Analysis of the Impact of Vehicular Communication Technologies in Intelligent Transportation Systems

Mendes et al. [12] presented an analysis of the advantages and drawbacks of V2X communications. On the upside, the implementation of V2X technology using the 5G network has been built upon a strong framework of regulatory requirements set forth through the global 3rd Generation Partnership Project (3GPP) standards and the local regulatory requirements of spectrum usage in each country where V2X technology will be deployed. The use of New Radio Unlicensed (NRU) spectrum has allowed for the establishment of dedicated private networks solely to support V2X operations and perform a variety of other types of vehicular services. Global Navigation Satellite Systems (GNSS) such as GPS and Galileo provide the high-accuracy location services needed for the proper operation of autonomous driving and cooperative maneuvers between vehicles and infrastructure by allowing vehicles to securely share their positioning information with each other and their environments. However, there are challenges associated with deploying V2X technologies including limited availability of public spectrum opportunities, which could result in monopolization of the spectrum and diminished com-petition among manufacturers of V2X devices; complexities in managing the different channels established for safety vs. non-safety communications; privacy implications due to the independent and uncoordinated nature of GNSS-based positioning systems that do not adequately safeguard the security and privacy of data exchanged during V2X operations; and a substantial gap in research regarding precision positioning services in V2X-focused private 5G networks, indicating an urgent requirement for additional studies to adequately exploit these technologies for the benefit of V2X operations.

Liu et al. [13] described the advantages of adopting AI-V2X technologies in conjunction with large amounts of data and computational power/analysis to develop drivers’ awareness and forecasts, while increasing both drivers’ comfort level and their safety. These systems utilize AI technology to connect the digital world and the real world while providing both network and vehicle intelligence. In addition, Liu et al. detail how the 5G technology will affect the Smart City and its Impact on Transportation Networks, along with the technological, economic, and legal implications that will accompany the implementation of 5G. The authors suggest that transportation metrics could improve significantly post-5G deployment, citing potential improvements in overall transportation efficiency (84.2%), vehicle image monitoring (88.2%), V2X communication development (85.36%), driving comfort (82.15%), and a reduction in roadway congestion (91.84%).

Wang et al. [14] demonstrated technologies that provide accurate control of vehicle acceleration, braking, and steering to improve operator safety. Cellular Vehicle-to-Everything (C-V2X) connected vehicles can establish real-time operations and enhance safety through specific cellular communication methods. However, obstacles such as signal fading, time delays, packet loss, and malicious attacks can limit vehicle interaction in Cooperative Intelligent Transportation System (C-ITS) environments. The study examines improvements in C-V2X technology regarding communication latency and efficiency compared to Dedicated Short-Range Communications (DSRC). Simulation results using Simulation of Urban Mobility (SUMO) indicated that C-V2X vehicles experience a reduction of more than 99% in communication delay compared to DSRC vehicles. Furthermore, a 38% reduction in accident rates was observed with 60% AV saturation, attributed primarily to reduced latency. The authors found that C-V2X offers benefits, but also challenges, such as data packet loss and system vulnerabilities must be addressed to fully realize the potential of C-ITS.

Elassy et al. [8] studied the key components of ITS, like dedicated vehicle networks, intelligent traffic signals, virtual traffic signals, and traffic prediction, focusing on how to enhance transportation efficiency, safety, and sustainability. Also, the researchers tested several other problems related to safety and privacy in public transportation systems and provided many case studies to explain how ITS technology can be integrated into large urban areas, resulting in a sustainable smart city.

Many studies are addressing ITS and vehicle communication technologies. Some works have focused on general communication architectures between vehicles and infrastructure, while others have explored the application of AI and edge computing technologies to improve traffic management. A third group has been concerned with evaluating advanced communication technologies such as 5G and C-V2X, or developing simulation frameworks for collaborative transportation systems. Given this fragmentation of research directions and the lack of a unified evaluation that integrates these approaches within a comprehensive comparative framework, previous studies have been organized in a comparative table. This table illustrates their classification according to their main research themes, the technologies used, their contributions, and their shared limitations. To address this, Table 1 presents a comparative summary of relevant studies, highlighting the research gap addressed by this work: a unified evaluation framework combining 4G, 5G, and V2X technologies with AI and edge computing integration within a realistic simulation environment.

Table 1. Comparative summary of related works highlighting technological focus, application domains, and existing research limitations in vehicular communication for Intelligent Transportation Systems (ITS)

Figure 1 provides an overview of four performance metrics, namely, latency, packet loss, navigation accuracy, and throughput for our vehicle study for 25 simulation scenarios defined by the x-axis. The sub-figures show the important data trends and how communication technologies (4G, 5G, V2X) and environmental variables (network load, interference) affect performance.

4.1 Latency

All scenarios show that the latency was unstable, where the lowest latency (5–10 ms) was achieved in 5G and V2X configuration with minimal interference. The performance of 5G and V2X is higher than 4G in time-sensitive applications (e.g., collision avoidance), while less than 20 ms latency is critical. A spike (~25 ms) appears in scenario 18 because there is high network congestion during the simulation (Figure 1(a)).

4.2 Packet loss

In high interference (e.g., Scenarios 7 and 19), the packet loss rate is about 15%, but in other scenarios it is below 5%. These results reflect that V2X gives more reliability in city traffic; on the other side 4G is more susceptible to interference. Also, a "sweet spot" appears at (Scenarios 3–12) where the packet loss rate in 5G is near zero, therefore it is comfortable for safety-critical messaging (Figure 1(b)).

4.3 Navigation accuracy

The relationship between the error rate and network loads is linear, so when the error rate decreases, the network loads increase. The result shows that the error rate was from 2 m (ideal case) to 12 m (worst case), where high packet loss occurred due to positioning inaccuracy. This shows the need for robust communication in dense traffic. The most important results are that V2X has sub-5 m accuracy even in high-load scenarios (scenarios 20–25); therefore, it is useful for autonomous navigation (Figure 1(c)).

4.4 Throughput

In 5G low load scenarios, the throughput is about 150 Mbps; on the other side it drops to 20 Mbps in high interference V2X cases. It is clear that 5G is useful in bandwidth-heavy applications (e.g., infotainment), while for low-data safety messages, V2X provides stable throughput (Figure 1(d)).

4.5 Quantitative comparison of technologies

We evaluate several key metrics related to communication technologies in the context of V2X communication, including path optimality compared to ground truth, computation time to assess scalability with network size, and packet delivery ratio.

The performance comparison of communication technologies is summarized in Table 3. The average latency for 4G is 45.2 ms, while 5G shows a significant improvement at 12.7 ms. V2X can provide nearly real-time responsiveness at an average of just 8.3 ms of latency. In terms of packet loss, 4G has an average of 9.5%, while 5G only has a 2.1% loss, and V2X's average packet loss is 4.7%. For navigation accuracy, 4G has an accuracy of 8.9 meters, 5G improves this to 3.2 meters, and V2X enables precise positioning with an accuracy of 2.5 meters. Throughput is another critical metric: 4G supports 35.1 Mbps, 5G significantly outperforms it at 120.5 Mbps, while V2X provides a throughput of 55.6 Mbps.

Table 3. Performance comparison of communication technologies

The most significant findings from the comparisons show that in terms of throughput, 5G offers a substantially greater maximum throughput than 4G, which can be beneficial for applications with high amounts of data, such as HD Map Updates and infotainment services. Because V2X transmits via 5G Networks, it experiences lower latency and reduced positional error, both of which are critical for collision avoidance systems. Furthermore, 4G may have benefits during "off-peak" times; however, it has higher packet loss when operating within areas of high density, therefore representing and supporting both 5G and V2X for different respective applications/deployments. It is recommended to implement hybrid deployments of 5G and V2X to take advantage of both systems. Figure 2 presents the overall variation in latency, packet loss, navigation accuracy, and network congestion across the 25 simulation scenarios. It further illustrates how increasing traffic load and interference affect the communication performance of the evaluated vehicular networks.

Figure 2. Network performance metrics across 25 simulation runs, showing latency, packet loss, navigation accuracy, and congestion

4.6 Proposed AI Hybrid Adaptive Switching and Routing Algorithm

To maximize the benefits of 5G and V2X technologies and ensure stable connectivity in multi-hop mobile networks, we propose a two-stage decision-making algorithm that integrates the proposed Reinforcement Learning (RL)-switching mechanism with the Dijkstra algorithm to reach the optimal path. The overall workflow of the proposed AI Hybrid Adaptive Switching and Routing Algorithm is illustrated in Figure 3. The framework first selects the appropriate communication interface through the RL-based switching mechanism and then determines the optimal transmission path using the modified Dijkstra algorithm.

Figure 3. Flowchart of the proposed Hybrid Adaptive Switching and Routing Algorithm

4.6.1 Algorithm design and mathematical expression

The RL-based decision mechanism performs two sequential operations to ensure adaptive and context-aware communication selection under highly dynamic vehicular environments:

Phase 1: Interface RL decides (5G or V2X):

The system continuously monitors two key parameters: Application Criticality ($C_{a p p}$) and Available Bandwidth ($B_{\text {avail}}$) by using an RL agent to dynamically decide between 5G and V2X interfaces. The RL framework is defined as follows:

1. State Space (s): The agent continuously finds environment parameters: Application Criticality ($C_{a p p}$), Available Bandwidth ($B_{\text {avail}}$) and Inter-vehicle Distance ($D_{i v}$), and Signal-to-Noise Ratio (SNR).
2. Action Space (A): Based on the previous step, the agent selects the mode: (5G (network mode) or V2X (Sidelink)).
3. Q-Learning Update Rule:

The Q-values are iteratively updated using the Bellman optimality equation:

$Q\left(s_t, a_t\right) \leftarrow Q\left(s_t, a_t\right)+\eta\left[r_t+\gamma \max _{a \in A} Q\left(s_{t+1}, a\right)-Q\left(s_t, a_t\right)\right]$          (1)

where,

$\eta$: learning rate

$\gamma$: discount factor

$r_t$: immediate reward

$Q\left(s_t, a_t\right)$: action-value function

An $\varepsilon$-greedy exploration policy is used to balance exploration and exploitation.

4. Reward Function (R):

• Condition 1 (Safety Critical Reward): If $C_{a p p}$ is high (e.g., emergency braking or collision warning) and ($D_{i v}$) < Threshold ($T_{d i s t}$), a high positive reward is received by the agent to minimize latency; therefore, the agent selects V2X (latency = ~8.3 ms as shown in Table 2).
• Condition 2 (Bandwidth Intensive Reward): If $C_{a p p}$ is low (e.g., media streaming) and $B_{\text {avail }}$ > $T_{b w}$, a low positive reward is received to the agent to maximize throughput; therefore, the agent selects 5G (~120 Mbps as shown in Table 2).
• Condition 3 (Fallback): If (SNR) < Threshold for the active interface, a negative reward is received by the agent, and the algorithm initiates a handover to the secondary interface.

Phase 2: Optimal Path Selection (Dijkstra Integration):

After the RL identifies the appropriate interface, the optimal path to the destination must be determined, especially in VANET (Vehicle Wireless Networks), where multiple media connections may be required. To achieve this, we integrate a Modified Dijkstra's Algorithm with the proposed RL-based strategy. Instead of relying solely on a fixed distance, the algorithm dynamically calculates the edge weights based on real-time metrics. The edge cost function is defined as the weighted sum of latency and inverse throughput, effectively guiding safety-critical packets ($C_{a p p}$ = High) via the lowest-latency path (potentially V2X), while data-intensive applications ($C_{a p p}$ = Low) are routed via the highest-throughput path (potentially 5G). Table 4 shows the detailed simulation parameters for ns-3 environment.

Table 4. Detailed simulation parameters for ns-3 environment for the proposed algorithm

• Mathematical Expression:

1. Proposed Edge Cost Function: The cost of each edge (u,v,s_t) is defined as:

$w\left(u, v, s_t\right)=\alpha\left(s_t\right) \cdot L_{u v}+\beta\left(s_t\right) \cdot \frac{1}{T_{u v}}$          (2)

where,

• $L_{u v}$: latency of the link
• $T_{u v}$: throughput of the link
• $\alpha\left(s_t\right), \beta\left(s_t\right)$: dynamic weighting coefficients based on RL

2. Adaptive Behavior:

• For safety-critical applications ($C_{a p p}=High$):

$\alpha\left(s_t\right)>\beta\left(s_t\right)$

prioritizes minimum latency paths (typically V2X)

• For data-intensive applications ($C_{a p p}=Low$):

$\beta\left(s_t\right)>\alpha\left(s_t\right)$

prioritizes maximum throughput paths (typically 5G).

4.6.2 Impact of adaptive switching on performance metrics

To evaluate the efficacy of the proposed two-phase algorithm (Interface Switching Logic and Modified Dijkstra Integration), the system performance was compared against standalone technologies, namely Pure 5G and Pure V2X. The simulation results are summarized in Table 4. The simulation considers a traffic mix of 40% safety-critical and 60% bandwidth-intensive applications to reflect realistic urban driving conditions. The Q-learning agent was trained over 100 simulation episodes. Demonstrate that the Hybrid (Adaptive) approach successfully optimizes the trade-offs between latency and throughput, yielding superior overall performance.

1. Average Latency Analysis

The primary objective of Condition 1 in the switching logic is to prioritize safety-critical applications by leveraging V2X Sidelink. As shown in Table 4, the Pure 5G network exhibits an average latency of 12.7 ms, which is suboptimal for emergency braking or collision warnings. In contrast, the Hybrid (Adaptive) system achieved a significantly lower latency of 9.7 ms. This value is substantially better than 5G and also slightly closer to the Pure V2X baseline (8.3 ms). This improvement is attributed to the Modified Dijkstra's Algorithm employed in Phase 2, which dynamically calculates edge weights based on real-time latency metrics. By routing critical packets through the path with the lowest calculated cost—prioritizing V2X when $C_{a p p}$ is high—the system minimizes transmission delays more effectively than a static V2X connection alone.

2. Throughput Performance

For bandwidth-intensive applications, Condition 2 dictates a switch to 5G Network Mode to utilize its high throughput capabilities. The simulation results highlight the limitations of relying solely on V2X for data-heavy tasks, as Pure V2X recorded a throughput of only 55.6 Mbps. Conversely, Pure 5G achieved the highest throughput at 120 Mbps. The Hybrid (Adaptive) system closely matches this 5G performance, recording a throughput of 119 Mbps. This performance confirms that the adaptive logic successfully identifies and routes high-volume data (e.g., HD map updates) via the 5G interface. The Modified Dijkstra algorithm supports this by incorporating "inverse throughput" into the edge cost function, ensuring that data-intensive packets are guided along the highest-capacity paths available.

3. Navigation Accuracy

Navigation accuracy in VANETs is contingent upon receiving timely position updates (low latency) and precise map data (high throughput). The hybrid system achieves a navigation accuracy of 2.4 m, which is better than both Pure 5G (3.2 m) and Pure V2X (2.5 m) (Table 5). This enhancement is due to the algorithm's balanced approach: it keeps the ultra-low latency of V2X for real-time localization beacons while using the 5G bandwidth to obtain detailed environmental data. Furthermore, Condition 3 (Fallback) ensures continuity by measuring SNR; if the performance of the active interface deteriorates, the packet loss is prevented by the handover mechanism, which could otherwise lead to deterioration of localization precision.

Table 5. Performance comparison (hybrid vs. standalone)