⚡ electrical engineering

Vehicle-to-Infrastructure Connectivity Improvement Based on Heterogeneous Networks in Intelligent Transportation Systems

This paper proposes a quality-of-service-aware framework for selecting roadside units in heterogeneous vehicular networks that prioritizes dwell time and connection stability over instantaneous signal strength, thereby significantly reducing handover frequency and improving continuity in dynamic environments despite a moderate trade-off in throughput.

Original authors: Hakimeh Azimi, Vahid sattari Naeini, Omid Abedi

Published 2026-11-02

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Hakimeh Azimi, Vahid sattari Naeini, Omid Abedi

Original paper licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine driving through a bustling city where your car is constantly talking to the streetlights, traffic cameras, and other vehicles. This conversation is the backbone of the future of transportation, a system designed to prevent accidents, ease congestion, and guide autonomous cars safely through complex environments. For this digital dialogue to work, your car must stay connected to the roadside equipment, known as roadside units, that act as the local internet hubs. However, because cars move so quickly and the city is full of obstacles, these connections are fragile. As a vehicle speeds past a hub, the signal can drop, forcing the car to frantically search for a new connection before the conversation is lost. This constant switching, or "handover," creates gaps in communication that can be dangerous for safety alerts and frustrating for passengers trying to stream video. The challenge for engineers is not just finding the strongest signal at any single moment, but finding the connection that will last long enough to get the job done.

Researchers at Shahid Bahonar University of Kerman have tackled this problem by proposing a new way for cars to choose which roadside unit to connect with. Instead of simply grabbing the nearest or strongest signal available, their method asks a deeper question: how long will this connection last, and how much data can we actually send before the car drives out of range? In their study, the team created a framework that looks ahead, estimating the time a vehicle will spend within the coverage area of different potential hubs. They then weigh this predicted duration against the speed of the connection and the delay involved in sending information. The goal is to select a partner that offers a stable, long-lasting link, even if it isn't the absolute strongest one at the exact moment of connection. This approach is particularly important in a "heterogeneous" network, where a car might have access to several different types of technology at once, such as short-range Wi-Fi and long-range cellular networks, and must decide which one to use for the specific task at hand.

To test their idea, the researchers built a detailed computer simulation of a city grid, similar to the layout of Manhattan, populated with two hundred moving vehicles. They programmed these cars to interact with both fixed roadside units and mobile ones mounted on other vehicles, creating a dynamic and realistic environment. They compared their new, forward-looking selection method against three traditional strategies that cars currently use. The first strategy simply connects to the closest unit, the second picks the one expected to be in range the longest, and the third uses a basic mix of speed and time. The results showed a clear trade-off. The traditional methods, especially the one that picks the closest unit, achieved higher peak speeds and lower delays when the connection was perfect. However, they suffered from a major flaw: the cars were constantly switching connections, sometimes hundreds of times during a single trip. This "ping-pong" effect caused frequent interruptions and made the overall data flow erratic.

In contrast, the new method proposed by the researchers dramatically reduced the number of these disruptive switches. In the simulation, the traditional approaches triggered between 685 and 750 handovers, while the new method reduced that number to fewer than 100. This represents a reduction of more than 85 percent in connection changes. By holding onto a single, stable connection for longer periods, the system ensured that data could flow continuously without the breaks caused by searching for a new signal. While the total amount of data sent in a single session was slightly lower than the aggressive, fast-switching methods, the data that was sent arrived in a much smoother, more predictable stream. The researchers found that for applications like video streaming or sending large files, this stability is far more valuable than a momentary burst of high speed that is immediately followed by a disconnection.

The study also highlighted that not all data is the same, and the best connection depends on what is being sent. Safety messages, which need to be delivered instantly, have different needs than a passenger watching a movie. The researchers' system is smart enough to adjust its priorities based on the type of traffic. For a safety alert, it prioritizes low delay, ensuring the warning arrives quickly. For multimedia content, it prioritizes the total amount of data that can be transferred during the time the car is in range. This flexibility allows the network to serve different needs simultaneously without forcing every vehicle to use the same rigid rules. The simulation confirmed that by considering the time a car will stay in range, the system can avoid the instability that plagues current networks, making the digital infrastructure of a smart city more reliable.

Ultimately, this research suggests that the future of connected driving lies in patience and planning rather than speed and reaction. By teaching vehicles to evaluate the longevity of a connection before committing to it, the system can maintain a steady line of communication even in the most chaotic urban environments. The findings, derived entirely from computer simulations, indicate that a moderate sacrifice in peak speed is a small price to pay for the significant gain in connection stability. As cities become denser and traffic more complex, this ability to maintain a continuous, unbroken link will be essential for the safe and efficient operation of intelligent transportation systems. The work provides a blueprint for how vehicles can navigate the wireless landscape not just as fast-moving objects, but as thoughtful participants in a network that values continuity above all else.

Technical Summary: Vehicle-to-Infrastructure Connectivity Improvement Based on Heterogeneous Networks in Intelligent Transportation Systems

Problem Statement
In heterogeneous Vehicular Ad Hoc Networks (HetVNETs), maintaining stable connectivity between vehicles and Roadside Units (RSUs) is a critical challenge, particularly in high-mobility urban environments. Existing association methods often rely on greedy criteria, such as instantaneous signal strength or maximum data rate. While these approaches may maximize short-term throughput, they frequently result in unstable connections, excessive handovers (the "ping-pong" effect), and service interruptions as vehicles rapidly move out of coverage areas. The authors identify a need for an association mechanism that prioritizes connection continuity and service stability over transient performance peaks, specifically addressing the trade-offs between dwell time, communication delay, and data transmission capacity in networks comprising both fixed and mobile RSUs.

Methodology
The paper proposes a Quality-of-Service (QoS)-aware RSU association framework designed to select the most suitable RSU for a vehicle based on predicted mobility and service requirements. The methodology is structured as follows:

Mobility and Dwell Time Estimation:
The framework utilizes a Manhattan mobility model where vehicle paths are predictable. It calculates the "dwell time" ( $T_{dwell}$ )—the duration a vehicle remains within the effective communication radius of a candidate RSU (fixed or mobile). This is derived by solving a quadratic equation based on the relative position and velocity vectors of the vehicle and the RSU.
Performance Modeling:
The system models three key metrics for each candidate RSU:
- Transferable Data ( $D_i$ ): The maximum data volume that can be transmitted during the predicted dwell time, calculated as the product of the nominal link rate and the effective connection time (accounting for connection setup overhead).
- Communication Delay ( $T_{delay}$ ): A composite metric including transmission delay (data size divided by link rate) and processing delay (data size divided by RSU edge computing capacity).
- Effective Link Rate ( $r_i$ ): The application-layer throughput, distinct from the nominal PHY-layer rate.
Multi-Criteria Selection Algorithm:
The selection process employs a two-stage filtering and ranking mechanism:
- Feasibility Filtering: RSUs are first filtered based on physical constraints (sufficient dwell time to complete a handover and capacity to transfer the required data size). A secondary filter applies strict QoS constraints (minimum rate and maximum delay) specific to the traffic class (e.g., safety, voice, multimedia). If no RSU meets strict QoS, the system falls back to the physically feasible set to prioritize connectivity.
- Weighted Utility Scoring: For the remaining candidates, a weighted-sum utility function is computed. The score combines normalized values of delay, link rate, and transferable data. Weights are dynamically assigned based on the traffic class (e.g., safety traffic prioritizes low delay, while multimedia prioritizes rate and dwell capacity).
- Hysteresis Control: To prevent unnecessary handovers, a handover is triggered only if the score improvement of a new candidate exceeds a specific hysteresis margin ( $H^{(c)}$ ) relative to the current RSU.

Key Contributions

Context-Aware Association: The paper introduces a selection mechanism that moves beyond instantaneous signal strength, integrating predicted dwell time, application-layer throughput, and processing delay into a unified decision framework.
Hybrid RSU Support: The framework is designed to operate in heterogeneous environments containing both static infrastructure and mobile RSUs (e.g., on buses or emergency vehicles), dynamically ranking them based on spatiotemporal factors.
Traffic-Class Adaptability: The algorithm adapts its selection criteria and weighting coefficients based on the specific QoS requirements of different data streams (safety, voice, text, multimedia).
Stability-Oriented Design: By explicitly modeling connection duration and applying hysteresis, the method aims to minimize handover frequency and signaling overhead, which are critical for maintaining session continuity in dynamic traffic.

Results
The proposed framework was evaluated using a simulation environment combining SUMO (for traffic mobility) and ns-3 (for network communication) in a 1000m x 1000m Manhattan grid with 200 vehicles. The results were compared against three baselines: Nearest RSU (distance-based), Maximum Stay-Time (dwell-time based), and a Rate-Dwell (RD) heuristic.

Handover Reduction: The most significant result is a reduction in the total number of handovers by more than 85% compared to the baseline strategies (reducing events from ~700 to <100).
Connection Stability: The proposed method achieved significantly longer connection durations. While baselines saw nearly 80% of connections lasting less than 100 seconds, the proposed scheme maintained a substantial fraction of connections beyond 400–600 seconds.
Throughput and Delay Trade-off: The study acknowledges a trade-off: the proposed method yields lower instantaneous throughput (averaging ~8 Mbps vs. 15–20 Mbps for greedy baselines) and higher end-to-end delay (approx. 3–5 seconds vs. <2.6 seconds). However, the data transfer per connection was approximately four times higher than the baselines, indicating more efficient use of established sessions.
Predictability: The cumulative distribution functions (CDFs) showed that while the proposed method sacrifices peak performance, it offers a more predictable and stable data flow, avoiding the volatility associated with frequent handovers.

Significance and Claims
The authors claim that their approach offers a robust solution for dense and highly dynamic vehicular environments where connection continuity is often more valuable than maximizing short-term data rates. By reducing the frequency of handovers, the framework minimizes signaling overhead and the risk of connection drops, thereby supporting time-sensitive services and autonomous driving applications that require reliable links. The paper concludes that while there is a moderate trade-off in throughput and latency, the overall improvement in connection stability makes the method better suited for sustainable V2I communication in intelligent transportation systems. Future work is identified as integrating machine learning to further refine trajectory and dwell time predictions in uncertain urban environments.

More like this