Several studies have explored the development and evaluation of adaptive interfaces and driver assistance systems to support drivers during cognitively demanding situations. An adaptive multi-modal interface model was developed to assist drivers during a take-over request (TOR) from autonomous to manual driving, focusing on balancing convenience with safety risks (Chen et al., 2021).

A comprehensive evaluation method for highly automated driving functions across Society of Automotive Engineers (SAE) levels 2 to 4 was proposed by Rösener et al. (2017). Notably, human involvement is still requisite up to Level 4 of automation (Murtaza et al., 2023b). This method integrates technical, user-related, in-traffic, and impact assessment components. It uniquely combines various testing tools and methodologies, including field tests, simulations, and user trials, to provide a holistic view of the multifaceted implications of automated driving. The cognitive and emotional impacts on drivers interacting with automated driving technologies have not been incorporated (Rösener et al., 2017).

The effects of haptic assistive driving systems on driving behaviors studied that highlighted the significance of a strong model-free controller for accommodating different driving styles. They ignore the high degree of uncertainties in human behavior. The impact of ambient lighting on driver performance and mental exertion during ten TORs have improved drivers' take-over performance. The sample collected with majority of young women may limiting the generalization (Noubissie Tientcheu et al., 2022).

How behavioral measures impact the braking performance of older drivers under cognitive workloads, particularly focusing on the left foot position as a preventive strategy were studies and, comparing simulated and real road environments to analyze braking errors and predict adaptive skills at intersections, achieving a 70% accuracy in forecasting adaptive skills by assessing the left foot posture of elderly drivers, enhancing the system's effectiveness and usefulness (Figalová et al., 2022). The study revealed that cooperative control strategy for lane-keeping enhanced cooperative driving quality by 9.4% and minimized conflict by 65.38% compared to the previous design without a sharing parameter, along with an 86.13% reduction in steering workload (Kajiwara and Murata, 2022).

They examined the acceptance of advanced safety features in Toyota Sienna and Prius models among 183 vehicle owners through telephone interviews, revealing a high level of interest in adaptive cruise control and forward collision avoidance, but slightly lower acceptance for lane departure warning and prevention, with variations observed based on drivers' age and gender, indicating potential differences in responses to emerging crash avoidance technologies in various vehicle models (Perozzi et al., 2022).

Their study focused on cognitive attributes impacting SA in driving scenarios, identifying eight cognitive states that influence drivers' SA. Their proposed guidelines show significant improvements in SA and decision-making through the new interface compared to existing systems (Eichelberger and McCartt, 2015). The research delves into the effects of training on drivers' capacity to engage with Advanced Driver Assistance Systems (ADAS) and Autonomous Vehicle (AV) technology, showcasing notable enhancements in precision, response time, and vehicle management post-training (Park et al., 2020). The research conducted by Murtaza et al. (2023a) and Murtaza et al. (2024a) underscored the significance of training in mitigating mental workload for operators of Advanced Driver-Assistance Systems (ADAS) and Autonomous Vehicles (AV). The studies compared the efficacy of video-based and text-based training modalities (Murtaza et al., 2024a) and evaluated conventional methods against those based on Large Language Models (LLM) (Murtaza et al., 2024b). These findings emphasize the necessity of designing tailored training programs to ensure the safe and proficient use of these technologies. The cited studies significantly highlight the imperative motivation behind the enhancement of driver safety and overall performance, which can be accomplished through the implementation of innovative adaptive technologies that utilize a variety of methodologies; these include, but are not limited to, the utilization of sophisticated driving simulators, the integration of haptic assistive driving systems, and the application of subjective workload assessments to gauge the efficacy of these interventions. In this context, it would be particularly beneficial to investigate and analyze how the findings from these studies could be further enriched and substantiated by the strategic incorporation of physiological measures that could provide deeper insights into the drivers' responses and behaviors during simulated driving tasks.

Researchers have investigated various methods to assess drivers' mental workload, including subjective measures, physiological indicators, and performance metrics. They reviewed the increasing popularity of remote eye measurements as tools for evaluating driver workload without causing distraction, emphasizing the importance of using multiple assessment methods for increased validity and reliability (Marquart et al., 2015). Researcher examined the effectiveness of eye-tracking metrics in evaluating a driver's cognitive load during semi-autonomous driving while performing non-driving tasks. They identified pupil diameter change, saccade characteristics, fixation duration, and 3D gaze entropy as consistent indicators of cognitive load in both visual and auditory multitasking situations (Chen et al., 2022). Their research found the impact of Non-Driving Related Task (NDRT) on takeover performance in Highly Automated Driving (HAD) contexts and the influence of workload on driver takeover performance. The study lacks details on sample size/participants, limiting generalizability (Yoon and Ji, 2019). Their study aims to analyze the relationship of driver reaction time, gaze-on time, road-fixation time, and take-over time. They solely examine the driver's reaction time, disregarding driving accuracy or safety (Ko and Ji, 2018). The study analyzed descriptive statistics and Friedman test results of pupil size under different HMI conditions at stop-sign intersections with and without traffic (Li et al., 2020). In summary, these studies demonstrate the motivation to develop reliable and non-intrusive methods for assessing driver mental workload, employing techniques such as eye tracking, subjective questionnaires, and secondary task performance. However, limitations include the need for more diverse driving scenarios and the potential influence of individual differences on workload measures.

Various studies have focused on the link between physiological indicators, particularly HRV and visual movements, and the cognitive states of drivers, involving mental workload, fatigue, and drowsiness. One particular study scrutinizes the application of in-vehicle physiological sensors to assess drivers' mental workload, focusing specifically on cardiovascular and respiratory data. This research delineates alterations in driving tasks that affect mental workload and the driver's capacity to regain control of the vehicle (Kanakapura Sriranga et al., 2023). Expanding upon this groundwork, an additional study introduces a methodology that leverages mutual information derived from EEG and vehicular signals to formulate a feature set for evaluating drivers' mental workload. This feature set exhibits promise in forecasting and classifying tasks with minimal error and substantial accuracy, with the Support Vector Machine (SVM) classifier outperforming alternative classifiers. Nonetheless, this investigation overlooks individual characteristics and experiential factors (Islam et al., 2020). Subsequent research emphasizes the evaluation of drivers' cognitive workload through cognitive assessments and physiological indicators such as heart rate variability and infrared thermal imaging. Machine learning models attained elevated levels of accuracy, although this accuracy may be affected by an increased sample size (Cardone et al., 2022). Another analysis underscores the potential of HRV as a metric for mental workload, indicating that the mean heart rate escalates with heightened demands in secondary tasks, and younger drivers display elevated heart rates compared to their older counterparts (Biondi et al., 2017). A systematic review focusing on the identification of driver fatigue through HRV indicates inconsistencies in HRV metrics between alert and fatigued drivers, underscoring the need for additional research into the relationship between HRV and the causes of fatigue (Lu et al., 2022). In conjunction with this, an article introduces a novel adaptive interface bridging human and machine interaction that facilitates the automatic rerouting of incoming phone calls to voicemail without alerting the driver when the estimated workload exceeds a specified threshold (Piechulla et al., 2003). Regarding methodological innovations, one manuscript evaluates the application of wavelet analysis on heart rate variability signals to differentiate between alert and drowsy driving events, contrasting it with the conventional fast Fourier transform approach. The wavelet-based technique demonstrated superior efficacy in classification performance, achieving an accuracy of 95% through leave-one-out validation (Li and Chung, 2013). Another investigation sought to establish a heart rate monitoring system utilizing an Arduino Kit and a heart rate sensor to evaluate drivers' mental workload and enhance road safety through timely notifications (Makhtar and Sulaiman, 2020). In investigating the effects of mental workload fluctuations on cerebral activity, a study examined military drivers in both combat and non-combat contexts using EEG. Elevated theta EEG power spectrum in the frontal, temporal, and occipital regions during complex scenarios signified increased mental workload. The data were scrutinized to assess the implications of workload on cerebral activity and driving efficacy (Diaz-Piedra et al., 2020). The Taiwan Bus Driver Cohort Study (TBDCS) employed a cohort methodology to examine the predictive capacity of HRV analysis concerning cardiovascular disease risk among occupational drivers over an 8-year duration (Wang et al., 2021). A physiological recognition framework was established utilizing HRV metrics for feature construction and selection, which incorporated the Best Individual N (BIN) search method, Kernel-based Class Separability (KBCS) selection criterion, Principal Component Analysis (PCA), and Linear Discriminant Analysis (LDA) for feature extraction, alongside the k-nearest neighbors (k-NN) algorithm for recognition purposes (Wang et al., 2010). Another investigation focused on the capacity to detect drowsiness through HRV analysis in drivers utilizing a driving simulator, integrating time-domain, frequency-domain, and fractal methodologies, in addition to theta brain wave activity (Mahachandra et al., 2012). Research concerning Automated Valet Parking systems explored the impact of scenario-based explanations on user trust and satisfaction, underscoring the necessity of customized solutions for diverse user demographics to bolster trust, enhance user experience, and mitigate cognitive load (Ma and Feng, 2023). The effectiveness of HRV in recognizing driver drowsiness during operation was also assessed through machine learning methodologies, achieving an accuracy rate of 85% with the random forest classifier, although challenges arising from confounding variables were acknowledged (Persson et al., 2020). Furthermore, a study investigated heart rate variability and perceived workload in simulated driving conditions, particularly within work zones on highways. The results indicated that participants encountered heightened workload in work zone environments and under conditions of increased traffic density, despite heart rate metrics remaining largely unaffected (Shakouri et al., 2018). In aggregate, these investigations underscore the impetus to devise trustworthy and non-invasive methodologies for monitoring cognitive states in drivers, leveraging subjective workload assessments, physiological sensors, and machine learning techniques. Notwithstanding the encouraging advancements, challenges such as the necessity for larger sample sizes, validation in real-world settings, and the consideration of individual variability in physiological responses persist. The integration of these multifaceted approaches and findings could facilitate the development of more effective and adaptive driver monitoring systems, ultimately driver welfare.

Numerous studies have been conducted to explore the relationship between HRV and MWL. These studies have consistently observed decreased HRV during mental tasks, particularly in the context of non-linear metrics such as D2. Additionally, alterations in HRV have been found to correlate with feelings of frustration and changes in myocardial oxygen consumption. Considering the real-world applications of HRV, it has proven to be a valuable tool in assessing pilot workload, monitoring stress levels in healthcare professionals, and optimizing cognitive performance. By leveraging the insights provided by HRV, individuals and organizations can develop strategies to effectively manage and maintain wellbeing in demanding environments. In this extensive investigation of driving performance, in Table 1 a variety of methodologies have been utilized to uncover crucial insights regarding safety, workload management, and cognitive aspects. The brain's electrical activity serves as a reliable indicator of workload among military drivers, showing significant potential for implementation in practical training situations. The incorporation of wearable sensor technologies, like ear electrode arrays, has the potential to enhance the real-time monitoring of cognitive states, thus improving road safety. Additionally, metrics such as SDNN and LF levels have displayed independent predictive abilities concerning heart and blood pressure conditions, offering valuable information to understand the mental workload in driving situations. The utilization of HRV-based recognition methods has demonstrated superior efficacy compared to other approaches, establishing it as a vital tool for identifying patterns in heart rate variability. Furthermore, biological signals such as RMSSD and SD1 have been recognized as sensitive parameters for detecting sleepiness, showing promise for the advancement of effective sleepiness detection systems. Particularly, explanations based on scenarios have proven to significantly boost situational trust, user experience, and mental workload management among drivers, resulting in quicker reaction times and reduced return times. Moreover, higher situation awareness scores have been linked to more efficient decision-making processes, despite some lingering indecision in decision-making. Interestingly, driving experience has not shown significant impacts on situation awareness or decision-making questionnaire scores. Research has unveiled that ambient in-vehicle lighting can improve drivers' takeover performance, while EEG markers of mental workload have shown no significant differences between experimental and control groups. Furthermore, older drivers have demonstrated increased cognitive errors and mistakes in actions with age progression; however, their accident likelihood has decreased when faced with collision risks. The accurate prediction of coping abilities based on left foot posture near intersections has achieved a commendable accuracy level of 70%. Additionally, evaluations of fatigue and workload levels have heavily relied on heart rate variability and infrared measurements. Adaptive phone settings have reduced mental workload for experienced drivers, with the SVM-based drowsiness detection system outperforming FFT-based methods, achieving an impressive accuracy of 95%. Lastly, heart rate readings during mental workload tasks have provided valuable insights into drivers' physiological reactions. Scholars have delved into methodologies related to heart rate variability and subjective workload in the driving realm, with the random forest classifier emerging as the preferred binary classification tool, despite a 20-percentage point decrease in performance for multi-class classification. Importantly, the development of personalized algorithms is crucial for the efficient monitoring of driver sleepiness based on HRV metrics. Additionally, weak correlations have been noted between heart rate variability measures and subjective workload, as well as between driving performance indicators and subjective workload.

We have explored various methods and their impact on driving performance and workload. Notable findings include improved driving performance with specific interventions (e.g., DS, AOSPAN, PASAT), safety consequences during obstacle encounters (F&Tt, AdaptIVe), and the influence of lighting conditions (DS, F&Tt, EEG). Metrics such as heart rate, eye movement, and pupillary response were used to assess workload. Additionally, SVM demonstrated high accuracy in workload classification tasks. Overall, these studies contribute valuable insights for enhancing driving safety and efficiency. In this comprehensive investigation encompassing diverse methodologies, in we have meticulously scrutinized critical parameters associated with driving performance. Notably, EEG theta power activity emerged as a robust workload indicator among army combat drivers, holding significant promise for real-world training scenarios. The integration of wearable sensor technologies, specifically around-the-ear electrode arrays, presents an avenue for real-time monitoring of cognitive states, thereby enhancing road safety. Furthermore, in Table 2 the study delved into cardiovascular health, identifying independent predictors of cardiovascular disease (CVD) and hypertensive disease through parameters such as standard deviation of normal-to-normal intervals (SDNN) and low-frequency (LF) levels. These insights hold valuable implications for assessing long-term CVD risk. The heart rate variability parameter-based recognition strategy outperformed alternative methods, positioning it as a valuable tool for identifying intricate patterns within heart rate variability. Additionally, biological signals, such as root mean square of successive RR interval differences (RMSSD) and standard deviation of Poincaré plot (SD1), exhibited sensitivity in detecting sleepiness, offering potential for the development of effective sleepiness detection systems.

In this specific scenario, an investigation is carried out concerning the impacts of different physiological parameters. In Figure 2 a comprehensive examination is performed regarding their individual contributions such as HR, Eye, ECG, and other parameters. Heart Rate (HR): With the most significant weightage of 46%, heart rate occupies a central position within the physiological framework. It potentially serves as an indicator of an individual's cardiovascular health or stress levels, thus representing a crucial aspect of their overall wellbeing. Eye: Holding a substantial 25% contribution, parameters related to the eye are important. These parameters may involve visual attention, eye movements, or cognitive processes, offering insights into various cognitive and visual functions. Electroencephalogram (EEG), Electrocardiogram (ECG), and Electromyogram (EMG): Each of these factors carries a 32% contribution, emphasizing their equal importance in the realm of physiology. EEG records the brain's electrical activity, ECG monitors heart function, while EMG assesses muscle activity levels, collectively providing essential insights into diverse bodily functions. Blood Flow (BF): Representing 11% of the overall impact, blood flow may seem less prominent compared to other parameters. Nevertheless, its interpretation within a particular context could yield valuable and detailed information concerning physiological responses. Hypertension (Hypt): Contributing minimally at only 4%, hypertension appears to exert a relatively minor influence on the ongoing physiological assessment. Reaction Time (RT): Playing a moderate role with a 14% contribution, reaction time introduces another layer to the thorough analysis of physiological parameters, underscoring the significance of prompt responses in various situations. Figure 3 illustrates the predominant utilization of computational tools in the calculation of physiological parameters. Statistical software, accounting for 64%, is primarily employed for data analysis and hypothesis testing. Machine learning techniques, representing 25%, play a significant role in predictive modeling and pattern recognition. Additionally, automotive hardware tools, constituting 18%, are crucial for vehicle control and performance optimization.

In summary the paper presented a thorough examination of the cognitive deficits resulting from mental workload and heart rate variability within the driving safety framework, emphasizing the relationship among cognitive impairment, workload, and physiological signals such as heart rate and eye movements. The research underscores the significance of understanding the interplay among cognitive abilities, workload requirements, and bodily reactions for ensuring driving safety. One of the limitations of the current research is the lack of consideration for individual differences in cognitive and physiological responses to mental workload. Age, driving experience, and cognitive abilities may influence how drivers perceive and respond to cognitive demands. Future studies should investigate these individual differences to develop more personalized and adaptive interventions for enhancing driving safety. Another limitation is the reliance on simulated driving environments in many reviewed studies. While driving simulators provide a controlled setting for assessing driver performance and cognitive states, they may not fully capture the complexities and unpredictability of real-world driving situations. Future research should aim to validate the findings from simulator studies in real-world driving conditions to ensure the generalizability and ecological validity of the results. Despite these limitations, the findings from this review have important implications for the development of driver monitoring systems and the design of adaptive interfaces. By incorporating physiological measures such as HRV along with ML techniques, it is possible to create more effective and personalized interventions for managing driver mental workload and preventing cognitive impairment. However, successfully implementing these approaches will require close collaboration between researchers, industry partners, and policymakers to develop reliable, valid, and ethically sound solutions.