1. Introduction
For the past several years, connected and automated vehicle technology and systems have become a rapidly growing focus of government-funded research and development activities in the field of intelligent transportation systems [
1]. Connected and automated vehicles have been widely recognized as potentially able to deliver significant benefits to society; they could greatly improve fuel efficiency and reduce emission, which has the potential to not only significantly contribute to environmental sustainability, but also to economic improvement and transportation decarbonization [
2,
3]. They could also significantly enhance road safety by potentially preventing traffic crashes caused by human error [
4]. In addition, connected and automated vehicles could also play an essential role in reducing traffic congestion by significantly decreasing the average trip times in heavy traffic conditions [
5]. Moreover, they also have a strong association with improvements in social inclusion and enhancing accessibility, especially by providing tailored support to user groups with different mobility needs [
6,
7,
8].
Vehicles equipped with connected and automated driving technologies can be grouped into several categories according to system functionalities, as well as the level of human input needed. One of the most commonly adopted classifications is the six-level definition of vehicle automation by SAE International (Society of Automotive Engineers) [
9]. Systems of the lower levels of vehicle automation (SAE Level 0 to Level 2) provide different levels of automated features and support the driver in different ways, from providing in-vehicle sensory and informational assistance to completely releasing the driver from the physical driving loop. However, they provide support under one common condition—that the human driver must always be mentally engaged in the vehicle control loop and constantly monitor the driving [
9,
10]. Compared to the lower levels of automation systems, Level 3 automation systems are the intermediate stage and potentially provide the driver with higher levels of autonomy. This works by enabling them to be temporarily disengaged from driving both physically and mentally and allowing them to safely engage in other non-driving related activities that could be prohibited when driving conventional vehicles, such as using mobile phones or watching movies [
1,
6,
7,
11]. However, the biggest challenge for a Level 3 automation system is that it relies on the human driver on-board to reassume control of the vehicle within the provided lead time when the automation system reaches the limitation of its operational design domains [
1,
9,
10]. There could be severe consequences for Level 3 automation systems if the driver fails to take over control of the vehicle effectively within the allocated time frame. Thus, the transition control between Level 3 automation systems and the human driver on-board has become a popular topic for research [
10,
12,
13,
14]. For Level 4 automation systems upwards, the system needs to be fail-operational, which ensures the full or degraded operation of the vehicle even if a failure or system limitation occurs, including situations where a human driver fails to take over control when requested [
9,
10,
15]. The failsafe mode has become one of the important challenges for the development of the Level 4 automated systems, which triggers the necessity to explore safe, effective and reliable fail-operational measures. One potential interim solution for the failsafe of Level 4 automated vehicles is a teleoperation system controlled by a remote driver [
16]. To enable the teleoperation solution for automated vehicles, 5G technologies play an imperative role, as they can profoundly enhance network connection and provide ultra-low latency communications [
17].
The 5G-enabled Connected and Automated Logistic project (5G CAL)cost GBP 4.9M, including GBP 2.4M from 5G Create, an open competition and part of the UK government’s 5G testbeds and trials programme (5GTT) from the UK Department for Digital, Culture, Media and Sport (DCMS). The 5G CAL project designed, developed, deployed and evaluated the 5G-enabled Level 4 automated vehicles (5G L4 AV) with a teleoperation system. The project applied the 5G L4 AV in the context of the logistic sector. The key focuses of the project included: to demonstrate the capability of 5G technologies in enabling safe, effective and secure operation, and the transition of control from the automated vehicle to the teleoperation system; to use 5G technologies to monitor the status of automated vehicles using real-time telematics; and to facilitate understanding of the deployment of 5G-enabled connected and automated logistics at scale. The key members of the project team included Newcastle University, StreetDrone, Vantec, North East Automotive Alliance, Sunderland City Council, Perform Green, Coventry University, Connected Places Catapult, with Nissan, Terberg and Fergusson as contributors.
1.1. State of the Art and Research Gaps
Existing research has attempted to explore the teleoperation of vehicle automation from different perspectives. Goodall [
18] reviewed the legal environment of the teleoperation of vehicles and developed a model to predict the required number of remote drivers for operating large-scale automated vehicle fleets. The model prediction results indicated that a high proportion of jobs currently belonging to professional human drivers in the USA could potentially be replaced by automated vehicles managed by remote drivers. The research highlighted the importance for the government of reviewing existing policy and paying more attention to teleoperation, which has the potential to significantly facilitate the development of vehicle automation. Some research focuses on the human–machine interface (HMI) of the teleoperation of automated vehicles. Kettwich et al. [
16] evaluated a prototype human–machine interface design consisting of video screens, a details screen, disturbances screen, map screen and touchscreen for the teleoperation of SAE Level 4 automated vehicles with thirteen end-users from public transport control centres in Germany. The prototype HMI has received positive feedback from the end-users. Other research focuses on the classification and definition of teleoperation concepts and terminologies. Majstorovic et al. [
19] conducted a systematic review on different teleoperation concepts that are used as fallback solutions to deal with critical situations and operation design domain limitations in automated vehicles. Teleoperation concepts were grouped into six categories, including Direct Control, Shared Control, Trajectory Guidance, Waypoint Guidance, Interactive Path Planning and Perception Modification. Moreover, Bogdoll et al. [
20] conducted a survey on the terminologies of remote human input systems for automated driving and proposed a taxonomy which potentially provides clarity and reduces confusion in the field of the teleoperation of vehicle automation. Finally, Zulqarnain et al. [
21] proposed algorithms to solve the largest challenge of teleoperation systems in automated vehicles—that is, the latency between the teleportation workstation where the remote drivers are located and the automated vehicles. Their research potentially enlightens the selection of suitable locations for teleoperation workstations.
The existing literature and previous studies have researched the teleoperation or remote operation of automated vehicles from various perspectives, including its legal environment and its implications on existing driver jobs, designing and evaluating the HMI and algorithms, as well as defining and classifying terminologies and concepts. However, there are still significant research gaps. For example:
There is limited research regarding the deployment and evaluation of full-scale authentic automated vehicles incorporating a teleoperation solution in the real world.
Existing research regarding the teleoperation or remote operation of vehicle automaton has neglected one of the most essential elements of the teleoperation system—the remote drivers (operators).
Understanding the acceptance, perception and requirements of support from the remote driver’s perspective is important to design and develop safe, effective and user-friendly teleoperation systems for automated vehicles. However, knowledge regarding the needs and requirements of remote drivers when interacting with automated vehicles is limited.
The lack of the above knowledge could potentially thwart the development and real -world deployment of automation systems, especially those of higher levels (SAE Level 4 and over), thereby preventing automated vehicles from delivering the expected environmental, economic, safety and social benefits.
1.2. Purpose of the Research
To address the research gaps identified above, the overall aim of this study was to provide new knowledge and unique insights into the teleoperation of vehicle automation by qualitatively investigating remote drivers’ perceptions, attitudes, needs and requirements when teleoperating 5G-enabled Level 4 Automated Vehicles in the real world.
4. Conclusions and Future Work
Connected and automated vehicles have the potential to deliver great benefits in terms of enhancing road safety, reducing traffic emission, optimising road efficiency and improving social inclusion [
23]. They also potentially bring new opportunities to fundamentally change urban mobility and logistics services [
34]. Among the vehicles equipped with higher levels of automation systems (SAE Level 3 and over), Level 4 automated vehicles could include a failsafe mode which ensures the safety of the vehicle in urgent situations where the vehicle is out of the designed service areas [
9,
10]. The failsafe mode could be achieved by a teleoperation system in which a human driver takes over control and operates the vehicle remotely [
16]. However, to date, limited research has explored the perception, attitudes, needs and requirements of the remote drivers of Level 4 automated vehicles. To fill this research gap, this research used a qualitative methodology to investigate the remote drivers’ perception, needs and requirements when operating the L4 AV remotely via a 5G-enabled teleoperation system in the real world. The study identified six core themes representing the needs and requirements of the remote drivers who perform the essential role of teleoperation control of the system. The research found that the remote drivers exhibited positive attitudes towards the 5G L4 AV. They also highlighted the need for further improvement of the human–machine interaction in the teleoperation system of the 5G L4 AV in terms of the visual field for driving, as well as the perception of feedback from the vehicle. In summary, the main findings of this study are as follows:
Remote drivers have positive attitudes towards the 5G L4 AV.
Remote drivers would be monitoring the road when the 5G L4 LV is performing automated driving. They expect to be informed if something happens.
In terms of the human–machine interface, remote drivers would like to have verbal communication if there is a safety driver on-board. If there were no safety drivers on board (ultimately the desired scenario), a visual, audible and vibrational HMI would be beneficial.
The main difference and difficulties remote drivers experienced when controlling the vehicle remotely (compared to conventional manual driving) was lack of depth vision, as well as not being able to feel the feedback from the vehicle when executing a manoeuvre.
Remote drivers would like more support regarding their visual field driving when teleoperating the 5G L4 AV.
Remote drivers would like more support in terms of enhancing the perception of physical feedback when teleoperating the 5G L4 AV.
Possible support includes introducing Virtual Reality, wide angle mirrors, as well as full motion feedback systems into the teleoperation workstation of the 5G L4 AV.
Remote drivers collaborated smoothly with the on-board safety driver, and their strategic decision making in urgent situations was consistent with the on-board safety drivers’.
This study is one of the world’s first pieces of research adopting a qualitative methodology to explore the user requirements from the remote driver perspective of a 5G L4 AV system deployed in the real world. The findings of this work have important implications for stakeholders of the 5G L4 AV, including manufacturers, policymakers, academics and researchers, in terms of the development and deployment of safe and efficient human–machine interactions of teleoperation systems in the 5G L4 AV. This study strengthened the importance of implementing a user-centred approach when designing, developing and deploying future mobility technologies and connected and automated vehicles [
11,
23,
35,
36]. The findings and new knowledge gained in this research could be developed and used in the following directions to enlighten future work. To begin with, the remote drivers of the L4 AV perceived some differences between operating the vehicle on board and via the teleoperation system. Therefore, ensuring the fidelity of a teleoperation system in terms of enhancing the remote driver’s sensing (vision and hearing) of the driving environment, as well as improving their perception of physical motion feedback from the vehicle, could be an important challenge for the future development and deployment of L4 AVs. Future work could focus on exploring potential measures to enhance the perceptual fidelity and motion fidelity of the teleoperation workstation of the 5G L4 AV. Apart from that, tailored training could also be potentially useful for remote drivers to adapt the teleoperation system more efficiently and rapidly. Future work could explore what kinds of training are needed for a conventional vehicle driver to become a remote driver and assess the training quality and outcome. In addition, future work could further explore and evaluate which modality of the human–machine interfaces could lead to better performance among remote drivers. This study investigated remote drivers’ requirements when teleoperating the 5G L4 AV with a safety driver on-board. Future work could explore what their needs are when there is no safety driver. Moreover, the study found that remote drivers would like to monitor the vehicle driving when they are not teleoperating the vehicle. In future work, it would be worth seeking to quantify and evaluate remote drivers’ attention, mental workload and situation awareness while monitoring the system. It is also necessary for future work to investigate the potential impact of distraction and driving disengagement [
8], as well as performing non-driving related tasks, on remote drivers’ attention and performance. Moreover, this study is from the perspective of the remote driver who is an essential part of the teleoperation process—the failsafe mechanism of the 5G L4 AV. Future work could focus on different stakeholders of 5G L4 AVs, for example, from the perspective of the end-users and customers who are planning to adapt 5G L4 AVs in their service and business. Although the study yielded novel findings and has key implications for future work, there are still some limitations. Firstly, this study is qualitative in nature; it does not aim for the generalizability of the results but to develop contextualized and in-depth knowledge regarding the remote drivers’ interaction with the 5G L4 AV [
37]. Future studies could research the teleoperation of the Level 4 AV with a larger sample size and adopt quantitative methodologies to further analyse the findings. The key knowledge gained in this study could facilitate the experimental design. Moreover, future research could explore whether the demo-graphic factors [
6,
7] of the remote driver would impact their needs and requirements when interacting with the 5G L4 AV. The ultimate goal may be that by understanding these critical interactions better, we will be able to transition to fully driverless L4 vehicles with a remote driver monitoring a number of vehicles. This would deliver a real economic benefit to the operator. Overall, the knowledge gained in this study could be important evidence potentially informing the development of strategy, policy, practice and service provision [
38] in the field of 5G-enabled automated vehicles.