Blockchain-Enabled M2M Communications for UAV-Assisted Data Transmission

Abstract

Internet of Things (IoT) technology has uncovered a wide range of possibilities in several industrial sectors where smart devices are capable of exchanging real-time data. Machine-to-machine (M2M) data exchange provides a new method for connecting and exchanging data among machine-oriented communication entities (MOCE). Conspicuously, network services will be severely affected if the underneath IoT infrastructure is disrupted. Moreover, it is difficult for MOCEs to re-establish connectivity automatically. Conspicuously, in the current paper, an analysis is performed regarding potential technologies including unmanned aerial vehicles, blockchain, and mobile edge computing (MEC) that can enable the secure establishment of M2M communications networks that have been compromised to maintain the secure transmissible data. Furthermore, a Markov decision process-based joint optimization approach is proposed for blockchain systems that aims to elevate computational power and performance. Additionally, the dueling deep Q-network (DDQ) is incorporated to address the dynamic and complex optimization issue so that UAV selection is ensured to maximize performance. The results of experimental simulation with several statistical attributes suggest that the proposed framework can increase throughput optimally in comparison to state-of-the-art techniques. Additionally, a performance measure of reliability and stability depicts significant enhancement for the proposed framework.

1. Introduction

The Internet of Things (IoT) consists of a large number of interconnected devices [1]. Wearable, automotive electronics, smart grids, industrial automation, and other machine-to-machine (M2M) communications are illustrations of machine-oriented communication entities (MOCEs) that can interact intelligently with each other to formulate an M2M communication system [2,3]. M2M connections are expected to reach nearly 15 billion by the beginning of 2024 as per reports by research institutes and corporations [4,5]. M2M connections will undoubtedly overwhelm that of human-to-human (H2H) communication in the near future [6]. M2M communication applications including smart meters, ubiquitous surveillance, and intelligent grids are implemented in distant locations without external supervision. Most MOCEs send and receive data automatically and periodically [7].

1.1. Research Domain

IoT networks and interconnections are complex to recover in the event of a natural disaster that damages the infrastructure. Damaged M2M communication networks require quick network repair and the re-establishment of data links. Conspicuously, the unmanned aerial vehicle (UAV)-based IoT system has already been suggested as it has received a lot of attention from academics and the industry. Such devices can be used as flying objects for data acquisition or mobile base stations (BSs) or even relay nodes to communicate with IoT devices. Figure 1 shows the wide scope of UAVs in industrial applications [8]. IoT and UAV-assisted wireless networks have been researched for several state-of-the-art studies in recent years. Furthermore, an emergency wireless network in catastrophes can be set up with the use of UAVs, and a multi-hop communication technique can be incorporated to share data. IoT swarm network architecture, described by Rovira et al. [9], is designed to reduce link average latency and boost data delivery throughput.

Furthermore, a model for UAV-specific M2M linking for escape-route detection during a natural disaster has been developed by Karem et al. [10]. It was able to limit the energy utilization at the UAV when it was relaying information for MOCEs. However, limited studies have focused on data security and privacy when it comes to UAV-assisted IoT or M2M communications. Machines communicating without human interaction must maintain data security and privacy as they are capable of communicating sensitive information including invoices and other financial records [6]. On the other hand, UAVs operating as BSs or relay nodes in M2M communications networks can be deployed by various operators for temporal incorporation in the underneath network. Henceforth, UAVs are not trustworthy for MOCE-assisted networks. As a result, a distributed trust mechanism is needed for M2M communications aided by UAVs.

1.2. Research Motivation

The motivation for using blockchain or decentralization is to create a more secure, transparent, and decentralized system that is less prone to manipulation or fraud. By leveraging a distributed ledger, blockchain technology may allow safe and transparent transactions without the need for a central authority or mediator. This may also lead to improved efficiency and cost savings, as well as enhanced privacy and control for users. Additionally, decentralization can help to reduce the risk of single points of failure or censorship, making the system more resilient and resistant to attacks or disruptions.

1.3. State-of-the-Art Contribution

Once the MOCEs transmit the data, UAVs act as a node in the blockchain, storing the information in the blockchain system for data verification and consensus technique implementation. Complex computing operations cannot be performed primarily for blockchain systems, resulting in restricted batteries and resources in UAVs. Fortunately, another promising technology known as mobile edge computing (MEC) is used in the M2M communication system to offload and execute processing workloads on the MEC server. Henceforth, elevated dependability and efficacy of data transmission and processing in UAV-assisted M2M networks are registered. By using UAVs, blockchain, and MEC to repair broken M2M communications networks, an optimization framework is presented to address the aforementioned challenges. The proposed framework aims to enhance the information computing capacity and performance of the blockchain system. For learning, training, and deducing the best solution, the Markov decision process (MDP) is defined. Moreover, an effective technique dubbed the dueling deep Q-network (DDQ) is presented for enhanced efficacy. Based on the aforementioned aspects, the specific contributions of the presented research are as follows.

• An M2M communications network with a blockchain-enabled UAV-assisted data transfer is proposed. MOCEs ensure a reliable transmission link to the BS via suitable UAVs during broken communication links. UAVs are used to operate as mobile base stations and offer ubiquitous coverage and transmission assistance.
• The proposed network architecture includes blockchain technology to assure the security and reliability of M2M communications facilitated by UAVs. The confidentiality and integrity of the transmissible data are assured via the system’s distributed ledger and contract mechanism.
• MEC is presented to perform computing jobs and increase cooperative computation efficiency. Information computing capacity and performance optimization of the blockchain framework are analyzed in the context of dynamic networks and available resources. Moreover, DDQ-assisted optimization is presented to address complex features for an efficient decision process.

UAV-assisted M2M communication networks bring up several intriguing open research questions and problems. The conceptual view of the blockchain-based data acquisition technique is presented in Figure 2 [11]. Moreover, Figure 3 shows a generic block structure in the blockchain [12].

1.4. Paper Organization

Section 2 discusses some of the vital contributions in the current domain of study. The proposed blockchain-inspired UAV data communication architecture is discussed in Section 3. The presented architecture is validated in experimental simulations in Section 4. Finally, Section 5 concludes the current study with future research directions.

2. Related Works

In the current section, state-of-the-art research developments in UAV, blockchain, and MEC research are explored. Moreover, an research analysis of MEC and blockchain integration in M2M communications using UAVs is presented.

2.1. UAV-Assisted Communication Drones

UAVs can be used as flying BSs or transmission nodes to assist terrestrial communication networks [13]. UAVs have been incorporated in several significant perspectives as communication frameworks [14]. The use of UAVs as computing and caching nodes in smart cities is ensured by researchers [15]. The detailed features and uses of UAVs in communication networks are explored ahead.

2.1.1. Communications Using UAVs

Research review has deduced that 5G and future wireless cellular networks have scalability and interference issues that cannot be disregarded even if several technologies can be combined to formulate larger capacity, wider coverage, and reduced latency [16,17]. UAVs embedded with communication facilities allow communication services to map service needs, including increasing communication service domain, reducing overload in the large number of small networked cells, and developing a dynamic and scalable network architecture [18,19]. Creating a dependable and competent communication system quickly and efficiently is essential in variable heterogeneous emergency scenarios [20]. Aljumah et al. [21] proposed an IoT-based application for secure data communication in a vehicular environment. The authors incorporated a secure hashing technique for secure data storage. The proposed technique was able to perform better as compared to state-of-the-art techniques for securing data. Dahmane et al. [22] proposed a novel blockchain-empowered AI technique utilizing edge computing for secure data communication. The proposed technique provisioned decentralized computation without a centralized server. Enhanced results were registered by the proposed technique upon deployment. With UAVs as a potential solution, wireless communication can be flexible and dependable [23]. Damaged or full destruction of existing terrestrial infrastructures can occur after natural catastrophes [24]. There are ways to re-establish communication links using UAV-assisted wireless networks [25]. When compared to traditional means of communication, UAVs have significant benefits in emergency scenarios, as shown in [26].

2.1.2. IoT Communication Using UAVs

As the IoT-inspired smart city notion grows in popularity, the need for reliable wireless communication between smartphones and MOCEs has been elevated [27]. IoT devices are not able to send data across long distances because of energy limitations [11]. While MOCEs are incorporated in remote places including mountains and terrains, traditional communication frameworks are not available for data transmission [28]. As a result, UAVs can be used to connect MOCEs and BSs via dependable communication channels. With the help of mobility, UAVs can provide enormous MOCEs with real-time location updates [29].

2.1.3. UAVs for Data Computation

With its centralized processing component (micro-CPU), UAV acts by computing and caching nodes for a variety of low-power data computation and caching activities [30]. Incorporating UAVs, the computation and storage load on the network can be reduced by caching, computing, and forwarding delay-tolerant data over a reasonably extended period [31].

2.2. MEC-Inspired Blockchain System

The MEC-inspired blockchain system is a combination of blockchain technology and the mobile edge computing paradigm [32]. Bitcoin transactions are initial industrial services to incorporate blockchain peer-to-peer (P2P) ledgers [33]. Transacting in an anonymous, trustful manner and eliminating any middlemen ensures the privacy and security of data [34]. The advent of blockchain technology in communication networks and the IoT has provided several benefits, including trustlessness, transparency, automation, decentralization, and security [35]. Without a centralized third party, the blockchain system operates correctly. A single point of failure in a decentralized system based on blockchain is impossible to achieve [36]. To ensure elevated security, the blockchain-enabled IoT network uses one-way cryptographic hash functions and consensus procedures to link all collected data [37]. IoT will benefit from the advantages of blockchain technology because of its ability to provide a secure, trustworthy, and transparent network environment [38]. Even though the IoT and communication networks gain greatly from blockchain technology, they must compete with computer power. As a result, the rising costs of computing are an inevitable issue. It is conceivable to provision compute transmission services with reduced delay and energy effectiveness by deploying MEC servers at the edge of networks [39]. The BSs with MEC servers near the devices running blockchain applications can outsource complex calculation chores. Since both blockchain and MEC have similar decentralized qualities and interdependent functions, their integration becomes inevitable.

2.3. Blockchain-Inspired M2M Communications

M2M communications incorporate device deployment in distant places that are deployed without human intervention, including ubiquitous surveillance, intelligent grids, or smart meters [40]. Consequently, network failure or disconnection is unavoidable in these situations. With the advent of UAV-assisted communications, these issues can be efficiently addressed. Using micro CPU or storage as a BS or relay node, UAVs may communicate data to BSs and MOCEs on the ground. UAVs can quickly restore communication links that have been destroyed and construct a self-organizing network architecture from the ground [41]. However, data security and privacy are also significant considerations for M2M connections. Security hazards and problems are inherent in the usage of UAVs since they are managed and operated by several people. Blockchain technology’s transparent, decentralized, and immutable features make it an appealing tool for ensuring security and dependability [9]. Even though UAV-assisted M2M communications can benefit from blockchain, the need for powerful computing is thought to be another tough issue, and the complex calculation jobs cannot be performed directly and independently by MOCEs or UAVs [42]. Depending on its characteristics, MEC helps to perform computing jobs that bring MEC into the network architectures. Network congestion and energy use will be erased as a result [27]. Blockchain and MEC will be huge assets to the M2M industry when used in conjunction with UAVs to facilitate M2M communications. Based on the aforementioned related works,

Table 1. Comparative assessment (Y: available, ||: not available).

Ref.	UAV	Blockchain	M2M Communication	Cogn. Decision	Security	Real-Time	Efficiency
Li et al. [32]	Y	Blockchain in communication	Y	||	||	Y	Y
Zuo et al. [33]	Y	Blockchain	||	Y	||	||	||
Zhang et al. [34]	Y	Blockchain in cloud	Y	Y	Y	||	Y
Bhattacharya et al. [36]	Y	Blockchain in edge communication	Y	||	Y	Y	Y
Xu et al. [37]	Y	Blockchain in communication	Y	||	Y	||	Y
Feng et al. [38]	Y	Blockchain in data communication	Y	||	Y	Y	Y
Liu et al. [39]	Y	Y	Y	||	||	Y	Y
Aljumah et al. [21]	Y	Blockchain in data communication	Y	Y	||	||	Y
Dahmane et al. [22]	Y	Blockchain in data communication	Y	||	Y	||	||
Wen et al. [43]	||	Optimal reinforcement learning	||	||	Y	Y	Y
Pham et al. [44]	||	Optimal reinforcement learning	||	||	Y	||	||
Proposed Technique	Y	Blockchain with optimized data communication	Y	Y	Y	Y	Y

is formulated to depict the comparative analysis with the state-of-the-art frameworks in the current domain of study.

3. Proposed System Architecture

The generic layered architecture of the proposed network framework is presented in Figure 4. Specifically, a blockchain framework is designed for the UAV network for M2M data communication. The proposed framework is used for a multi-rotor UAV model with wireless communication and computation capability, capable of aerial applications. A detailed explanation of each layer is discussed ahead.

3.1. Network Design

The network is designed for blockchain-enabled M2M communications by including N tiny cells, M MOCEs for heterogeneous services, and 1 BS empowered with MEC servers and controllers for M2M data transfer assistance. U UAVs are also deployed in the network to service tiny cells, such that U = N. Because of its micro-CPU and blockchain subsystem, the U_nth UAV in the n^th (∀ n = 1, 2,…, N) tiny cell can connect directly to MOCEs and create high-quality and short-range linkages with the MOCEs to gather data effectively. It is assumed that there are M_n MOCEs in the n^th small cell that can transfer the information to the U_n^th UAV within the enhanced wireless communication range when wired communication fails. After the information is acquired from MOCEs, UAVs serve as the computing nodes for the blockchain in the presented framework to send the information. The blockchain nodes use consensus procedures to validate data, ensuring that only data that has not changed will be accepted by the system. This procedure necessitates a lot of processing power. Henceforth, it can be assessed locally on UAVs or transmitted to servers. The controller, which is connected to the BS, can perform the decision-making.

3.2. Network Modeling

The modular structure of the proposed network’s communication is discussed ahead.

3.2.1. Effective Communication Module

UAVs and MOCEs are assumed to have symmetrical up- and down-link channels in the communication model when data communication is performed in the simultaneous coherence window. It is believed that channel radio propagation consists of Rayleigh fading and route loss. C_mn, U_n($\Delta$t). and C_Un, BS($\Delta$t) indicate the data traffic capacity from n_m^th MOCE to the U_n^th UAV or from U_n^th UAV back to the base station while the temporal window $\Delta$t is active.

3.2.2. Computational Module

UAVs and MEC servers are capable of performing computation tasks. Henceforth, the proposed network model has two distinct data computation modes: local computing and MEC. It is assumed that the computational job at the $\Delta$t temporal instance is represented as M_Un($\Delta$t) = [$\beta$_Un($\Delta$t), $\alpha$_Un($\Delta$t)] where the input data size is indicated as $\beta$_Un($\Delta$t) and the total number of CPU cycles necessary to complete the calculation task at $\Delta$t are assumed to be denoted as $\alpha$_Un($\Delta$t).

Simple computing activities can be performed by the U_n^th UAV outfitted with an integrated micro-CPU. As a measure of UAV’s processing speed, the quantity G_Un represents the total CPU cycles/s. It is not viable to conduct complex computational activities such as mining and smart consensus with blockchain systems exclusively with UAVs. As a result, the first computations must be delegated to a BS equipped with MEC servers. Offloading computation data transmission time is denoted by the notation S_{Un, off}. After the computations are delegated to the MEC servers linked to the BS, S_{Un, comp} variable is used to reflect the time spent on these activities. The overall execution time in MEC is thus expressed as S_{Un, mec} taking into account S_{Un, off} and S_{Un, comp}. Moreover, computation capacity in local computing, as well as MEC, are represented as r_{n, local}($\Delta$t), r_{n, mec}($\Delta$t), respectively. The overall processing capability of U_n^th’s UAV in time slot $\Delta$t is depicted as r_{n, total}($\Delta$t) for each of the computation modes.

3.3. Blockchain Framework

UAVs serve as the blockchain node, forwarding data from MOCEs to the network. The current paper uses a consortium blockchain framework for data protection. The detailed procedure of block creation is depicted ahead.

Generation of Block: 

 

Once the presented solution identifies the vulnerable node, the information related to the identified UAV is transmitted to the ledger. Data accumulation acquires transactions to form a unique block. As the consortium blockchain is adopted, the procedure of validation is not necessary. The usage of the locking mechanism in the presented technique is used to provision a warning in case of faulty UAV determination. Furthermore, each stakeholder is a component of the procedure that uses blockchain. In other words, blocks are formulated effectively in the presented framework. On the contrary, compromised nodes can be assessed in the future.

Consensus of Blocks: 

 

As can be seen in Figure 5, the suggested architecture makes use of practical Byzantine fault-tolerant consensus (PBFT) to actualize the consortium’s network [45]. When applied to distributed systems, the PBFT method ensures that all nodes in the network have the same understanding of the system’s state. With PBFT, a chosen leader node proposes a new block of transactions, which is then validated and accepted by the network as a whole. PBFT’s optimization strategy is a three-step procedure: initially, the head node proposes a new block to the rest of the network. After validating the proposal, the other nodes will respond to the leader node with a message indicating that they too agree with the proposed block. When the majority of other nodes have confirmed their agreement with the proposed block, the leader node will notify all other nodes in the network to add it to the blockchain. Using this three-step procedure, we can guarantee that all nodes in the network will eventually reach the same consensus on the system’s state and that no malicious nodes will be able to disrupt it. PBFT is also well-suited for high-performance distributed systems due to its optimization for high throughput and low latency. Moreover, PBFT provides the essential benefits of security, dependability, decentralization, and transaction finality [46]. The suggested approach uses a modified version of the PBFT to get a group decision. As a new block, $Bi$ is generated, the validator announces it to the network. This is how a blockchain is formed. The blockchain network’s drones sign the block’s header after approving the transaction. By signing their identities to the blocks, each user effectively casts their vote. Consortium blockchain and its vote-based approach assist achieve high throughput and low latency, preventing 51% attacks. Algorithm 1 explains the entire consensus procedure in detail. The number of consortium UAVs is denoted by N, and the signature performed by the UAVs on block Bi for verification is denoted by the unique symbol $\mathsf{\mu }$ in the Algorithm.

Algorithm 1 An algorithm of customized PBFT and blockchaining

Require:  $B_i$ Ensure:  Consensus decision Transaction verification for Peer = i to N do     Calculate Signature $B_i$     j++ end for if j ≥ $\mathsf{\mu }$ then     NTP server-based request generation     Perform timestamping and hashing     Consensus = true     Process of blockchain updating = True else     Wait for definite time event $\Delta T$ end if Return consensus

3.4. Hybrid Approach for Secure Data Communication

A discrete MDP formulation is utilized to define the state space, action space, and reward function to optimize system rewards. In light of the dynamic and large-dimensional properties of the presented issue, dueling DDQ is proposed to address it, which involves convolutional neural network formulation for prediction and the Q-learning procedure.

3.4.1. Action Set

Data communication request b_tr($\Delta$t), node identification b_comp($\Delta$t), block size selection b_Sl ($\Delta$t), block delay decision b_Tk ($\Delta$t) in the blockchain network during $\Delta$t are all included in the action domain. Mathematically, b($\Delta$t) = [b_tr ($\Delta$t), b_comp ($\Delta$t), b_Sk ($\Delta$t) and b_Tk ($\Delta$t)]. $\sigma$ is the action vector and b($\Delta$t) ∈$\sigma$ is the composite action that must be satisfied in formal terms.

3.4.2. State Set

It is assumed that the state space is denoted by the S-State. If the state is mapped at every time window $\Delta$t, then it is expressed as s($\Delta$) ∈ S-State. Moreover, the s($\Delta$) in the n^th cell, it is indicated as [s_mn,U_n($\Delta$t), d_mn,U_n,s_mec,s_Stake($\Delta$t)], where s_nm, U_n($\Delta$t) is the wireless link state. It is gained between the m_n^th MOCE and U_n^th UAV. The distance between the m_n^th MOCE and U_n^th UAV is given by the expression d_mm, U_n($\Delta$t). The Shannon theorem can be used to determine the available transmission capacity. s_mec($\Delta$t) describes the status of MEC servers and may be represented as s_mec($\Delta$t) ∈ (0,1), such that s_mec($\Delta$t) = 0 indicates that the edge computation nodes are vacant and ready to perform computations during the time slot. s_mec($\Delta$t) = 1 indicates that the servers are not idle and are unable to do computations. When it comes to blockchain systems, the number of nodes and performance setup affects how many stakes each node can hold; this is represented as s_stake($\Delta$t).

3.4.3. System Reward

With the passage of every temporal slot, the network will get an immediate incentive or reward depending on a specific action. Based on the aforementioned aspects, the incentive function should be correlated to its goal function, which in turn should be aimed at maximizing the system’s data computing capacity and throughput. As a result, the role of anticipated overall system benefits reward is defined as

$$\mathrm{Reward}=\mathrm{ArgMax}\mathrm{F}[{\sum }_{t=0}^{T-1}{\sum }_{n\in N}^{}{\sum }_{u\in U}^{}(\nu r_{n, \mathrm{total}}(\Delta t)+\mu \theta (\Delta t))]$$

where r_{n, total}($\Delta$t) is the cumulative computing capacity of the compute jobs, and $\theta$($\Delta$t) is the transaction performance that has been studied and specified before. Additional weights, $\nu$ and $\mu$, are used to differentiate system rewards such that $\nu$ + $\mu$ = 1. It should be noted that the weights might be dynamic, indicating a heterogeneous preference for the aforementioned optimization goals. To make the modeling process simple, it is assumed that the weights remain constant throughout time. In addition, the reward function takes into account the impact of all MOCEs and UAVs throughout the period.

3.4.4. Deep Reinforcement Learning

To maximize long-term rewards, Q-learning and other reinforcement learning methods rely on learning from the network state and making the best decision possible. Traditional real-time systems have agents that monitor and collect data on system status and then make decisions based on the data. A new time slot is set up, with an agent receiving both a new system state and instant rewards. Using a smart contract is one way to implement reinforcement learning on the blockchain. The conditions of the agreement between end users in a smart contract are encoded into lines of code, making the contract self-executing. To create a decentralized and open system, the reinforcement learning algorithm’s rules and parameters may be encoded in smart contracts and run on the blockchain. By using predetermined criteria and circumstances, such as the current blockchain state or the occurrence of certain events, the smart contract may be configured to perform the reinforcement learning algorithm. Tokens or other types of value may be used inside the smart contract to motivate users to donate data or resources to the algorithm. Using a smart contract to include reinforcement learning in the blockchain allows for the development of a decentralized, transparent, and tamper-proof system. However, it is crucial to design the smart contract carefully, taking into account the restrictions and difficulties of both reinforcement learning and blockchain technology. Figure 6 illustrates the procedure used to discover the best policy to maximize long-term value by repeating the steps above. This is a result of the beneficial aspects of deep reinforcement learning (DRL). To expand the range of applications of Q-learning, neural networks may be added to RL algorithms. This is because neural networks can approximate any nonlinear function. As an additional training aid, the deep Q-learning algorithm incorporates a target DDQ, which is updated at a slower pace to maintain stability and smoothness throughout training. Furthermore, an upgraded method of DRL can determine Q measures with minimal variability and employ the dynamic approach to assure sufficient diversification in the action domain. For estimating Q-measure, the Q-table is initialized with a random value for every state/action pair. This value is updated during training using the Bellman equation: Q(t, a) = Q(t, a) + $\alpha$ ∗ (reward + $\gamma$ ∗ max(Q(t’, a’)) − Q(t, a)), where Q(t, a) is the Q-value for state-t and action-a $\alpha$ is the learning rate, the reward is the immediate reward for taking action a in state t, $\gamma$ is the discount factor, max(Q(t’, a’)) is the maximum Q-value for all possible actions a’ in the next state t’. It is feasible to determine state value and action rewards using dueling DDQ, unlike typical value-based DRL. In this way, the computation and derivation of the dueling DDQ algorithm may be expressed as a combined measure of executive actions and the environmental state. Repetition of the same state value may be handled by dueling DDQ, and environmental state estimation can be improved with a defined optimization goal. Henceforth, dueling DDQ is incorporated to determine the best strategy for data transmission requests, node identification, size detection, and block interval in M2M communications. Figure 7 depicts the overall procedure of the proposed DDQs.

4. Experimental Simulation

This section evaluates the performance of the presented technique in comparison to the state-of-the-art techniques. Numerous baseline techniques are considered for the estimation throughput enhancement. Specifically, computational analysis, statistical analysis, reliability, and stability are computed. Each of these has been described ahead in detail.

4.1. Simulation Environment

An Intel W-2133 processor with 32 GB of RAM is used in the simulation. There are 8 CPU cores on the machine. Tensorflow 1.11.0 with Python 3.6 on Ubuntu 18.04.2 LTS is the software environment utilized in the simulations. Using 6 small cells, 1 BS, 200 MOCEs, 7 UAVs, a BS empowered with MEC servers, and a controller in an area measuring 900 m ∗ 900 m, the implementation of an M2M communications network powered by blockchain is performed. It is assumed that UAVs and MOCEs are always in the same place. A UAV outfitted with a micro-CPU provides data re-transmission and data calculation functions in each tiny cell. UAV-based data accumulated dataset is acquired from IEEEDataPort data repository. It comprises several imagery data segments for aerial data communication (Source: https://ieee-dataport.org/documents/uav-aerial-photograpy-vehicle-data-set, 7 April 2023). Moreover, for mapping images, photogrammetry software can be used. It includes utilizing common points to create the final output. However, for image mapping, sufficient overlapping is to be ensured in terms of altitude, and angle of accurate mapping. Optimal camera calibration is required for consistency. Furthermore, training and testing data were considered in the ratio of 75:25 for reducing over-fitting. Initially, the MOCE and UAV’s channel bandwidths are set at 15 MHz. The transmission power is set at 12 mW and 2 W for each MOCE and UAV, respectively, with the background noise of –181 dBm/Hz. The route loss exponent is prefixed to 4 and the channel gain follows a Gaussian distribution with a mean of 0 and a variance of 1. For simulation of reinforcement learning, the Open AI platform was used. Data calculation offloading requires 800 KB of data, and the total number of CPU cycles is 1200. UAV and MEC servers will be able to run at 0.4 GHz and 3.1 GHz, respectively, for CPU processing. In addition, the block size is set at 16 MB, the average time to construct the next block is 3 s, and the average transaction size is 220 KB for the blockchain. The $\mu$ and $\nu$ weights are each set to 0.6.

4.2. Computational Analysis

The CNN serves as the computation framework for determining the desired Q measure for each experiment. In this simulation, a CNN with 5 layers is built and implemented. Figure 8 compares the system’s payouts under various scenarios. It can be seen that the presented approach has a clear benefit over state-of-the-art techniques of greedy-based, logistic regression, and random-selection strategies. The presented method, which has a quick convergence as shown in Section 4.5, has also been proven to be effective. The reason for this is that training via battling DDQ allows for the ideal decision to be taken. The system performance may be improved by making acceptable decisions and selections based on diverse network settings and system statuses. The influence of the blockchain framework on performance is too critical to overlook as part of the proposed scheme. Figure 9 depicts the comparison of system rewards based on the block size computation. With rising block sizes in all schemes, it is shown that blockchain-enabled M2M communications networks can obtain higher system incentives. However, due to the inherited limitations of blockchain technology, it is not possible to grow indefinitely. The blockchain system’s data computation capacity and throughput can be improved because of the training provided by DDQ. When compared with current greedy-based strategies, the suggested method outperforms them since DDQ and focusing on long-term benefits in complete time frames are more important than short-term gains. Based on the experimental simulation, it can be depicted that the proposed system improves data transmission throughput for M2M communications allowed by blockchains using UAV-assisted transmission. Varied dueling DDQ training settings and different blockchain block size constraints reveal significant benefits in the outcomes. In addition to the aforementioned comparative analysis, the state-of-the-art comparison is performed for the incorporated CNN model with several prediction techniques as shown in

Table 2. Statistical result.

Algorithms	Recall (%)	Precision (%)	Accuracy (%)	ROC AUC (%)	F1 Measure (%)	Error (%)
Support Vector Machine	92.23	91.0	91.6	92.56	91.25	6.38
K-Nearest Neighbor	93.25	91.25	94	93.25	92.36	6.51
Naive Bayes	88.02	88.65	89.25	90.01	91.02	6.71
Artificial Neural Network	91.02	91.25	94.25	89.1	94.15	3.71
Decision Tree	90.2	90.12	90.45	90.85	90.55	5.33
Proposed	98.6	95.2	94.5	97.14	96.53	2.23

. Specifically, it can be seen that in the current scenario, the proposed technique can register enhanced performance in terms of recall (98.6%), precision (95.2%), accuracy (94.5%), ROC (97.14%), and F1 measure (96.53%). Additionally, the error rate of 2.23% is registered, which is a significant enhancement over state-of-the-art techniques.

4.3. Stability

This metric is used to measure the stability of a system in a range of 0 to 1. Figure 10 shows the results of the stability estimation. An average MSS value of 0.73 is obtained using the proposed model, with a minimum MSS value of 0.51 and a maximum MSS value of 0.82. Accordingly, the model presents a higher level of stable performance.

4.4. Reliability

This section determines the reliability of the presented model. For performance enhancement, results are compared with state-of-the-art techniques. However, it is important to mention that the proposed model is compared based on the selection criterion used for security purposes as shown in Figure 11. Furthermore, 10% falsified data was used to determine overall reliability estimation. Based on the results, random selection is able to acquire a measure of 72.45%. Greedy selection was able to register a mean value of 69.45%, followed by logistic regression with a 61.34% measure. Based on the results, it is concluded that the proposed model is better compared to other techniques.

4.5. Temporal Delay

Temporal delay is computed to determine the time take by the proposed model for decision-making. It is computed in terms of difference of time of decision-making and data acquisition. Overall results for temporal delay estimation is presented in Figure 12. It can be seen that average delay incorporated for the data assessment and pre-processing is 32.78 s and 45.59 s, respectively. Moreover, the decision-modeling incorporates the overall time of 78.65 s. indicating that the proposed model is temporally effective.

4.6. Limitations

It is important to thoroughly examine the drawbacks of combining blockchain technology with reinforcement learning before incorporating it.

• Both technologies are resource-heavy and need a lot of computer power, making them difficult for smaller businesses or people to use. With the growth in data and transactions, this might also cause scalability problems.
• Both methods may raise privacy issues if widely used. Data privacy and security issues may arise because reinforcement learning needs access to massive volumes of data. On the other side, the blockchain’s inherent transparency and immutability might make it difficult to safeguard private data.
• There is a high bar for entry when it comes to technical expertise and specialized knowledge when combining blockchain and reinforcement learning.
• Regulatory and legal considerations must be made when putting these technologies into action. There may be concerns regarding data ownership, responsibility, and accountability when using blockchain with reinforcement learning.

4.7. Open Challenges

Even though the advantages of data transmission in M2M communications architectures based on blockchain and MEC have been emphasized, still several open research questions and obstacles persist that need to be solved in the future.

4.7.1. Efficient Energy

Energy efficiency cannot be disregarded because of battery limitations when developing UAV-assisted communications. In particular, for UAV-aided machine-to-machine communications in distant places, it is hard to maintain a long-term energy supply. The optimization challenge for energy efficiency in UAV-inspired communications is indispensably vital.

4.7.2. Management of Interference

In the current research, several UAVs have not been taken into account. It has also been made easier to keep the MOCEs in one cell and UAVs in a different cell from colliding. Additionally, UAV-aided M2M data transfer must contend with the issue of managing interference.

4.7.3. Network Planning

There are a variety of channel models for UAV-assisted communications due to the dynamic fluctuation features of location and position. Only basic UAV-aided communications are discussed in this article, not multi-hop relays. In other words, future work should focus on network planning and channel modeling.

5. Conclusions

The current paper presented a unique strategy to simultaneously examine the resource distribution of computational nodes and a blockchain system for UAV-assisted M2M communications to increase the data processing capacity and system performance. Data transmission security and anonymity are both ensured by utilizing the blockchain system, which is integrated into the proposed framework via MEC servers. Dueling the DDQ technique is used to tackle the collaborative decision-making optimization issue because of its dynamic and intricate features. The results of the simulations showed that the suggested framework can enhance system rewards compared to the existing schemes while also maintaining the reliability of the proposed scheme. Efficiencies in UAV-assisted M2M communications can be explored in future research works.