Simulation framework for connected vehicles: a scoping review

Introduction

In recent decades, a significant increase in vehicle use has increased traffic congestion and fatalities¹. According to the World Health Organization, 1.25 million people are killed and severely injured involving vehicle accidents². Hence, connected vehicle technology responds to this constraint, aiming to leverage inter-vehicle communication to produce safe, user-friendly, and fuel-efficient vehicle assistive technologies^3,4. One of the main aspects of connected vehicle research is to optimise traffic flow through the exchange of information⁵. This communication can be sorted in terms of vehicle (V2V), infrastructure (V2I), a pedestrian (V2P), and network (V2N)^6,7. The exchange of information, collectively known as V2X communications, could assist drivers in preventing accidents by providing warnings of danger invisible to drivers and other sensors (e.g. collision avoidance, lane departure warning and speed limit alert)^8,9.

Nevertheless, the adoption of connected vehicle technology poses a range of challenges, particularly in urban environments. It is challenging to analyse the effectiveness of the application of connected vehicles under traffic conditions^10–12. As such, simulations using traffic and network simulators as well as mobility generators are viable alternatives to modelling and determining the effectiveness of such deployments in the real world^13,14, as it provides an affordable and scalable method for analysing model compliance in various contexts and parameters.

Traffic simulations are categorised by level of detail into three separate categories¹⁵. First, the most precise information on each vehicle in the system is microscopic simulations¹⁶. Second, mesoscopic simulations exploit aggregate velocity-density functions to represent their behaviour and view traffic as a continuous stream of vehicles¹⁷. Finally, macroscopic simulation is the large-scale traffic model, which focuses on combined traffic status¹⁸. Microscopic simulations provide the highest degree of detail for modelling, although they are the slowest to execute^19,20.

In addition, mobility generators are a possible option for modelling vehicle elements such as traffic, temporal and spatial mobility, and generating mobility traces^21,22. These traces are then uploaded to a network simulator, which mimics vehicle-to-vehicle communication. Furthermore, these traces can be generated by observing real-world vehicles on the road and then used in network simulations^23,24. The effect of network parameter modifications on traffic mobility is a strategic objective simulation²⁵. It is also restricted to the use of the trace controlled by the mobility model. Another option is to use a simulator that directly integrates the mobility framework.

This study focuses on the Vehicular Adhoc Networks (VANET) which relies on network protocols to assess their performance^22,23, given that actual experiments are not possible. Over the last decade, efforts have been made to produce a full transport simulator for VANET solutions, including a wireless network simulator for modelling and evaluation^24,25. A wide range of simulators can be used for VANET simulation modelling, both commercial and open source. Older simulators provide a network simulator to communicate with stationary mobility models. Many researchers have examined various mobility models with simulation tools for several contexts. Such simulator tools are not yet well explored since many researchers base their simulations depending on their use case settings. Thus, this motivates the identification of different simulators do not yet exist. Therefore, this study conducted a systematic scoping review to identify the applicability and availability of existing mobility generators, network simulators, and combination simulators.

The rest of the paper is organised as follows. The following section explains the methods used for analysing existing mobility generators and network simulators. Further, the outcome of the review scoping is depicted with several discussions. The final section concludes by considering future directions.

Methods

This work involves identifying the research question, identifying relevant studies, selecting the studies, charting the data, collating, summarising, and reporting results. The review was carried out in compliance with the PRISMA Extension for Scoping Review²⁶.

Results

The initial search turned up 269 matches. After removing duplicates, a total of 184 titles and abstracts were screened, from which 72 publications were subjected to full-text review. 10 studies fulfilled the criteria for inclusion and were included in the analysis (see Figure 1 for the PRISMA Flow Diagram).

Figure 1. PRISMA Flow Diagram.

We found that open-source mobility and network simulators were popular among researchers. Microscopic models were preferable for research related to vehicular communications since the simulations provide the most precise information of each vehicle or mobile node and the highest degree of detail for modelling compared to macroscopic and mesoscopic models. Common network simulators were NS-2, Ns-3 and OMNeT++. However, not all mobility simulators supported active development, which is important in current active research domains such as vehicular communications. The mobility generators and simulators available after 2015 are further shown in Table 1 and Table 2, respectively. Besides, summarised previous studies are shown in Table 3 of using mobility generators or simulators.

Discussion

Since this area of study is considered as a relatively new but rapidly growing field, this scoping review process only considers relevant papers published from 2015 onwards, which shows that extensive research has been conducted to create security standards for communication technologies, particularly the vehicular network. Although various simulators can be enhanced with library extensions, none of the simulators is related to security and privacy. Ultimately, researchers and professionals cannot compare their security measures to a given circumstance. For instance, ensuring the privacy of a vehicular user in a fast-moving network and disseminating messages in a secure vehicular environment. However, there is no simple practice of extending existing simulators to the desired security standard, which implies that future development research will need to be done.

In addition, the quality of a simulation depends largely on the precision of the models. The range of precision has increased dramatically recently, where several modules contain signal attenuation components, multiple antenna models, and environmental interferences. However, one continuous barrier to producing accurate simulations is the evolution of rapid prototyping and its increasing use in-vehicle networks. For example, vehicle nodes would depend on three-dimensional scenarios to communicate with other nodes. It would be crucial for current and future simulators to extend the current simulators to these new conditions.

Apart from that, integration with real-time system modelling based on non-real-time events creates additional challenges. Due to resource limitations, current simulators do not correspond with the physical properties of the hardware prototype while simulating a comprehensive network with multiple vehicles. Several alternatives have been put forward to reduce the complexity that could speed the simulation. However, this approach usually does not include indirect outcomes, which could seriously impact the behavior of real-world network components. It is, therefore, necessary to examine the interconnection between simulators and hardware devices with the security standards concerned.

Conclusions

Studies have led to the discovery of comprehensive and realistic simulation tools due to the increasing popularity and interest for the future transportation system. This work has examined the current availability of simulators. While testing VANET with essential performance, it is necessary to deploy a mobility generator and mobility network that accurately represent real vehicle traffic. Based on our comparative identification, NS-3 and SUMO has been the optimal choice for real-time VANET modelling. Although several simulators have many features, it is worth exploring further the improvement of the simulators for specific scenarios. In addition, this work can be further be expanded in future by investigating relationships of appropriate simulator for a V2V or V2X application to different scenarios and protocols. We plan to study the used simulators in this context and the extent of benefit and development achieved using the simulators.