This literature survey explores the domain of remote driving of road vehicles within autonomous vehicles, focusing on challenges and state-of-the-art solutions related to driving feedback, latency, support control, as well as remote driving platform and real applications. The advancement towards Level-5 autonomy faces challenges, including sensor reliability and diverse scenario feasibility. Currently, remote driving is identified as vital for commercialization, however, it comes with challenges like low situational awareness, latency, and a lack of comprehensive feedback mechanisms. Solutions proposed include enhancing visual feedback, developing haptic feedback, employing prediction techniques, and use control methods to support driver. This paper reviews the existing literature on remote driving in these fields, revealing research gaps and areas for future studies. Additionally, this paper reviews the industry applications of remote driving and shows the state-of-art use cases.

Remote driving is emerging as an important backup for automated vehicles, yet the adaptation process and learning rate of remote drivers remain unexplored. In this study, we present a simulator-based experiment designed to evaluate how drivers adapt to remote driving environments characterized by realistic delays and feedback. The experiment employs a high-fidelity driving simulator that replicates real-world remote driving conditions—including motion cueing, steering force, and auditory feedback—while introducing controlled driving feedback delays. Participants navigate a dynamic scenario incorporating changing curvature roads, slalom manoeuvres, lane changes, and parking tasks over 10 training rounds. Both objective performance metrics (e.g., time consumption, lane following deviation, velocity, lateral acceleration) and subjective assessments (e.g., trust, controllability, familiarity, workload)are collected and analysed using repeated measures ANOVA. Results indicate that drivers rapidly adapt to the remote driving environment, achieving stable familiarity and reduced mental workload within the first4-5 rounds. These findings provide valuable insights for designing effective training protocols and improving remote control tower systems.

Remote driving plays an essential role in coordinating automated vehicles in some challenging situations. Due to the changed driving environment, the experiences and behaviors of remote drivers would undergo some changes compared to conventional drivers. To study this, a continuous real-life and remote driving experiment is conducted under different driving conditions. In addition, the effect of steering force feedback (SFF) on the driving experience is also investigated. In order to achieve this, three types of SFF modes are compared. According to the results, no SFF significantly worsens the driving experience in both remote and real-life driving. Additionally, less force and returnability on steering wheel are needed in remote driving, and the steering force amplitude appears to influence the steering velocity of remote drivers. Furthermore, there is an increase in lane following deviation during remote driving. Remote drivers are also prone to driving at lower speeds and have a higher steering reversal rate. They also give larger steering angle inputs when crossing the cones in a slalom manoeuvre and cause the car to experience larger lateral acceleration. These findings provide indications on how to design SFF and how driving behavior and experience change in remote driving.

Steering feedback is one essential aspect to provide real world information, and can influence driving performance during remote driving. In this work, the classical feedback models based on physical characteristics (Physical Model) and modular characteristics (Modular Model) of the steering system are constructed separately, and the influences of it on the remote drivers are studied. Objective and subjective measurement methods are separately used for evaluating the performance of the feedback models. In the subjective assessment, a multi-level assessment method is used for studying the influence of steering models on driver’s intuitive feeling. In the objective assessment, lane following accuracy, steering reversal rates, vehicle speed, time consumption, and throttle engagement are studied for different feedback models and scenarios. Moreover, the human biological information of electroencephalogram and heart rate variability are measured for studying the workload differences. The results showed that the physical model gave drivers a better steering characteristic feel and confidence in remote driving while the modular model could provide better real world feel. Returnability was an important parameter in remote driving, and the level of feedback force and returnability speed could be lower in remote driving compared to real car driving. It was also found that drivers had a higher workload in remote driving compared to real car driving.

Driving feedback is an important factor that can affect the perceptions of remote drivers of the surrounding environment during teleoperation. This paper focuses on investigating the influence of motion-cueing, sound and vibration feedback on driving behaviour and experience. A prototype teleoperation station is developed with feedback from audio, vibration actuators, and motion cues. Using this prototype, the experiment is carried out in two scenarios: a low-speed disturbance scenario with 30 participants and a dynamic driving scenario with 22 participants. Objective and subjective assessment methods are used to evaluate driving behaviour and experience separately. The results indicate that the combination of motion-cueing, sound and vibration feedback provides the most favourable driving experience for the participants. Specifically, sound and vibration feedback enhance drivers' sense of speed, while motion-cueing feedback helps in road surface sensing, leading to increased throttle reversal rate in the low-speed disturbance scenario. However, it is noteworthy that motion-cueing feedback does not significantly improve driving performance in the dynamic driving scenario of this study.

Driving feedback is an important way of providing remote drivers with physical world information during teleoperation. In this study, a teleoperation experiment is conducted to explore how sound, vibration and motion-cueing feedback influence the drivers’ driving experience and behaviour. To this end, four types of driving feedback modes are used as variables to investigate this, including no feedback, motion-cueing feedback, sound and vibration feedback, and a combination of sound, vibration, and motion-cueing feedback. A prototype of teleoperation platform is first built, which includes a teleoperated vehicle and a driving station capable of generating sound, vibration, and motion-cueing feedback. Then, the scenario with disturbances is built to investigate how the driving behaviour changes under various driving feedback modes. Both subjective and objective assessments are used in this study. For driving experience, the driving feeling, such as presence feeling, road surface feeling, etc, are explored. For driving behaviour, the throttle reversal rate is investigated. Furthermore, the relationship between throttle reversal rate and driving experience is studied. The results show that the combined feedback mode could provide drivers with the highest rated driving experience; the motion-cueing feedback could provide better road surface feeling while the sound and vibration feedback could provide better speed feeling. The throttle reversal rate with motion-cueing feedback is higher than without it, which may be caused by the increased road surface feeling provided by motion cues.

Remote driving plays a vital role in coordinating automated vehicles in challenging situations. Data transmission latency, however, can cause several problems in remote driving. Firstly, it can degrade the performance of remote-controlled vehicles, evident in issues like lane-following deviation and vehicle stability. Additionally, the remote control tower's driving feedback is affected by delayed vehicle signals, leading to delayed driving experience. To address this, a model-free-based predictor is employed to compensate for the delay in remote driving. This approach does not require any dynamic model of the system and only needs tuning of two parameters to reduce communication delay. This study enhances the previous work by mitigating the amplitude of overshoot around peak points. It leverages the principle of the second-order derivative to predict the signal's peak time and uses it to address the predictor's overshoot issue. The effectiveness of the proposed method is validated using real car data from multiple participants in two scenarios, including Slalom and lane-following. Simulation results indicate that the proposed method can reduce prediction error by nearly 25% compared to previous works. Moreover, the solutions in this study are capable of managing not only delays in remote driving vehicles but also in traditional mechanical systems, such as CAN bus delays in conventional cars.

Remote driving, as a backup system for automated vehicles, can play a vital role in their commercialization. However, delay is one of the major challenges in the practical application of remote driving. It not only degrades the stability of remote driven vehicles (RDVs) but also introduces delayed driving feedback, such as motion cueing feedback, to remote drivers. This can result in an unpleasant driving experience. This study proposes a square root cubature Kalman filter-based predictor (SRCKP)to compensate for driving feedback delays in remote driving. The SRCKP reduces the limitations of both model-based and model-free predictors (MFPs). Additionally, this paper presents an overshoot compensator to address the overshoot problem associated with traditional MFPs. Furthermore, a packet loss predictor (PLP) is designed to mitigate the influence of packet loss during data transmission. Both simulation and hardware-in-the loop(HIL) experiments during comprehensive driving scenarios are conducted to verify the effectiveness and robustness of theproposed method. The findings indicate that, compared to MFPs, the SRCKP reduces the L2-norm error by up to 81.2% in simulations and by up to 54.0% in HIL experiments for the best-case conditions.

Remote driving has emerged as an effective solution to bridge the gap between Level-4 and Level-5 autonomous vehicles by allowing human operators to take control when full autonomy is challenged by real-world complexities. However, delay remains a critical issue in remote driving systems. To address this, various predictors—such as the model-free predictor(MFP) and the square-root cubature Kalman filter-based predictor(SRCKP)—have been proposed. In this study, a simulator based remote driving experiment is conducted to evaluate the influence of these predictors on driver performance, experience, and workload. In addition, differences in driving behavior and performance across diverse driver profiles are explored, including comparisons between gamers and non-gamers, experienced and non-experienced drivers, as well as racing gamers and non-racing gamers. The findings indicate that the SRCKP nearly replicates the performance of a no-delay condition, whereas the MFP introduces oscillations due to overshoot issues, resulting in a less comfortable driving experience. Furthermore, experienced drivers consistently outperform non-experienced drivers. These results underscore the importance of advanced predictive algorithms and experienced drivers in enhancing the robustness and performance of remote driving systems.