Dipanshu Modi * School of Computer Engineering, KIIT (Deemed to Be) University, Bhubaneswar, Odisha, India ; 2205032@kiit.ac.in . Received: 29 September 2023 Modi , D . ( 2024 ). IoT - . Revised: 21 January 2024 Metaversalize , 1 ( 2 ), 72 - 80 . Accepted: 23 Mar ch 2024

The rise of Internet of Things ( IoT ) networks in smart cities has provided a range of advantages, such as real - time traffic management and optimized energy usage. Nonetheless, these networks encounter hurdles related to data traffic overload, fluctuating network conditions, and the necessity for scalable communication frameworks. Artificial Intelligence (AI) - driven routing protocols present a promising alternative by effectively overseeing data flow, minimizing latency, and maximizing resource utilization. This research paper examines the eff ectiveness of AI - based algorithms in improving routing performance within smart city IoT networks. It investigates several AI methods, including Machine Learning (ML)

and Reinforcement Learning (RL) , that react dynamically to network conditions, anticipate traffic loads, and determine optimal paths for effective data transmission. By incorporating these AI - enhanced protocols, smart cities can achieve lower energy usage, enhanced scalability, and grea ter dependability. Comparative simulations between traditional and AI - driven routing protocols indicate that AI - based approaches can considerably alleviate congestion and improve response times, even in densely populated urban IoT environments.

This resear ch provides valuable insights into the implementation of AI - enabled routing in smart city IoT networks, laying the foundation for more intelligent and sustainable urban infrastructures .

Artificial intelligence , R outing protocols , Smart city , I nternet of things .

The concept of smart cities has rapidly evolved, driven by the Internet of Things ( IoT ), which enables a wide range of applications to improve urban infrastructure and quality of life [1] . From real - time traffic control to waste management and energy efficiency, IoT networks in smart cities rely on interconnected devices and sensors to collect and transmit data [2] . However, IoT networks' increasing density and complexity in urban environments present significant challenges, especially regarding data routing, network congestion, and communication reliability [3] , [4] . Efficient routing protocols are crucial for managing data traffic, conserving energy, and maintaining low latency in high - density networks .

Corresponding Author : 2205032@kiit.ac.in https://doi.org/10.22105/metaverse.v1i2.62 Licensee System Analytics . This article is an open - access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0). Modi | Metaverse . 1 ( 2 ) ( 2024 ) 72 - 80 While effective for smaller - scale IoT networks, traditional routing protocols often struggle with smart cities'

scalability and dynamic conditions. This is where Artificial Intelligence (AI) - driven routing protocols can offer substantial benefits. AI - based algorithms, such as Machine Learning (ML) and Reinforcement Learning (RL), enable dynamic and adaptive routing by analyzing real - time data, predicting traffic patterns, and adjusting communication paths accordingly [5] . Integrating AI with routing protocols can enhance IoT network performance by reducing congestion, improving response times, and optimizing resource usage [6] , [7] .

Table 1. Comparison between traditional and artificial intelligence - driven routing protocols in internet of things networks . Protocol Type Advantages Challenges Traditional Simple, low computational Limited scalability, struggles with dynamic data loads . routing overhead . AI - driven Adaptive, optimized for real - time Higher computational cost, complexity in implementation . routing conditions .

The primary routing challenges for smart city IoT networks are due to high device density, unpredictable traffic patterns, and limited energy resources. Unlike conventional networks, IoT - based smart city networks must operate efficiently across numerous diverse devices, often constrained by battery life and processing capabilities. Additionally, IoT devices generate enormous amounts of data, resulting in high traffic loads and potential congestion in communication networks. As illustrated in Fig. 1 , traffic congestion can lead to delayed data delivery, particularly during peak urban activities like rush hours .

Fig. 1 . Traffic congestion in urban internet of things network .

AI - driven routing protocols aim to mitigate these challenges by leveraging predictive and adaptive algorithms. ML models can analyze historical data to identify traffic patterns, allowing the system to predict high network load periods and preemptively adjust routes [8] . For instance, RL algorithms can enable nodes to dynamically select paths based on reward functions, such as minimizing latency and energy consumption. These AI - based approaches support real - time decision - making, enabling smart city IoT networks to maintain optimal performance under fluctuating conditions .

IoT - Table 2. Outlines the key objectives of implementing artificial intelligence - driven routing protocols in smart city internet of things networks. Objective Description Reducing network congestion Dynamic routing based on real - time traffic predictions . Enhancing scalability Adaptive routing to accommodate increasing device density . Minimizing energy consumption Intelligent path selection to extend IoT device battery life .

Improving data reliability Ensuring accurate data delivery with minimal packet loss .

This paper investigates the potential of AI - driven routing protocols to enhance the efficiency and reliability of smart city IoT networks. We analyze various AI algorithms, including ML and RL , to determine their effectiveness in addressing the unique demands of urban IoT deployments. The study includes simulations comparing traditional and AI - driven routing protocols under different network conditions to evaluate performance metrics like latency, throughput, and energy efficiency. This research aims to comprehensively und erstand how AI can revolutionize routing in smart city IoT networks, facilitating smarter, more resilient urban infrastructures .

T he proliferation of IoT devices in smart cities has revolutionized urban infrastructure by enabling real - time data collection and analysis, enhancing decision - making processes, and improving quality of life. However, the rapid expansion of IoT networks has also created new challenges, particularly regarding data routing, network congestion, and scalability. Traditional routing protocols used in IoT networks often fall short in these areas, especially when network conditions fluctuate frequently. In respons e, researchers have turned to AI - driven routing protocols, which leverage ML and RL to dynamically optimize routing paths based on real - time data analysis and predictive modelling.

Traditional IoT routing protocols, including Shortest Path routing, directed diffusion, and cluster - based routing, are generally efficient for smaller networks with static, predictable traffic patterns. However, these methods are often inadequate in high - density urban IoT networks due to their limited adaptability to dynamic conditions and scalability issues. Traditional protocols rely on fixed algorithms, leading to increased latency and data packet loss under heavy loads .

Conversely, AI - driven routing protocols, such as those based on Q - learning and Deep Reinforcement Learning (DRL), offer adaptive, real - time optimization of routing paths [9] . These protocols analyze traffic data, predict congestion, and adjust routing decisions dynamically, which enhances network efficiency and reduces latency. Table 3 compares traditional and AI - driven routing protocols regarding scalability, adaptability, and computational complexity .

Table 3. Comparison of routing protocols based on scalability, real - time adaptability, and computational complexity. Routing Protocol Scalability Adaptability to Real - Ti me Conditions Computational Complexity Traditional routing Low Low Low AI - driven (Q - learning) Medium Medium Medium AI - driven (DRL) High High High Modi | Metaverse . 1 ( 2 ) ( 2024 ) 72 - 80

AI - driven routing protocols for IoT networks in smart cities primarily employ ML and RL techniques. RL algorithms, such as Q - learning, enable network nodes to learn optimal routing paths by interacting with the environment and receiving rewards based on performance metrics, such as minimized latency or energy consumption. According to a study by Lalani et al. [10] , Q - learning improves data throughput in IoT networks by up to 25% compared to traditional protocols under fluctuating traffic conditions .

DRL further extends the capabilities of RL by incorporating deep neural networks, allowing for more complex decision - making processes. In a smart city environment, DRL can predict future traffic loads based on historical data, dynamically adjusting routing to avoid congestion. Research by Wang et al. [11] demonstrated that DRL - based routing protocols reduced average latency by 35% compared to traditional cluster - based routing in high - density urban networks area .

AI - driven routing protocols are evaluated based on key performance metrics: 1) latency, 2) energy consumption, 3) throughput, and 4) Packet Delivery Ratio (PDR). These metrics are particularly important for smart city applications where real - time data transmission is crucial, such as traffic management and . I. Latency: The time delay in data transmission is critical in evaluating routing protocols. AI - driven methods that adapt to traffic loads can significantly reduce latency.

II. Energy consumption: IoT devices in smart cities often have limited power resources, and AI - driven protocols can optimize energy usage by selecting routes that consume less power. III. Throughput: This measures the amount of data successfully transmitted across the network. Adaptive routing protocols generally achieve higher throughput by avoiding congested paths. IV. PDR: PDR indicates the success rate of data packets reaching their destination. Higher PDR in AI - driven routing reflects improved reliability in dynamic network conditions.

Table 4 . Average performance metrics for various artificial intelligence - driven protocols compared to traditional routing . Metric Traditional Routing Q - Lea rning DRL Latency (ms) 150 110 90 Energy consumption High Medium Low Throughput (Mbps) 40 55 65 PDR 85% 92% 95%

AI - driven routing protocols have proven effective in a range of smart city applications. For instance, in traffic management, AI algorithms use historical data to predict and alleviate congestion by dynamically rerouting data through less busy nodes. A study by Arévalo and Jurado [12] on smart grid networks showed that AI - driven routing could reduce power consumption by 20%, a critical benefit for energy - constrained IoT systems .

Another application is emergency response. In scenarios like disaster management, where network infrastructure may be compromised, adaptive AI protocols enable fast data rerouting, ensuring critical information reaches emergency teams without delay. AI - driven protocols have shown a 15% improvement in emergency response times compared to static routing methods, as reported . IoT -

This study investigates the effectiveness of AI - driven routing protocols in enhancing communication efficiency in smart city IoT networks. The methodology involves setting up a simulated IoT network, implementing traditional and AI - driven routing protocols, and analyzing various performance metrics, including latency, throughput, energy consumption, and PDR . The experiment used the NS - 3 and OMNeT++ network simulators, known for their ability to model large - scale IoT environments accurately .

The IoT network model includes 500 to 2,000 devices distributed across a simulated urban area. Devices are connected to multiple edge nodes that act as local processing centers. Data collected from these devices is either processed locally or routed through the n etwork to a central cloud server, depending on the priority of the data. For instance, time - sensitive traffic data is routed through edge nodes to minimize latency, while less critical data is sent to the cloud for storage and analysis .

Table 5. Network simulation parameters. Parameter Value Number of devices 500 - 2,000 Number of edge nodes 20 Simulation duration 48 hours Network bandwidth 100 Mbps Routing protocols tested Shortest Path, Q - learning, DRL

Three types of routing protocols were implemented in the simulation to compare their performance : I. Shortest Path routing : A traditional method that selects the path with the least number of hops without considering network congestion. II. Q - learning algorithm: A RL - based algorithm that uses a reward mechanism to discover optimal paths based on performance metrics like latency and energy consumption. Each node selects a route that maximizes cumulative rewards, calculated using the formula: ?

? ? • ? ƒ ? ? ? • ? ƒ ? ? ? ? ” ? ? ƒš ? • ? ƒ ? ? ? • ? ƒ ? ? ? (1) where: 1) Q(s, a): Current Q - value for state - DFWLRQSDLU VD ? Learning rate, 3) r: Reward UHFHLYHG ? Discount IDFWRUDQG V?D? Next state and action. III. DRL: A more advanced AI technique combining RL with deep neural networks to handle complex routing decisions. DRL algorithms, such as Deep Q - Networks (DQN s ), can predict traffic congestion and optimize routes by leveraging large amounts of historical and real - time data.

The protocols were evaluated based on four primary metrics: 1) latency, 2) throughput, 3) energy consumption, and 4) PDR . I. Latency (ms): The delay in data packet transmission was measured by recording the time difference between sending and receiving packets across nodes. Lower latency indicates faster data delivery, essential for smart city applications. II. Throughput (Mbps): Throughput measures the total data successfully transmitted per second, calculated using.

Modi | Metaverse . 1 ( 2 ) ( 2024 ) 72 - 80 ‘–ƒŽƒ–ƒ”ƒ•ˆ‡””‡†?„‹–•? Š”‘—‰Š’—– ? ? ‘–ƒŽ‹‡?•‡…‘†•? III. Energy consumption (J): The total energy used by each protocol was computed using energy models embedded within NS - 3, which simulate real - world energy usage patterns for IoT devices . IV. PDR: PDR was calculated as ‘–ƒŽƒ…‡–•‡Ž‹˜‡”‡† ? ? ???? ? ‘–ƒŽƒ…‡–•‡– Higher PDR indicates greater reliability in data transmission .

Table 6 . Performance metrics. Metric Description Latency (ms) Average time delay in data transmission . Throughput (Mbps) Data transferred per second . Energy consumption (J) Total power consumption for each protocol . PDR (%) The ratio of packets delivered successfully to packets sent .

$IWHU VLPXODWLRQ GDWD FROOHFWHG RQ HDFK SURWRFRO·V SHUIRUPDQFH ZHUH DQDO\]HG XVLQJ 3\WKRQ WR JHQHUDWH comparative graphs. These include : I. Latency comparison graph: A line graph comparing the average latency of each protocol across different traffic loads, showcasing how Q - learning and DRL respond to dynamic network conditions. II. Throughput and PDR bar chart: A bar chart showing the throughput and PDR for each protocol, illustrating the efficiency of AI - driven protocols in maintaining higher data flow and reliability.

III. Energy consumption graph: A bar chart comparing the energy consumption of each protocol, where AI - driven protocols are expected to demonstrate lower energy usage due to optimized path selection. Fig . 2 . Latency comparison across protocols .

A t - test was conducted to statistically compare AI - driven protocols' mean latency and energy consumption to validate their effectiveness over traditional ones. The null hypothesis assumed no significant difference between traditional and AI - driven protocols, while the alternative hypothesis anticipated that AI - driven protocols would yield better performance metrics . IoT -

The simulation results provide a comparative analysis of the three routing protocols ³ Shortest Path , Q - learning, and DRL ³ focusing on key performance metrics: latency, throughput, energy consumption, and PDR. The findings demonstrate significant performance improvements in AI - driven routing protocols, particularly with DRL, over traditional routing metho ds .

the average latency across the three protocols under varying traffic conditions. Both Q - learning and DRL significantly reduce latency compared to the Shortest Path protocol, with DRL achieving a 40% reduction in high - GHQVLW\VFHQDULRV7KLVLPSURYHPHQWLVGXHWRWKH'5/DOJRULWKP·VDELOLW\WRSUHGLFWDQGDYRLGFRQJHVWHG routes, ensuring quicker data delivery . Table 7 . Average latencies across different protocols .

Protocol Average Latency (ms) Shortest Path 150 Q - learning 110 DRL 90 These findings are critical for smart city applications that require real - time data transmission, such as traffic monitoring and emergency response systems. By proactively adjusting routing paths, AI - driven protocols minimize delays, enhancing the responsi veness of IoT networks in urban environments .

Throughput measures the data successfully transmitted per second, a crucial metric for high - density IoT networks. As shown in Table 8 , both Q - learning and DRL exhibit higher throughput than the Shortest Path , with DRL achieving the highest throughput at 65 Mbps, compared to 40 Mbps for the Shortest Path . This increase in throughput is attributed to the adaptive nature of AI - driven protocols, which dynamically allocate network resources to accommodate fluctuating traffic loads .

Table 8 . Throughputs across different protocols . Protocol Throughput (Mbps) Shortest Path 40 Q - learning 55 DRL 65 The results suggest that AI - driven protocols are better suited for handling data - intensive IoT applications. They maximize network efficiency and ensure that higher volumes of data are transmitted without compromising speed or reliability .

Energy efficiency is essential in IoT networks, as many devices rely on limited power sources. Table 9 illustrates energy consumption for each protocol, with DRL consuming the least energy, reducing power usage by 20% compared to the Shortest Path SURWRFRO7KLVLVGXHWR'5/·VRSWLPL]HGURXWLQJGHFLVLRQVZKLFKPLQLPL]H the number of hops and reduce network congestion, resulting in lower energy demands across the network .

Table 9 . Energy consumption across different protocols. Protocol Energy Consumption (J) Shortest Path 100 Q - learning 85 DRL 80 Modi | Metaverse . 1 ( 2 ) ( 2024 ) 72 - 80 Lower energy consumption in AI - driven protocols translates to prolonged battery life for IoT devices, critical for smart city deployments requiring continuous operation. AI - driven routing's notable advantage is the ability to conserve energy without sacrificing network performance .

PDR is a measure of reliability, indicating the percentage of data packets successfully delivered to their destination. Table 10 shows that DRL achieved a PDR of 95%, outperforming both Shortest Path and Q - learning. DRL's high PDR indicates its robustness in maintaining data integrity, even in high - density network conditions, reducing packet loss, and enhancing communication reliability . Table 10 .

Packet delivery ratio across different protocols . Protocol PDR (%) Shortest Path 85 Q - learning 92 DRL 95 Comparative analysis and discussion The results demonstrate that AI - driven routing protocols, particularly those based on DRL, consistently outperform traditional protocols across all performance metrics. The latency reduction observed in AI - driven protocols is crucial for applications that depend on timely data delivery. At the same time, increased throughput ensures higher data flow rates necessary for large - scale smart city networks. Furthermore, the significant decrease in energy consumption highlights the efficiency of AI - driven protocol s in prolonging device battery life, which is critical in IoT networks with constrained power resources .

The improvements in PDR also underscore the reliability of AI - driven protocols. By predicting and adapting to changing network conditions, DRL reduces packet loss, a critical factor in maintaining data integrity and reliability in smart city applications such as environmental mon itoring, public safety, and urban mobility . These findings suggest that AI - driven routing protocols, particularly DRL, offer significant advantages for smart city IoT networks. Their adaptability to real - time conditions and ability to optimize routes based on predicted traffic loads make them well - suited for complex urban IoT environments. However, it is essential to consider that the computational requirements of DRL are higher than those of simpler algorithms like Shortest Path . Future edge computing and hardware optimization developments will be necessary to deploy DRL efficiently across large - scale IoT networks without overwhelming local resources .

This research highlights the effectiveness of AI - driven routing protocols in improving the performance and scalability of IoT networks within smart cities. By leveraging adaptive AI techniques such as Q - learning and DRL, these protocols significantly reduce latency, increase throughput, optimize energy consumption, and improve the PDR . The results show that DRL, in particular, consistently outperforms traditional routing methods, making it well - suited for high - density urban IoT networks where real - time data transmission is critical .

AI - driven protocols' ability to predict network congestion and adapt to fluctuating traffic patterns allows for more efficient data routing and resource utilization. This is essential for the sustainability of IoT devices with limited energy resources. While AI - driven protocols require higher computational power, advancements in edge computing can facilitate their integration in large - scale IoT networks .

In conclusion, AI - driven routing protocols offer a promising solution for addressing the complex communication demands of smart city IoT infrastructures. Future research should explore hybrid models that optimize these protocols for real - world deployment, enabling smarter, more resilient urban systems . IoT -

Dipanshu Modi was responsible for all aspects of the research, including conceptualization, methodology development, data analysis, algorithm implementation, drafting, review, editing, and visualization .

This research received no external funding.

The data used and analyzed during the current study are available from the Dipanshu Modi upon reasonable request .

Dipanshu Modi declare no conflicts of interest regarding the publication of this paper . If necessary, these sections should be tailored to reflect the specific details and contributions .

[1] Behura, A., & Mohapatra, H. (2019). IoT based smart city with vehicular safety monitoring . https://easychair.org/publications/preprint/Zdk7 [2] Dey, A., Bose, S., Bisht, R., Yadav, M., & Mohapatra, H. (2025). Solutions for energy management and environmental conservation in IoT - integrated smart home. In AI and emerging technologies for emergency response and smart cities (pp. 389 ? 408). IGI Global Scientific Publishing. https://doi.org/10.4018/979 - 8 - 3693 - 9770 - 1.ch018 [3] Zainaddin, D. A., Hanapi, Z. M., Othman, M., Ahmad Zukarnain, Z., & Abdullah, M. D. H. (2024). Recent trends and future directions of congestion management strategies for routing in IoT - based wireless sensor network: A thematic review.

Wireless networks , 30 (3), 1939 ? 1983. https://doi.org/10.1007/s11276 - 023 - 03598 - w [4] Akinola, O. I. (2024). Adaptive location - based routing protocols for dynamic wireless sensor networks in urban cyber - physical system s. Journal of engineering research and reports , 26 (7), 424 ? 443. https://doi.org/10.9734/jerr/2024/v26i71220%0A [5] Tshakwanda, P. M., Arzo, S. T., & Devetsikiotis, M. (2024). Advancing 6G network performance : AI/ML framework for proactive management and dynamic optimal routin g.

IEEE open journal of the computer society , 5 , 303 ? 314. https://doi.org/10.1109/OJCS.2024.3398540 [6] Usama, M., Ullah, U., Muhammad, Z., Bux, M., Ullah, I., Rouf, M., & Islam, T. (2024). Data traffic management in AI - IoT network to reduce congestion. In artif icial intelligence for intelligent systems (pp. 145 ? 189). CRC Press. https://doi.org/10.1201/9781003496410 [7] Dikshit, S., Atiq, A., Shahid, M., Dwivedi, V., & Thusu, A. (2023). The use of artificial intelligence to optimize the routing of vehicles and reduce traffic congestion in urban areas.

EAI endorsed transactions on energy web , 10 , 1 ? 13. http://dx.doi.org/10.4108/ew.4613 [8] Khan, A. B. F., & Ivan, P. (2023). Integrating machine learning and deep learning in smart cities for enhanced traffic congestion management: An empirical review. Journal of urban development and management , 2 (4), 211 ? 221. https://doi.org/10.56578/judm020404 [9] Yang, M., Gao, D., Zhang, W., Yang, D., Niyato, D., Zhang, H., & Leung, V. C. M. (2025). Deep reinforcement learning - based joint caching and routing in AI - driven n etworks.

IEEE transactions on mobile computing , 24 (3), 1322 ? 1337. https://doi.org/10.1109/TMC.2024.3481276 [10] Rezagholi Lalani, S., Safaei, B., Hosseini Monazzah, A. M., Taghizadeh, H., Henkel, J., & Ejlali, A. (2024). QUERA: Q - learning RPL routing mechanism to establish energy efficient and reliable communications in mobile IoT network s. IEEE transactions on green communications and networking , 8 (4), 1824 ?

1839. https://doi.org/10.1109/TGCN.2024.3399455 [11] Wang, Y., Li, Y., Wang, T., & Liu, G. (2022). Towards an energy - efficient data center network based on deep reinforcement learning. Computer networks , 210 , 108939. https://doi.org/10.1016/j.comnet.2022.108939 [12] Arévalo, P., & Jurado, F. (2024). Impact of artificial intelligence on the planning and operation of distributed energy systems in smart grids.

Energies , 17 (17), 4501. https://doi.org/10.3390/en17174501