Car-to-Pedestrian communication with MEC-support for adaptive safety of Vulnerable Road Users
Introduction
In recent years, cooperative collision detection has emerged as a complementary approach to traditional safety solutions using vehicle sensors like cameras, laser scanner, and RADAR to detect nearby objects. Different from the sensor-based approach, which requires a direct Line-Of-Sight (LOS) between vehicles and detected objects, cooperative collision detection can even operate in Non-Line-Of-Sight (NLOS) scenarios by exchanging movement information between entities via wireless communication. In the context of safety systems for Vulnerable Road Users (VRUs) like pedestrians and bicyclists, vehicles exchange information with VRUs’ mobile smart devices, e.g., smartphones, tablets, or wearable devices [1], [2], [3], [4].
The general concept of a VRU safety system is shown in Fig. 1 (extending our earlier work in [5]), in which vehicles and VRUs have two possible options for data transmission, namely direct Device-to-Device (D2D) communication and an infrastructure-based approach. As an initial idea, D2D technologies, such as Wi-Fi Peer-to-Peer (P2P) or Dedicated Short Range Communication (DSRC), are designed to serve the need of cooperative safety systems, which allows messages to be directly transferred from sender to receiver. In particular, DSRC has been assumed for Car-to-Car (C2C)/Vehicle-to-Vehicle (V2V) and Car-to-Everything (C2X)/Vehicle-to-Everything (V2X) communication and is likely to be available for smartphones in the next few years [6], [7]. However, these D2D technologies do not offer good coverage and range, i.e., less than 200 m for Wi-Fi Direct and less than 1000 m for DSRC, especially around intersections in urban areas [8], [9]. To overcome these limitations, cellular technology is proposed as an alternative.
Cellular V2X (C-V2X), or more specific LTE-V2X, includes two interfaces: the LTE interface (named Uu), that supports the communication between vehicles/end-devices and mobile base stations and the newer D2D interface (named PC5), that enables V2V communications based on direct LTE sidelink [10], [11]. C-V2X is indicated to have many advantages in different aspects, such as communication range, performance, and reliability, especially with the current evolution from LTE to 5G [12]. PC5, like DSRC, has been studied by several existing works and is deployed mainly for vehicles [10], [11], [13]. This direct V2X communication can significantly reduce the latency and is expected to become superior to DSRC [12]. Meanwhile, the Uu radio interface, which is already available for smartphones, still remains an important part of the next generation wireless V2X system with the ability to support long-range communication. However, the feasibility of applying this interface and network infrastructure to provide an additional approach for different V2X use cases is still unclear. Recognizing this gap along with the desire to leverage existing networks for V2X communications, we focus on studying the performance of cellular-based systems over the Uu interface and exploiting this approach for VRU safety applications/services. To be more specific, Car-to-Pedestrian (Car2P), a typical case of VRU systems, in which the safety for pedestrians is the main objective of our investigation in this paper.
In order to enable collision detection, pedestrians’ smartphones and vehicles need to send beacon messages periodically to each other and after obtaining the beacons, a Collision Detection Algorithm (CDA) is executed. While these tasks are not a big deal for vehicles, they could be a significant obstacle when running on smartphones with limited resources. Especially, it would be a heavy burden on smartphones in terms of energy consumption and executing time when more information from other vehicles is needed and a more complex CDA is used. In this paper, we concentrate on addressing this bottleneck by considering the use of computation offloading using Multi-access Edge Computing (MEC). Applying this mechanism for smartphone applications, such as gaming, image/video processing, object/face recognition, or accelerated web browser has been shown to have a noticeable improvement in energy efficiency [14]; however, its applicability in the context of Car2P safety systems is still unclear.
The original idea of our approach was presented in our recent work [15], in which we proposed a dynamic selection of execution location (local or remote) for calculating pedestrian context information. Conceptually, the beacons transmitted between objects comprise some fundamental information like current position, heading direction, and speed [16], [17], [18], [19]. Some recent works [3], [20] suggested using additional context information to improve the accuracy of prediction results. Looking in more details, we can see that it is quite simple to calculate some information like position or direction directly from smartphone sensor data [6], [21]. But for others, e.g., motion states or activity, more extensive data handling like pre-processing, extraction of features, and training machine learning models are needed [20], [22], [23]. Depending on the amount of raw sensor data collected, an appropriate computational scheme is selected to save energy on smartphones.
In this paper, we extend and improve our proposed approach in [15], to support not only context information calculation but also the CDA. Generally, in cellular-based safety systems, context information computation and collision detection could be done either on a pedestrian smartphone itself or on a remote server. Deciding the strategy to not only prolong the battery life of the smartphone but also to meet the timing constraints in a Car2P safety system is a challenging task. It requires to take into consideration different parameters, such as energy consumption and processing time of local computation, network communication overhead, as well as real-time requirements of the system.
To provide insights on this adaptive approach, we carry out experimental and simulation studies on the performance of a Car2P system when applying both Local and Offload computational schemes on smartphones. In addition, energy consumption and latency of each scheme are investigated in the relationships with different parameters, such as the machine learning algorithm, the sensor sampling rate, the window size, and the sending interval of messages on the smartphone. We see the results of this paper as an important step towards energy-efficient VRU safety systems.
Our main contributions can be summarized as follows:
- •
We propose an adaptive computational approach for smartphones, which considers the possibility of offloading tasks in both data and service levels of Car2P safe applications.
- •
We measure and analyze the energy consumption and processing time of a lightweight machine learning application for determining pedestrian activities as well as a CDA for detecting potential collisions.
- •
We simulate a scenario of a Car2P system using the simulation framework Veins LTE to evaluate the end-to-end performance and scalability of our adaptive approach.
Section snippets
Related work
The concept of a cooperative Car2P safety system has been described in many previous works [1], [5], [24], [25]. In particular, cellular-based safety systems have received much attention over the last years [5], [8], [18], [26]. Additionally, the communication performance of cellular in vehicular networks has been studied and compared to other wireless technologies, such as IEEE 802.11p-based DSRC [9], [27], which demonstrates the great potential of this technology in safety applications.
In
System architecture
Basically, our proposed Car2P safety system relies on the (centralized) architecture depicted in Fig. 1. All communication within the system is established via LTE. Direct Car2P communication is beyond the scope of this paper. In the following, we explain our system architecture in which part of the information is processed locally and other parts are offloaded to the cloud supported by MEC.
Experimental study
In this section, we report on the experiments performed to evaluate the performance of our proposed adaptive approach in both data and service levels. As described in Section 3.2, energy consumption and processing time are the two main metrics, which directly affect the selection of operation schemes. In our experiments, we mainly focus on estimating the local computation costs as well as the energy consumed by network operations (transferring data to/from the server).
Simulation study
In this section, we evaluate the performance of our adaptive approach by means of network simulations. We conduct our study with the Veins LTE simulator [47], which is based on the popular Veins vehicular network simulation framework [48] building upon OMNeT++.2 The road network and the movement patterns of vehicles and pedestrians are generated using SUMO.3
Discussion
Based on the experimental and simulation results, we derived some very interesting insights, which will help to better optimize the Car2P safety systems in terms of energy efficiency for pedestrian smartphones and packet latency as well. As described in Section 3.2, context information calculation and CDA work together in a safety system for pedestrians as a two-phase process. The resulting context information of the first phase in the data level will be the input for the second phase, CDA, in
Conclusion
In this paper, we studied options for Car2P communication with MEC support for improving the safety of VRUs. Building upon established C2C communication principles, we extend this concept to also integrate pedestrians into the system. The challenge here is that all computation and communication needs to be performed by a smartphone carried by the user. This poses both energy and latency issues given that dedicated activity detection algorithms need to be executed and the results need to be
Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.
References (48)
- et al.
DSRC Versus 4G-LTE for connected vehicle applications: A study on field experiments of vehicular communication performance
J. Adv. Transp.
(2017) - et al.
Human activity recognition with smartphone sensors using deep learning neural networks
Expert Syst. Appl.
(2016) - et al.
A survey of adaptation techniques in computation offloading
J. Netw. Comput. Appl.
(2017) - et al.
Efficient data handling in vehicular micro clouds
Elsevier Ad Hoc Netw.
(2019) Fast effective rule induction
- et al.
Vehicle to pedestrian communications for protection of vulnerable road users
- et al.
Modeling cycling behavior to improve bicyclists’ safety at intersections – a networking perspective
- et al.
Improving smartphone based collision avoidance by using pedestrian context information
- et al.
Vehicle-to-pedestrian communication for vulnerable road users: Survey, design considerations, and challenges
Sensors
(2019) - et al.
CAR-2-X And pedestrian safety
IEEE Veh. Technol. Mag.
(2010)
Cars talk to phones: A DSRC based vehicle-pedestrian safety system
A case for leveraging 802.11p for direct phone-to-phone communications
A cluster based architecture for intersection collision avoidance using heterogeneous networks
LTE-V For sidelink 5G V2X vehicular communications: A new 5G technology for short-range vehicle-to-everything communications
IEEE Veh. Technol. Mag.
Vehicle-to-everything (V2X) services supported by LTE-based systems and 5G
IEEE Commun. Stand. Mag.
An assessment of LTE-V2X (PC5) and 802.11p direct communications technologies for improved road safety in the EU
LTE For vehicular networking: A survey
IEEE Commun. Mag.
Mobile edge computing: A survey on architecture and computation offloading
IEEE Commun. Surv. Tutor.
Adaptive safety context information for vulnerable road users with MEC support
Transmission control for reliable pedestrian-to-vehicle communication by using context of pedestrians
Performance and channel load evaluation for contextual pedestrian-to-vehicle transmissions
Cloud-based pedestrian road-safety with situation-adaptive energy-efficient communication
IEEE Intell. Transp. Syst. Mag.
Backwards compatible extension of CAMs/DENMs for improved bike safety on the road
5G / LTE based protection of vulnerable road users: Detection of crossing a curb
Cited by (39)
Mobility in pedestrian communication simulations: Impact of microscopic models and solutions for integration
2023, Computer CommunicationsOn safety design of vehicle for protection of vulnerable road users: A review
2023, Thin-Walled StructuresCitation Excerpt :It can be seen from the above statistics that more than half of the road traffic deaths are VRU in most motorized regions. For this reason, VRU’s safety has drawn extensive attention all over the world [16–19]. In order to reveal the casualty/injury mechanisms of VRU, numerous researchers have carried out in-depth investigations into various traffic accidents [20–24].
A survey on vehicular task offloading: Classification, issues, and challenges
2022, Journal of King Saud University - Computer and Information SciencesCitation Excerpt :With the growing number of mobile and IoT devices, more mobile and vehicular applications compete for computing capacity to provide a customized experience. In a vehicular environment, 5G automotive association (5GAA) has identified the following potential groups of use cases such as safety, convenience, advanced driving assistance, and vulnerable road user (VRU) (Malinverno et al., 2020; Nguyen et al., 2020). Most of these applications have strict requirements for real-time communication and intensive computation (Yao et al., 2018).
A survey on road safety and traffic efficiency vehicular applications based on C-V2X technologies
2022, Vehicular CommunicationsCitation Excerpt :Many of the results of the works about V2V periodic messages are also relevant for V2P periodic messages, as they test the ability of the technology to enable the reception of periodic messages under timing constraints. Specific works considering V2P periodic messages with C-V2X technologies are [115] [116], which explored the use of the LTE Uu interface to implement applications to protect pedestrians, in combination with MEC technology to reduce latency. In [115], periodic messages are sent from pedestrians' devices to vehicles through the infrastructure, so vehicles can warn drivers of risks with pedestrians when appropriate.
Mobile phone user behavior's recognition using gyroscope sensor and ML algorithms
2021, Materials Today: Proceedings