85%, respectively, at 0.5 SIC coefficient for equal power and channel inversion. The proposed pairing scheme improved intrusion energy efficiency by about 27.48% and 26.59%, respectively, for equal power and channel inversion at a SIC value of 0.5.

VI. CONCLUSION
We developed a mathematical model to analyze and compare the pairing schemes of multiple CUAV-assisted mMTC and URLLC services. The NOMA approach was employed to avoid collisions between mMTC and URLLC information. We derived mathematical expressions for the data rate, energy efficiency, and latency for such a network using the FBLT scheme. In addition, we formulated an optimization problem to achieve the maximum energy efficiency for paired m/uCUAV devices by considering the presence and absence of intrusion. The simulation results revealed that the CUAV pairing scheme is essential for minimizing latency while optimizing energy efficiency. Our simulation results demonstrated the superior performance of the proposed scheme compared to the existing method. Also, the proposed method outperformed the traditional OMA method. Moreover, according to simulation study, the proposed scheme outperformed three benchmark schemes in terms of SIC coefficient. To further improve the applicability of the presented methods and analyzes, future work would explore a more general scenario, such as a network including uplink and downlink scenarios. In addition, it would be interesting to examine interference mitigation and CUAV trajectory design in this scenario. Moreover, the physical layer security of the proposed method will be analyzed.

APPENDIX PROOF OF EFFECTUAL CASE
According to (28a), the Hessian matrix of the objective function is negative. Hence, the optimization problem is non-convex. Therefore, we determined the minimum SINR requirements of ED k −1 u and ED k u using (19b) and (19c). From (26), we obtain Substituting ED k −1 u and ED k u into (A.1) and (A.2), we obtain the equivalent expressions of (19b), (19c), (20b), and (20c) as follows  [31].