Smart city traffic optimization using IoD and IoT integration

Abstract

Urban traffic congestion, along with the resulting fuel waste and pollution, poses increasing challenges as city populations grow. There is a critical need for techniques that mitigate these effects while maintaining traffic efficiency. This study proposes an integrated framework that combines the Internet of Things (IoT) infrastructure and the Internet of Drones (IoD) to enhance urban traffic management. IoT sensors monitor real-time traffic conditions, while Roadside Units (RSUs) collect and process data to deliver timely updates to vehicles. Drones dynamically extend communication coverage by acting as mobile relay nodes, accelerating traffic information dissemination across wider areas, particularly those with sparse connectivity. Drone placement was optimized for maximum coverage using the Particle Swarm Optimization (PSO) algorithm. Experiments conducted in two urban scenarios—Dammam (Saudi Arabia) and Doha (Qatar)—used the SUMO simulator, Traffic Control Interface (TraCI), and Python to implement communication protocols and adaptive rerouting mechanisms. The findings suggest that integrating IoT and drones can substantially reduce travel delays and vehicular emissions while maintaining real-time operational efficiency. In Dammam, emissions and travel times decreased by up to 40.99% and 32.05%, respectively, and in Doha by up to 48.78% and 43.92%. These results indicate that coordinated IoT–IoD systems can support sustainable urban mobility by improving traffic flow and contributing to cleaner air.

Introduction

Urbanization is rapidly increasing, with 56% of the world’s population—around 4.4 billion people—now living in cities, and nearly 70% projected to do so by 2050¹. This concentration of people creates significant pressure on urban transportation systems, which must provide efficient and sustainable mobility.

Transportation is a critical component of this challenge, facing persistent problems such as congestion, accidents, and air pollution². As vehicle numbers increase, so do the risks of traffic delays, fuel waste, and emissions. Therefore, addressing these issues requires smarter, technology-driven methods.

In response to these urban challenges, the smart city paradigm has emerged, integrating advanced technologies to manage urban resources and improve quality of life³. At the core of modern intelligent transportation, the Internet of Things (IoT) connects road infrastructure, Roadside Units (RSUs), and vehicles through real-time sensing and communication⁴. Drones further extend these capabilities through aerial monitoring and communication, providing enhanced sensing coverage and acting as mobile relay nodes⁵.

Together, these technologies can be integrated into Intelligent Transportation Systems (ITS), which enable dynamic decision-making to improve traffic flow, efficiency, and safety. Research indicates that ITS implementations can reduce travel times by up to 25%, greenhouse gas emissions by 15–20%, and accident rates by 20%^6–8.

Despite these advances, conventional traffic management systems often remain limited. They typically rely on fixed-schedule signal control and respond slowly to incidents, struggling to adapt to real-time disruptions such as accidents or unexpected congestion. This inflexibility leads to prolonged delays, increased fuel consumption, and higher emissions. This research addresses the limited adaptability of traditional systems to dynamic traffic conditions.

Developing such integrated systems requires addressing several limitations of existing IoT–drone frameworks. Many rely on simplified geometric coverage models that overlook realistic wireless channel behavior, including the effects of fading, interference, and obstacles^9,10. Moreover, most prior works focus on isolated functions such as monitoring or control, rather than complete pipelines that integrate incident detection, information dissemination, and vehicle rerouting. Drone placement strategies are also often heuristic and lack rigorous consideration of signal propagation and coverage consistency.

A key technical challenge involves managing the interaction between dynamic traffic demand, incident conditions, and sensing coverage across large urban networks. IoT sensors generate large volumes of data that must be processed quickly to detect congestion and update routes. Drones add a mobile layer that expands coverage but also increases coordination complexity.

This study proposes a system that integrates IoT infrastructure and Internet of Drones (IoD) to improve traffic management. The system is designed for real-time data collection, rapid incident detection, and adaptive rerouting. Its goal is to achieve measurable improvements in travel time and emissions reduction through city-wide information dissemination. This study integrates IoT sensing, RSU-based data aggregation, and optimized drone deployment to create a responsive and sustainable traffic management system.

The research models probabilistic air-to-ground communication links, simulates alert propagation through drones and RSUs, and reroutes vehicles using a cost function that balances emissions and delay. The design aims to achieve adaptive, environmentally aware traffic control suitable for smart cities.

The specific objectives of this work are:

1. Reduce traffic congestion through faster detection and response to incidents.

2. Lower vehicular emissions and fuel consumption to improve air quality.

3. Enhance overall traffic efficiency through adaptive real-time routing.

The proposed IoT–IoD framework detects and disseminates incident alerts, enabling adaptive rerouting to achieve measurable reductions in congestion and emissions. Inspired by¹¹, the routing mechanism employs a modified Dijkstra algorithm¹² that excludes affected roads during route computation. This ensures timely rerouting and improved network efficiency. The study demonstrates the potential of coordinated IoT and drone integration for adaptive and sustainable traffic management.

The remainder of this paper is organized as follows: Section “Literature review and related works” reviews related works on IoT- and drone-based traffic management. Section “Methodology” describes the proposed system architecture, communication models, and rerouting methodology. Section “Simulation results and discussion” presents the simulation setup, performance metrics, and experimental results. Finally, Section “Conclusion” concludes the paper and outlines potential directions for future work.

Literature review and related works

A smart city is commonly defined as “an urban ecosystem that uses advanced technologies, data-driven decision-making, and interconnected systems to improve quality of life, optimize resource use, and enhance urban service efficiency”¹³. At its core, it integrates Industry 4.0 infrastructure—including the Internet of Things (IoT), Artificial Intelligence (AI), Unmanned Aerial Vehicles (UAVs), and modern communication networks—with traditional urban systems to collect data from road, vehicle, and signal sensors¹⁴. This integration supports decision-making processes aimed at reducing congestion, improving energy efficiency, and mitigating environmental degradation.

IoT sensors monitor traffic and environmental conditions in real time to optimize signals, reduce congestion, and lower emissions^15,16. The Internet of Drones (IoD) extends these capabilities through aerial monitoring, emergency response, and by acting as relay nodes in Vehicular Ad-hoc Network (VANET)-based Intelligent Transportation Systems (ITS)^5,17. Together, these technologies enable continuous data collection, efficient resource utilization, and improved public services. However, most existing systems focus primarily on sensing or control. They do not establish a complete operational pipeline that includes incident detection, alert dissemination, and vehicle rerouting.

Internet of things (IoT)

Coined by Kevin Ashton in 1999¹⁸, the IoT refers to the interconnection of machines through the Internet, enabling data exchange and automated decision-making^19,20. It allows city planners to enhance essential services such as waste and energy management and to optimize transportation systems²¹. By supporting adaptive signaling and efficient routing, IoT contributes to sustainability goals and aligns with Sustainable Development Goal (SDG) 11²².

Several studies have explored IoT’s role in traffic management. For example,²³ proposed a system that combines IoT, cloud computing, Fifth Generation (5G), and big data to manage vehicle flow using magnetic detectors. Similarly,²⁴ introduced a three-layer architecture leveraging IoT and Vehicle-to-Everything (V2X) communication to reduce congestion. In another work,²⁵ designed a decentralized system with adaptive signaling based on magnetic sensors and microcontrollers. Furthermore,²⁶ analyzed recurring and non-recurring congestion and proposed IoT-based solutions to mitigate phantom jams and enhance safety. These studies demonstrate IoT’s potential in improving traffic control and reducing congestion. However, they do not address real-time rerouting based on incident detection or assess environmental impact.

In a related contribution,²⁷ presented an IoT-based traffic management system employing Radio Frequency Identification (RFID), Infrared (IR) sensors, and Long Range Wide Area Network (LoRaWAN) to improve emergency response and congestion management. Likewise,²⁸ introduced an Intelligent Traffic Management System (ITMS) leveraging Vehicular Ad-hoc Networks (VANETs) and Internet of Vehicles (IoVs) to optimize signal operations. Additionally,²⁹ implemented a smart monitoring system in Cagliari that integrates mobile and fixed sensors to collect both traffic and pollution data. In a similar vein,³⁰ proposed an Internet of Robotic Things (IoRT) system combining IoT sensors with mobile robots (acting as Roadside Units (RSUs)) to enhance flow and reduce collisions by transmitting real-time traffic information. While these methods are effective in localized sensing and control, they lack mechanisms for dynamic city-wide alert propagation or environmentally aware, vehicle-level routing.

According to³¹, an IoT-driven framework for optimizing urban road networks was developed using an improved chaotic particle swarm optimization (IS-APCPSO) algorithm. Their bi-level planning model jointly minimizes traffic impedance, total vehicle emissions, and infrastructure cost by coordinating road capacity, signal timing, and IoT investment parameters. Simulation results showed that integrating IoT sensing with the IS-APCPSO algorithm led to significant emission reductions compared with traditional models. Although this method effectively addresses large-scale emission management and signal control, it does not consider real-time incident-driven rerouting or vehicular communication dynamics.

In another study,³² proposed an IoT and fog computing-based traffic management system that integrates real-time data from smart sensors with Dijkstra’s algorithm and OpenStreetMap (OSM) for dynamic route optimization. The system achieved low-latency traffic updates through edge processing and demonstrated improved congestion management in simulated urban scenarios. However, it focuses solely on ground-level IoT and fog coordination without incorporating aerial communication layers or multi-entity message dissemination.

In their work,³³ developed an IoT-enabled adaptive traffic management framework that integrates real-time sensor data with predictive analytics and a multiagent simulation model to optimize urban mobility. Using London as a case study, the system employed adaptive signal control and dynamic rerouting to reduce travel times, congestion, and CO emissions by 30%, 50%, and 28%, respectively. The study highlights the value of IoT sensors and agent-based modeling for developing scalable and sustainable traffic management systems. Nonetheless, it remains limited to ground-level adaptive control and excludes aerial relay communication or multi-layered IoD coordination.

Finally,³⁴ proposed an AI–IoT-powered multimodal optimization framework for sustainable urban mobility. The system combines reinforcement learning (RL) and graph neural networks (GNNs) with IoT data from sensors, Global Positioning System (GPS), and public transport systems to dynamically recommend optimal travel modes and routes that minimize emissions and travel time. Simulations conducted in Simulation of Urban MObility (SUMO) showed a 45% reduction in average CO emissions per trip and an 83% user acceptance rate for suggested routes. This framework demonstrates how AI and IoT can enable data-driven, low-emission mobility aligned with several United Nations Sustainable Development Goals (SDGs). However, it focuses on mode selection and emission-aware routing without incorporating IoD-based communication or drone-assisted information dissemination, which are central to this work.

Internet of drones (IoD)

The IoD refers to the use of drones as part of the IoT ecosystem³⁵. As defined by³⁶, it is a system where drones communicate with ground stations via the Internet to synchronize activities and exchange data. Similar to IoT, IoD forms an interconnected network used for applications such as logistics, surveillance, and traffic monitoring³⁷.

Drones frequently operate in coordinated swarms, enhancing system performance through improved spatial coverage, redundancy, and collaborative sensing capabilities. In traffic management, they enable real-time congestion detection, enhance mobility, and assist in incident response³⁸. Deployed across road segments, drones detect events and transmit data to nearby peers, base stations, or vehicles. They also serve as relay nodes that extend VANET coverage. With aerial mobility and adaptive communication, drones strengthen traffic management through dynamic monitoring and rapid data dissemination. However, many such systems do not incorporate vehicle-level rerouting or evaluate the end-to-end impacts of their methods on emissions and travel times.

The authors in⁹ presented a UAV-swarm-based traffic density estimation method using the SUMO simulator. This approach improves urban sensing but does not include communication-aware relay behavior or dynamic route adjustment in response to incidents.

Building on related efforts,³⁹ introduced a UAV-based architecture for estimating traffic densities using sparse data with Kalman filtering and least-squares optimization. Although this method enhances traffic state awareness, it does not enable real-time responses, limiting its ability to mitigate congestion or improve flow.

Within the Saudi context,⁴⁰ addressed high accident rates by proposing a UAV-5G surveillance system to replace SAHER. In another study,¹⁷ explored UAVs as mobile RSUs to ensure line-of-sight connectivity, reduce signal loss, and minimize communication delay. These studies improve communication infrastructure but do not measure downstream effects on overall traffic performance, such as delay reduction or emission minimization.

Further advancing this line of research,¹⁰ proposed AVARS, a drone-assisted traffic control system that employs deep reinforcement learning (DRL) to mitigate congestion caused by sudden events such as road closures. Drones were deployed at key intersections to adjust signal phases using a pre-trained DRL policy. In a SUMO-based simulation of a 1 km urban area in Dublin, the system achieved notable reductions in CO emissions (72.9%) and travel time (67%) after just 10–30 minutes of UAV activity. However, the approach is limited to a few intersections and does not generalize to large-scale multi-point alert dissemination or flexible rerouting across an entire network.

More recently,⁴¹ developed a traffic optimization framework that combines UAV and IoT-based sensing (using induction loops and drone-like sensors) with a Large Language Model (LLM) to dynamically adjust vehicle speeds in SUMO simulations. The AI model receives traffic data and provides speed recommendations every five simulation steps. Applied in three small urban areas for 25 minutes, the approach produced substantial reductions in congestion (up to 82.2%) and CO emissions (up to 75%). However, it does not include decentralized alert propagation or structured RSU–drone coordination for routing decisions, which remain open challenges addressed by this work.

The reviewed literature highlights the need for an integrated IoT–IoD architecture capable of real-time rerouting and environmentally aware traffic optimization. While existing studies demonstrate individual strengths in sensing, communication, or control, none provide a complete pipeline that integrates incident detection, multi-tier alert dissemination, and emission-aware vehicle rerouting across large urban networks. These research gaps motivates the proposed integrated framework that combines IoT sensing, RSU-based data aggregation, and optimized drone deployment for comprehensive traffic management.

These observed gaps directly correspond to the study objectives stated in Section “Introduction”: limited adaptability of IoT-based systems motivates Objective 1 (reducing congestion through faster detection and response); the absence of coordinated IoD–RSU communication motivates Objective 3 (enhancing adaptive real-time routing); and the lack of emission-aware rerouting motivates Objective 2 (lowering vehicular emissions and improving air quality). This alignment shows that each objective directly addresses a specific shortcoming identified in prior work, establishing a clear link between the literature and the goals of this study.

Methodology

Overview of the system architecture

The proposed system architecture integrates key components to manage and optimize urban traffic in real time. It comprises IoT sensors, RSUs, drones, and vehicles that collaboratively gather and exchange data for dynamic decision-making.

IoT sensors are deployed along roads to collect traffic and environmental information such as vehicle count, speed, and emission levels. The collected data are transmitted to nearby RSUs, which act as local data collectors and analysis units.

RSUs communicate with both vehicles and drones to support incident detection and alert dissemination. Drones complement the ground-based infrastructure by serving as mobile relays, extending coverage and ensuring reliable communication across the city. The overall interaction among these components, including their communication links and coverage areas, is illustrated in Fig. 1.

Fig. 1.

Simplified traffic management architecture showing RSUs, drones, and vehicles.

Two types of alerts are considered in the system: accidents and congestion. Accidents are simulated by stopping a vehicle on a given road segment, while congestion is detected when traffic density exceeds a defined threshold. These alerts are distributed across the communication network, allowing vehicles to receive updates and adapt their routes accordingly.

The overall architecture enables real-time monitoring and cooperative response among IoT sensors, RSUs, drones, and vehicles, supporting efficient data flow and dynamic traffic control for reduced delays and emissions.

System components

IoT sensors

IoT sensors are virtually positioned along each road edge in the SUMO network via TraCI, where they continuously monitor vehicle counts, speeds, and emissions. The collected data are transmitted to nearby RSUs through the Message Queuing Telemetry Transport (MQTT) protocol using a local broker (Mosquitto⁴²). Each RSU subscribes to its corresponding topic to receive relevant sensor data, which it processes to generate congestion alerts when specific thresholds are exceeded (Fig. 2).

Fig. 2.

MQTT-based IoT network architecture showing the flow of sensor data from publishers (IoT sensors) to subscribers (RSUs and Logger) via the MQTT broker.

Roadside units (RSUs)

RSUs are deployed within the SUMO network using a greedy set cover algorithm that selects candidate locations based on road midpoints and iteratively chooses those covering the highest number of uncovered road segments while enforcing a minimum separation distance to avoid overlapping coverage⁴³. In total, nine (9) RSUs were deployed in the Dammam network and twelve (12) in the Doha network, each covering a 300 m radius with minimum spatial separation to prevent overlap. Each follows the United States Department of Transportation Federal Highway Administration (USDOT FHWA) Dedicated Short-Range Communications (DSRC) Roadside Unit Specification v4.1 (Siemens RSU)⁴⁴.

Drones

Drones are implemented in the SUMO network using Python scripts through TraCI to support aerial communication and monitoring. Each drone is modeled as an object in three-dimensional Euclidean space, , with position vector . The collection of all drone positions forms the matrix .

The IoD system is represented as an undirected communication graph , where is the set of drones and the set of wireless links. A communication link exists if the pairwise distance and the received power . This forms a binary connectivity matrix :

Each drone follows the specifications of the DJI Phantom 4⁴⁵, operating at 2.4 GHz WiFi with 5 km maximum transmission range, 23 dBm transmit power, -97 dBm receiver sensitivity, and hover accuracy of ±0.3 m horizontally and ±0.1 m vertically.

Optimal placement ensures maximum RSU coverage while minimizing overlap (lower cost). Using the PSO algorithm adapted from⁴⁶, the system achieved full RSU coverage even with only two drones, demonstrating higher spatial and resource efficiency compared with the Greedy and Genetic Algorithm approaches. As summarized in Table 1, PSO attained 100% normalized coverage with as low as two drones, whereas the Greedy algorithm required three drones to reach the same coverage. The slight rise in cost for the two-drone PSO setup remained marginal, showing that PSO provides superior coverage efficiency with minimal additional cost. This confirms the suitability of PSO for optimizing drone placement due to its effective balance between coverage performance and cost efficiency. PSO was implemented with a swarm size of 10, cognitive and social coefficients , inertia weight , and 50 iterations.

Table 1.

Normalized coverage (%) and cost comparison of optimization algorithms.

Algorithm	Number of drones
	1	2	3
PSO	66.67/0.4	100.00/3.2	100.00/1.0
Greedy	55.56/4.2	77.78/2.5	100.00/0.9
GA	11.11/8.2	55.56/4.4	55.56/4.4

Values shown as coverage / cost. Coverage represents the normalized percentage of RSUs served; cost is a normalized measure inversely related to the PSO fitness, combining coverage performance and the penalty for drones placed too close to one another.

Significant values are in bold.

PSO was selected because it is well suited for continuous, non-linear, and typically non-convex optimization problems such as spatial drone placement. Unlike greedy algorithms that often converge to local optima or genetic algorithms that require extensive parameter tuning and longer convergence times, PSO achieves an effective balance between exploration and exploitation with relatively few control parameters. Its velocity–position update rule promotes smooth convergence toward a stable global solution, maintaining steady fitness improvement across iterations observed in this study. This behavior demonstrates PSO’s inherent convergence stability and computational efficiency for continuous search spaces. Its swarm-based search operates directly in continuous 3D space, enabling it to adapt naturally to irregular urban geometries and discontinuous coverage regions, which are well aligned with the dynamic coverage objectives of this study. Compared with GA and Greedy, PSO reached optimal coverage faster, confirming its suitability for real-time optimization in dynamic urban environments.

Communication channel models

Communication among system entities is modeled using established RF channel models commonly applied in ITS and VANET studies^47–49. Both air-to-ground and ground-based channels are mathematically represented to capture physical-layer propagation effects relevant to traffic management.

Air-to-ground (A2G) communication links

Air-to-ground communication between RSUs and drones follows the probabilistic RF propagation model proposed in⁵⁰, which captures signal power variation with distance and environmental conditions. The positions of the RSU and drone are denoted as and , respectively (Fig. 3).

Fig. 3.

Drone-to-RSU communication model.

The channel capacity is defined as:

1

where denotes the channel bandwidth, is the transmit power, represents the total thermal noise power, and is the average air-to-ground path loss (evaluated in linear scale for computation).

The total noise power is given by , where is the Boltzmann constant and the absolute temperature. In the simulations, a bandwidth of ⁵¹, carrier frequency of ⁵², and a noise spectral density of ⁵⁰ were used. Based on these parameters, the total thermal noise power corresponds to approximately over the bandwidth.

The average air-to-ground path loss is expressed as⁵³:

2

Ground-based communication links

Ground communication between RSUs and vehicles follows the Two-Ray Interference model⁵⁴, which accounts for both direct Line of Sight (LoS) and ground-reflected components (Fig. 4).

Fig. 4.

RSU-to-vehicle communication model using the two-ray interference model.

The path loss is calculated as:

3

where is the distance between transmitter and receiver, the signal wavelength, the reflection coefficient, and the phase difference:

4

The reflection coefficient is:

5

where is the relative ground permittivity and the incident angle. The direct and reflected path distances are:

6

and the incident angles are given by:

7

Parameter values from⁵⁴ are  m,  m, and .

This study focuses on physical-layer modeling of A2G and G2G communication to capture realistic signal propagation effects relevant to traffic management. Higher-layer aspects such as network protocols, packet-level simulation, and MAC processes are excluded. These simplifications allow scalable large-network simulation while retaining representative propagation behavior for analyzing alert dissemination and coverage.

Integration and interoperability of IoT and IoD components

The message dissemination mechanism follows a VANET-like emergency broadcast model⁵⁵, where nodes exchange safety alerts using controlled flooding⁵⁶. Each receiving node rebroadcasts only new messages to neighbors within range. To prevent redundant transmissions and broadcast storms, as noted in^57–59, nodes maintain a buffer of unique message identifiers, filtering duplicates and reducing network overhead. In this implementation, a congestion alert is triggered whenever an RSU detects ten or more vehicles on a road edge for a sustained interval.

Vehicle rerouting methodology

Transportation systems contribute substantially to global emissions, accounting for roughly 20–25% of CO output and a similar share of energy consumption^60,61. Road transport, particularly diesel vehicles, is a dominant NO source in urban areas⁶². Reducing these pollutants is therefore a key sustainability goal^63,64.

Modern traffic management combines dynamic traffic assignment (DTA) and eco-routing strategies. While DTA minimizes travel time⁶⁵, eco-routing seeks energy-efficient paths that reduce fuel use and emissions^66,67. Research shows these objectives can conflict^68–71, motivating dual-objective optimization approaches^72–74.

This study’s rerouting methodology optimizes vehicle paths when an incident or congestion alert is received. Each vehicle evaluates its route using a weighted additive cost function that prioritizes emission reduction over travel time:

8

where represents total emissions and the estimated travel time, approximated as distance divided by maximum allowed speed⁶⁸. Various weight combinations were systematically tested through grid search experiments, with values ranging from 0.1 to 0.9 in 0.1 increments. The optimal ratio of (travel time) to (emissions) was selected empirically, as it provided a consistent balance between environmental benefits and traffic efficiency. To ensure comparability, both travel time and emissions were normalized prior to weighting, allowing the cost function in Eq. 8 to remain dimensionally consistent and interpretable.

Emissions estimation model

Emissions are estimated using a two-part model:

9

where is route length (m) and is the base emission factor derived from simulation data, with and denoting total emissions and distance for vehicle .

Delay-induced emissions are estimated as:

where is the total waiting time (s) and the average emission rate (mg/s). Emission rates are obtained from SUMO using the default HBEFA3/PC_G_EU4 model for gasoline Euro 4 passenger cars. For each edge, the mean emission rate for CO, CO, NO, and PM is:

and the network-wide mean rate is:

Total emissions are then:

10

To maintain dimensional consistency, emission rates obtained from SUMO (in mg s) were converted to kg s before aggregation, ensuring that all terms in Eq. 10 are expressed in kilograms (kg) for subsequent analysis and reporting.

Vehicle rerouting process

Upon receiving an alert, each vehicle retrieves its current route and checks for affected edges. If any are marked as avoided due to incidents or congestion, an alternative route is computed using Dijkstra’s algorithm, similar to the approach in⁷⁵. If the new route’s cost (Eq. 8) is lower, the vehicle is rerouted; otherwise, it continues along its current path. Dijkstra is chosen for its preference in SUMO simulations⁷⁶, guaranteed optimality and computational efficiency under dynamic conditions^77,78.

Algorithm 1.

Vehicle rerouting.

Simulation environment and experimental setup

Simulations were conducted using SUMO, TraCI, and Python. SUMO⁷⁹ is an open-source microscopic simulator where each vehicle behaves as an independent agent governed by built-in car-following and lane-changing models (default being Krauss⁸⁰). Vehicles and their mobility pattern were entirely provided by the SUMO environment, following established traffic signals, speed limits, and road rules to realistically represent urban driving behavior and congestion dynamics. All simulated vehicles were homogeneous in characteristics within their respective types (cars, buses, and trucks). Each was assumed to be equipped with On-board Units (OBUs) enabling Vehicle-to-Vehicle (V2V) and Vehicle-to-Infrastructure (V2I) communication for rerouting and alert exchange. TraCI⁸¹ provides real-time control via a TCP-based Python interface.

Two map sections were used: Dammam City and Doha City, extracted from OpenStreetMap and converted into SUMO-compatible networks using the OSM Web Wizard with default parameters. Traffic demand for each network was generated using the default SUMO trip generation utility integrated with the OSM Web Wizard, which randomly assigns origin–destination (OD) pairs across the network based on available edges. Departure times were uniformly distributed over the simulation period to emulate typical peak-hour flow patterns and intersection interactions observed in urban traffic. The Dammam scenario simulated with three vehicles types (cars, trucks, and buses) for approximately 4 hours (scaled to represent peak traffic), while the Doha scenario simulated for 1 hour. These configurations allowed evaluation under different urban densities and traffic demands.

The simulations operated in step-based mode with a 1 s timestep. Vehicle behavior followed SUMO’s default models, and all data were collected via TraCI and processed in Python. Experiments were conducted on a standard desktop system (16 GB RAM, Windows 11).

Simulation results and discussion

Performance metrics and experiments

This study evaluates two of the key performance indicators widely used in intelligent transportation systems research: vehicular emissions and travel time^73,82,83. Vehicular emissions capture reductions in harmful pollutants, while total travel time reflects overall network efficiency. Together, these metrics assess mobility improvement and environmental sustainability.

The pollutants analyzed include CO₂, CO, NO_x, and PM_x, initially measured in milligrams (mg). For consistency, all emission values were converted to kilograms (kg) for analysis and reporting. Emissions were computed by aggregating per-step data across the simulation period to capture total environmental impact.

Three configurations were tested to assess system integration effects: (i) baseline with only vehicle-to-vehicle (V2V) communication, (ii) baseline with added roadside units (RSUs), and (iii) RSU–drone integration. Vehicular emissions and travel times served as the KPIs for all experiments.

Validation of the communication model

The air-to-ground (A2G) channel model was validated to ensure realistic representation of radio propagation between drones and RSUs. Three parameters–received power, attenuation, and data rate–were analyzed against drone–RSU distance, as shown in Fig. 5.

Fig. 5.

Effect of distance on communication metrics: received power, data rate, and attenuation for the A2G link.

As expected, received power and data rate decreased with distance, while attenuation increased. This pattern agrees with standard propagation behavior, confirming that the adopted A2G model accurately captures drone–RSU communication characteristics and provides a reliable basis for later analyses.

Traffic management results

Accidents were simulated on the network road and cleared sequentially after one hour to evaluate the system’s response. The framework was tested under the three configurations described earlier in Section “Performance metrics and experiments”. Travel time (expressed in vehicle-hours, veh-h) and emissions (expressed in kg) were used as evaluation metrics, and a one-dimensional moving average filter smoothed the raw data for clarity. Both the peak and total values are reported in the same units since each represents an accumulation across all vehicles: the total corresponds to the time-integrated value (area under the curve), whereas the peak denotes the maximum instantaneous accumulation during the period. The area under each curve provides the best measure of overall performance.

Travel times

Baseline 1 (V2V) recorded a peak of 106.1 vehh and Baseline 2 (V2V + RSU) recorded 84.7 vehh, serving as references for all configurations. Figure 6 shows total travel times for the Dammam network with one drone. The drone-assisted configuration (V2V + RSU + Drone) reached a peak of 81.1 vehh, while the integrated total travel time across the simulation decreased by 24.08% and 6.3% relative to Baseline 1 and Baseline 2. This reduction likely occurs because the aerial link speeds up how quickly congestion or accident alerts spread, allowing vehicles to reroute before queues form and travel delay increases.

Fig. 6.

Travel time for one drone (Dammam dataset).

Applying the same configuration to the Doha network (Fig. 7) produced a similar pattern. The drone-assisted case reached a peak of 24.38 vehh, much below the baseline peaks, and the integrated total travel time was reduced by 24.48% and 1.7% compared with the fixed Baseline 1 (31.17 vehh) and Baseline 2 (24.61 vehh). This performance similarity shows that the improvement depends mainly on faster information flow rather than city size or density.

Fig. 7.

Travel time for one drone (Doha dataset).

With two drones in Dammam (Fig. 8), the peak value reached 75.18 vehh, while the integrated total travel time fell by 29.88% and 13.4% compared with Baseline 1 and Baseline 2. Wider coverage enabled faster alert delivery and shorter congestion periods. Adding the second drone shortened the time needed for alerts to reach RSUs and vehicles at the network edges, which prevented congestion from spreading to nearby links.

Fig. 8.

Travel time for two drones (Dammam dataset).

In Doha (Fig. 9), the peak value reached 18.06 vehh, while the integrated total travel time was reduced by 43.92% and 27.0% relative to the fixed baselines. These consistent reductions across both networks form a simple parametric sensitivity test over drone count, confirming that performance gains remain stable under different traffic densities. Beyond two drones, additional units offered little improvement, indicating coverage saturation for the network.

Fig. 9.

Travel time for two drones (Doha dataset).

Adding a third drone in Dammam (Fig. 10) produced a peak value of 71.32 vehh, while the integrated total travel time decreased by 32.05% and 16.1% relative to Baseline 1 and Baseline 2. The third drone strengthened coverage between RSUs, improving alert dissemination and routing. Across the 1–3 drone scenarios, travel times decreased predictably as coverage expanded. At three drones, nearly every RSU had a direct link to at least one drone, so additional units provided minimal further benefit. The system therefore reached full communication coverage, marking the point of saturation. Beyond this point, performance stabilized while computation cost rose, indicating that three drones represent the most balanced configuration in the network.

Fig. 10.

Travel time for three drones (Dammam dataset).

Vehicular emissions

For emissions, Baseline 1 (V2V) showed a maximum instantaneous value of 0.33 kg, while Baseline 2 (V2V + RSU) recorded 0.29 kg, which serve as reference levels for comparison. Emission trends followed a similar pattern to travel time. In the Dammam scenario with one drone (Fig. 11), the drone-assisted configuration (V2V + RSU + Drone) exhibited a lower peak of 0.24 kg, and the integrated total emissions over the simulation period decreased by 26.04% and 11.9% relative to Baseline 1 and Baseline 2. The improvement arises from smoother acceleration profiles and reduced idling once congestion is detected earlier.

Fig. 11.

Emissions for one drone (Dammam dataset).

Doha results (Fig. 12) displayed a comparable trend. The drone-assisted configuration reached a peak emission value of 0.12 kg, notably lower than both baselines, while the integrated total emissions over the simulation period decreased by 31.67% and 11.9% relative to Baseline 1 and Baseline 2. Once traffic flow became continuous, the reduction in stop–go cycles naturally led to lower overall emissions.

Fig. 12.

Emissions for one drone (Doha dataset).

With two drones in Dammam (Fig. 13), the peak emission level dropped to 0.22 kg, while the integrated total emissions declined by 34.94% and 22.5% relative to the fixed baselines. This outcome indicates that deploying a second drone enhanced alert dissemination and rerouting efficiency, further reducing idling times and overall emissions.

Fig. 13.

Emissions for two drones (Dammam dataset).

In Doha (Fig. 14), the peak emission value was 0.09 kg, while the integrated total emissions decreased by 48.78% and 34.0% compared with the baselines. These improvements resulted from smoother traffic flow and fewer acceleration spikes as alerts propagated more rapidly through the network. No additional benefits were observed beyond the two-drone configuration. The Doha network, which includes twelve (12) RSUs compared with nine (9) in Dammam, required fewer drones to achieve full connectivity due to its denser ground infrastructure.

Fig. 14.

Emissions for two drones (Doha dataset).

The addition of a third drone in the Dammam network further lowered emissions, resulting in a 40.99% and 29.7% reduction in integrated total values relative to the baselines, as shown in Fig. 15. At this stage, communication coverage was already complete, so additional drones increased cost without yielding further emission benefits. Adding a fourth drone produced no additional improvement, confirming three as the optimal configuration for the current network.

Fig. 15.

Emissions for three drones (Dammam dataset).

Table 2 summarizes total travel time and emissions across both networks. Consistent trends in Dammam and Doha verify that integrating drones with RSUs improves mobility and environmental performance. Together, these parametric comparisons demonstrate sensitivity stability, where small variations in drone number or network type lead to predictable performance changes and indicate a robust, repeatable system response. Reduced idling, smoother acceleration, and shorter trips collectively produced reductions in vehicular emissions, corresponding to a potential air-quality improvement under the tested conditions.

Table 2.

System performance for Dammam and Doha networks under varying drone configurations.

City	Metric (unit)	Scenario	System total	Abs. reduction versus B1	Rel. reduction (%)
Dammam	Travel time (vehh)	1 Drone	35.21	11.17	24.08
		2 Drones	32.53	13.86	29.88
		3 Drones	31.52	14.86	32.05
	Emissions (kg)	1 Drone	0.26	0.093	26.04
		2 Drones	0.23	0.12	34.94
		3 Drones	0.21	0.15	40.99
Doha	Travel time (vehh)	1 Drone	3.60	1.17	24.48
		2 Drones	2.67	2.09	43.92
	Emissions (kg)	1 Drone	0.17	0.077	31.67
		2 Drones	0.13	0.119	48.78

Travel time values represent cumulative totals over the full simulation period for all vehicles (vehh). Percent reductions are computed relative to Baseline 1 (V2V) for each metric and city. Emissions are expressed in kilograms. All absolute values are rounded to two decimal places

Figure 16 compares computational performance across configurations. Drone integration increased average latency from 0.01 s to 0.37 s and CPU usage from 50.36% to 76.93%, while memory remained nearly constant. Despite these overheads, real-time operation was maintained, and the improvements in mobility and emissions clearly outweighed computational cost.

Fig. 16.

Computational performance metrics (latency, CPU usage, and memory usage) for scenarios with and without drones.

Since SUMO operates as a deterministic microscopic simulator where each configuration runs under identical network, demand, and routing conditions, the observed variations arise purely from system design differences rather than random effects. Consequently, statistical significance testing is not applicable, and performance evaluation focuses on comparative trends and percentage improvements across scenarios. Under this deterministic setting, the multi-scenario parametric results serve as a practical sensitivity validation, confirming that observed improvements come from design parameters rather than random variability.

Discussion

The results show that integrating IoT sensors, RSUs, and drones enhances traffic management performance in urban environments. Three configurations were tested: V2V only, V2V with RSUs, and the full IoT–RSU–drone system. Each added layer expanded communication coverage and reduced the time needed for incident detection and response.

The travel-time and emission plots rise as traffic builds up and decline as vehicles complete their trips, reflecting realistic urban flow patterns. The greater reductions observed in the drone-assisted setup indicate that earlier alerts help prevent secondary congestion and support a faster return to free-flow conditions. In the baseline V2V case, alerts remained confined to short ranges, which delayed rerouting and caused longer idling periods. RSUs improved this limitation by broadcasting alerts from fixed positions, while drones further accelerated dissemination by transmitting messages over wider areas. Their elevated position provided line-of-sight access to regions otherwise disconnected, enabling quicker and more coordinated route updates. This layered communication process explains the substantial combined gains in both travel time and emissions.

Performance improved as more drones were added, up to two in the Doha network and three in Dammam, after which the gains leveled off. This plateau occurs once all RSUs and major roads are covered, as additional drones mainly increase processing cost without improving alert delivery. Therefore, efficient allocation is preferable to simply adding more units.

The integrated IoT–RSU–drone framework illustrates how combining ground sensing with aerial relays enhances both mobility (reducing travel times) and air quality (lowering emissions). The two improvements are interconnected: steady traffic flow reduces idling, and less idling lowers emissions. These results suggest that communication speed, rather than node quantity, is the main factor driving performance in this setup.

However, some limitations remain. Vehicle types were uniform, and communication was idealized without packet loss or fading. The analysis focused on network-wide effects rather than intersection-level details. Real-world deployments would also need to consider regulatory constraints, energy limitations, and potential interference. Future work can include varied vehicle classes, longer runs, and adaptive drone movement to test scalability.

Conclusion

This study demonstrated that integrating IoT sensors, RSUs, and drones can significantly improve traffic efficiency and environmental sustainability in urban settings. The proposed framework combines real-time sensing with adaptive rerouting to enhance responsiveness during congestion and traffic incidents. Through the coordinated operation of these components, the system supports rapid incident detection and timely vehicle rerouting across the network.

An adaptive rerouting strategy based on the Dijkstra algorithm enabled vehicles to avoid affected roads, reducing overall congestion and travel delays. Drones extended communication coverage and accelerated alert dissemination in areas with limited RSU connectivity. The analysis indicates that this multi-tier communication approach strengthens both network reliability and responsiveness.

Simulations in Dammam and Doha produced consistent performance improvements. The integrated framework reduced total vehicular emissions by up to 40.99% and travel time by up to 32.05% in Dammam, and up to 48.78% and 43.92% in Doha, relative to the baseline Vehicle-to-Vehicle configuration. These outcomes suggest that combining fixed and aerial communication layers can improve congestion management and reduce environmental impact. Although latency increased from 0.01 s to 0.37 s and CPU usage from 50.36% to 76.93%, the operational gains outweighed these computational costs, supporting feasibility for real-time applications.

The results provide evidence that the proposed IoT–IoD integration can serve as a practical model for smart-city traffic management, balancing mobility efficiency with environmental benefits. The study achieved its stated objectives of reducing congestion through faster detection and response, and lowering vehicular emissions to improve air quality. While safety was not explicitly measured, the observed improvements in traffic flow imply indirect safety benefits.

Limitations

Several limitations are acknowledged:

• Vehicle behavior was modeled uniformly, without capturing driver variability or vehicle heterogeneity in detail. This simplification may reduce the spread of travel times and emission variability observed in the results, though it provides a consistent baseline for comparison across scenarios.

• Communication links were assumed reliable, without explicitly modeling issues such as packet loss. While this assumption isolates system performance, it may slightly overestimate effective communication coverage relative to real-world wireless conditions.

• Drone operations were idealized, excluding regulatory and energy-related constraints. This approach emphasizes algorithmic evaluation, but real deployments would need to account for endurance limits and airspace regulations that could influence sustained coverage.

Future work

Future research can extend this study by:

• i.incorporating heterogeneous vehicle models and more realistic driver behavior to reflect diverse traffic dynamics,
• ii.testing the framework under varied traffic densities and longer durations to assess scalability and robustness,
• iii.improving the communication layer through secure, interference-resilient, and adaptive networking protocols, and
• iv.applying machine-learning-based algorithms for autonomous drone repositioning and real-time route optimization.

In conclusion, the findings support the view that coordinated IoT and IoD integration provides a viable foundation for next-generation intelligent transportation systems, promoting cleaner, safer, and more efficient urban mobility.

Author contributions

A.Y. and T.R.S. conceptualized the study and designed the research methodology. A.Y. implemented the simulation environment, performed the experiments, and analyzed the data. T.R.S. supervised the project, guided the development of the optimization and rerouting algorithms, and provided critical feedback throughout the research. A.M. and M.I. contributed to refining the methodology, improving the experimental design, and interpreting the results. A.Y. drafted the initial manuscript. T.R.S., A.M., and M.I. reviewed and edited the manuscript critically for important intellectual content.

Funding

This research was supported by Project Number INML 2520 under the Interdisciplinary Center of Smart Mobility and Logistics at King Fahd University of Petroleum and Minerals.

Data availability

The data generated and/or analysed during the current study are not publicly available but are available from the corresponding author on reasonable request.

Declarations

Competing Interests

The authors declare no competing interests.