Effects of Vehicle Air Temperature on Drivers’ Cognitive Abilities Based on EEG

Abstract

Vehicle air temperature affects drivers’ physiology, psychology, and cognitive abilities. However, the effects are difficult to quantify, especially for jobs related to driving tasks. In this research, 10 male subjects were directly exposed to four different vehicle air temperatures of 20, 23, 26, and 30 °C for 160 min. They were asked to perform cognitive tasks and subjective questionnaires, and 16 channels of EEG signals were monitored in a vehicle cabin. Based on the assessment of the EEG characteristics, the impacts of vehicle air temperature on cognitive abilities and EEG were investigated. The results showed that the cognitive ability of drivers decreased with the rising of the ambient temperature. The subjective questionnaire scores for thermal sensation, thermal comfort and brain load increased as ambient temperature rose; meanwhile, the scores for environmental acceptance, job satisfaction and willingness to work declined. As the ambient temperature rose, the normalized power of θ activity and α activity elevated, and the vigilance and frontal EEG asymmetry decreased. At 20 °C, the completion time of cognitive ability test was the shortest, the number of errors was the smallest, and the drivers could maintain high cognitive ability. At this time, the β activity component of the EEG signal increased, and the level of alertness (AL) and prefrontal asymmetry (FEA) also increased. At 23 °C, drivers’ subjective thermal comfort reached its peak: the EEG wavelet entropy values of the two segments before and after the experiment were the largest, and the wavelet entropy difference was also the largest. A suitable vehicle air temperature aroused β activity and motivation, increased driver alertness and thus enhanced cognitive performance. Therefore, to achieve high cognitive ability and thermal comfort, the vehicle air temperature should be maintained between 20 °C and 23 °C. The research results can provide a reference for the design standards of vehicle air temperature and improve the safety of driving.

1. Introduction

A vehicle’s ambient temperature affects the driver’s physical, psychological, and thermal comfort, as well as cognitive abilities. Hot and uncomfortable interior environments can threaten driving safety, lead to driver inattention and poor decision-making, ultimately resulting in traffic accidents [1,2,3].

Most of the existing studies on the effects of vehicle air temperature on drivers have focused on subjective thermal comfort. Yun et al. [4] studied the key factors affecting vehicle thermal comfort and found that changes in air temperature would lead to changes in thermal comfort. Li and Chen [5] explored the relationship between human thermophysiological response and subjective thermal sensation in the driver’s compartment, and found that the heart rate and blood pressure of the driver had a strong correlation with their thermal sensation scores. Rosaria and Chiara [6] studied the sensitivity of different human body parts to changes in vehicle temperature and found the most significant effect on the shoulder and back regions. Several studies [7,8] have shown that drivers make more driving errors in hot environments, which can lead to traffic accidents.

Existing studies on people’s cognitive abilities have been conducted in indoor environments, and found that changes in indoor ambient temperature not only affect people’s thermal comfort but also impact their cognitive abilities. In 2020, Kim et al. [9] analyzed the factors affecting college students’ cognitive abilities in different indoor thermal environments and the changes induced in their physiological and psychological parameters. The best cognitive performance was observed at 25.7 °C, while indoor thermal conditions could affect cognitive performance by activating physiological responses. Zhai and Tham [10,11] found that indoor air temperature significantly affected work performance through simulations of office environments. Lower room temperatures reduced thermal comfort, stimulated the nervous system that controls thermoregulation, and increased individual alertness and attention.

Electroencephalogram (EEG) is a physiological and neuroscience approach that rapidly reflects the whole cortical and scalp surface electrophysiological activities [12]. The EEG can monitor and record the brain’s central nervous system and area rhythm activities [13]. EEG is also used in diagnosing several neurological disorders such as Alzheimer’s and epileptic seizures [14]. The EEG signal is weak, its amplitude range is below 100 μV, and the signal-to-noise ratio is low [15]. This causes some artifacts (mainly eye movement artifacts produced by blinking and eyeball rolling, sweat artifacts usually appear in the hyperthermal environment) to greatly impact the experimental results. Electroencephalogram (EEG) signals may get easily contaminated by muscle artifacts, which may lead to the wrong interpretation in the brain-computer interface (BCI) system, as well as in various medical diagnoses [16]. The existing studies [17] are proposed algorithms that combine unsupervised eye blink artifact detection (eADA) with modified Empirical Mode Decomposition (Fast EMD) and Canonical Correlation Analysis (CCA), to automatically identify eye blink artifacts and remove them in an online setting. A sleep disorder is a medical condition that affects an individual’s regular sleeping pattern and routine, hence negatively affecting the individual’s health [18]. Emotion recognition using EEG has been widely studied to address the challenges associated with affective computing [19].

In the study of cognitive performance under brain load in 1997, Jung et al. [20] first used EEG power spectroscopy to assess alertness, and reported a decreasing trend in alertness levels with the progression of time. In addition to acting on the body surface, ambient temperature also serves on the brain and other components of the nervous system, affecting a person’s cognitive abilities. Christensen et al. [21] found that EEG signals can identify tasks of different difficulty as different brain loads, which can be used as a critical indicator of cognitive load and mental state. Zheng et al. [22] revealed a significant negative correlation between forehead temperature and reaction time under high-temperature radiation, which can be considered a critical factor affecting efficiency. Several studies have used parameters such as (α + θ)/β, α/β, θ/β, and (α + θ)/(β + θ) as indicators of alertness [23,24], and also nonlinear indicators such as forehead asymmetry [25] and wavelet entropy [26] to monitor brain load.

In summary, changes in the vehicle air temperature affect drivers’ subjective comfort and significantly influence their cognitive abilities. Due to the difference in space size and intensity of personnel activities, the law of the effect of indoor thermal environment on cognitive ability does not apply to the evaluation of drivers’ cognitive ability. In this paper, by simulating driving tasks in different vehicle environmental temperatures and obtaining the data of the subject’s physiological parameters, subjective perceptions, and cognitive ability tests, we further analyzed the mechanisms of the effects of different vehicle environmental temperatures on drivers’ cognitive abilities based on EEG signals. The research results can provide a reference for the design standards of vehicle air temperature and safe driving.

Therefore, the main purpose of this study is to establish the relationship between driver response and vehicle air temperature. The primary objectives of this study were:

• To explore how the vehicle air temperature influences the driver’s cognitive abilities, EEG features, vigilance, and thermal comfort;
• To investigate the changing pattern in the driver’s EEG under vehicle air temperature;
• To study the potential explanatory mechanisms of changes in cognitive abilities underlying the vehicle air temperature and practical contribution.

2. Methodologies

2.1. Experimental Facilities and Conditions

Subjects were asked to perform single-environment road (high-speed) driving [27], cognitive ability tests, and questionnaires on a constructed simulation platform. The co-driver support person measured the subjects’ physiological data and recorded the cognitive ability test completion time and the number of errors. The test layout and in-vehicle environment are shown in Figure 1.

2.2. Measurement and Instruments

Modern EEG devices, especially wearable EEG monitors, contain a large number of electrodes that can be placed in different regions of the brain using the head and bilateral mastoids as reference electrodes. In this research, data collection and transmission utilized wireless Bluetooth technology. (Figure 2a depicts various brain activation in different locations, and Figure 2c presents brain activation caused by changes in cognition. For example, electrodes are positioned in the frontal region (e.g., AF3, AF4, F7, F8, F3, F4)).

These devices, as shown in

Table 1. Parameters of devices.

Type	Name	Measuring Devices	Versions	Range
Environmental parameters	TA	Testo-Smart	405i	−20 °C~ +60 °C
	RH	Probes		0–100%
	V			0–30 m/s
	TW	WBGT index meter	2006	5–40 °C
Physiological indices	EEG	EEG device	Open BCI	128 Hz 16 Channels

below, are placed near body areas sensitive to thermal ambient temperature and collect data every 10 s. The temperature and humidity ranges of these instruments are −20 °C to +60 °C and 0–100%, with ±accuracy of 0.5 °C and ± for 2%, respectively.

2.3. Participants

Herein, a within-subjects design was used to determine the sample size based on statistical power (Sp) and effect size (Es) [28]. A total of 10 male college students aged 21–24 years (M = 22.3 years, SD = 1.06 years) with a BMI of 22.51 ± 1.38 kg/m² were enrolled in the experiment. The subjects wore summer clothing, including long sleeves and long pants with a clothing thermal resistance of approximately 0.6 clo [29]. Subjects had regular daily routines, adequate sleep, and were in good health. The experiment included four different temperature conditions, and each subject was exposed to all of them. In this way, individual differences could be counteracted more effectively [30]. All subjects were informed of the experimental protocol and given hands-on training before the test to ensure stable proficiency in cognitive testing. A balanced Latin square design was used to control the test sequence to reduce fatigue and practice effects [31]. Adequate food and water intake was ensured before the test, and these were not supplemented during the trial to reduce the effects of thirst, hypoglycemia, and other factors. Female participants were excluded from this investigation, because the menstrual cycle could affect thermal comfort and other results [32].

2.4. Experimental Protocol

A graphic figure of the participant experimental framework is described in Figure 3. The whole study was performed in the spring from 1 to 30 May 2022. The total duration of each experiment was 160 min, including 30 min of pre-exposure preparation. To simulate the natural vehicle environment, we used air conditioning to control the ambient temperature while keeping the humidity within the range of 55 ± 5%. The air conditioner was turned on 30 min before the test started. Meanwhile, the participants were requested to sit in the measurement area for 30 min as an acclimatization period to stabilize their physiological parameters after arriving at the measurement chamber [33]. After the test started, the first and second half of the simulated driving tasks were performed sequentially. At the same time, the EEG acquisition device was turned on, and measurements and cognitive ability tests were conducted. The subjective questionnaire was filled out in the last 10 min of the trial.

2.5. Subjective Questionnaire

The questionnaire was composed of two main sections: (1) basic information, such as name, height, weight, age, etc.; and (2) the comprehensive questionnaires, which include surveys on indicators such as TSV, TCV, TAV, motivation, and NASA-TLX. The subjects’ thermal sensation, thermal comfort, environmental acceptability, brain load score, and psychological feeling condition were measured using a subjective questionnaire. The thermal sensation score, thermal comfort score (cold discomfort~hot discomfort) and environmental acceptability were designed based on ASHRAE [34,35]. The personal assessment of subjects’ brain load and time demands was performed using the NASA-TLX scale [36]. The range of this scale was changed to [−3,3], and the content of job satisfaction and job willingness scores was [−5,5] to facilitate the rapid assessment of scores.

2.6. Cognitive Ability Test

To assess the effect of vehicle air temperature on drivers’ cognitive abilities, cognitive ability tests were conducted in the experiment. The test included memory, thinking, perception, and vigilance abilities, and test time and several errors were used as indicators of cognitive ability evaluation. These programs are based on neurobehavioral tests and cognitive resource allocation, which are the most frequently utilized to evaluate neurocognitive processing [37].

One short test on the tablet was selected for each task, and the task completion time and the number of errors will be recorded and analyzed. We quantified subjects’ performance (productivity, intelligence level, and cognitive ability) on the task by the average completion time and the number of errors in the trials, the picture as shown in

Table 2. Cognitive ability test content.

Testing Capability	Test Project	Test Content	Legend
Memory	Graphics memory	Given a variety of patterns, select the correct pattern position by memory
Thinking	Shorthand simplified	In a certain period of time, arrive at the answers of five mixed numbers through mental arithmetic and then select the correct option
Perception	Find the target graphic	Find out the figure given by the topic from 36 different figures
Vigilance	Reaction speed	First press the left green key to start the test. When the middle light is on, click the right red key immediately

[38].

2.7. Data Processing Method

The subjects’ objective physiological data, cognitive performance, and subjective questionnaires were statistically analyzed using SPSS Statistics 26 and Origin 2018 [10]. Normality was tested by Shapiro-Wilk’s-W test, Analysis of Variance (ANOVA) and Least Significant Difference (LSD) methods were used for multiple comparisons of variables at different ambient temperatures, and Pearson analysis was employed to calculate each parameter [39]. The correlation between the parameters and the ambient temperature was calculated using Pearson analysis, and the significance level was set at p < 0.05.

2.8. EEG Features

The offline EEG was preprocessed using the EEG lab 2020 toolbox plug-in for MATLAB R2019b, with bilateral mammillary synapses re-referenced and the Finite Impulse Response (FIR) digital filter being band-pass filtered from 1 to 30 Hz. Eye rotation and blink artefacts were isolated using Independent Component Analysis (ICA), and artefacts were calibrated and removed using the IC Label plug-in [40].

2.8.1. Power Spectrum Features

Power spectrum characteristics are widely used in EEG feature extraction based on the Welch method using the Hamming window function for EEG signals. The power spectral density (PSD) at different rhythms is integrated to calculate the average power P. The PSD was calculated as in Equation (1).

$$PSD\left(\omega \right)=\underset{T\to \infty }{\lim }\frac{{\left|X_T\left(\omega \right)\right|}^2}{T}$$

(1)

$$P=\frac{1}{2\pi }{\int }_{-\infty }^{+\infty }\underset{T\to \infty }{\lim }\frac{{\left|X_T\left(\omega \right)\right|}^2}{T}d\omega$$

(2)

In the equation, $\omega$ represents the frequency domain function of the brain wave after the Fourier transform, and T is the signal period. EEG signals are classified into the following four categories based on frequency: δ (Delta, 1~4 Hz), θ (Theta, 4~8 Hz), α (Alpha, 8~13 Hz), β (Beta, 13~30 Hz).

2.8.2. Vigilance and Arousal

Since the intensity of brain waves varies from person to person, the power spectrum ratio is used to express the Alertness Level (AL). The larger the value of this index, the higher the level of alertness, which is calculated as follows:

$$AL=\frac{P\left(\beta \right)}{P\left(\alpha +\theta \right)}$$

(3)

where P(β) is the average power of β wave, P(α + θ) is the sum of the average power of α wave and the average power of θ waves. θ waves are intermediate between drowsiness and wakefulness, α waves are prominent during relaxation and diminish or disappear with concentration, and β waves are associated with active thinking, high levels of engagement, alertness, attention, or solving specific problems [20].

2.8.3. Frontal EEG Asymmetry

The existed research shows [41] that the left hemisphere of the brain is more associated with positive, forward emotions, and the right hemisphere is stronger associated with negative, withdrawn feelings. The Forehead EEG Asymmetry (FEA) is calculated as follows:

$$FEA=\frac{P_{FP1}\left(\beta \right)}{P_{FP1}\left(\alpha \right)}-\frac{P_{FP2}\left(\beta \right)}{P_{FP2}\left(\alpha \right)}$$

(4)

where P_FP1(β), P_FP1(α), P_FP2(β), and P_FP2(α) denote the average power of FP1 and FP2 electrodes corresponding to β and α waves, respectively.

3. Results

3.1. EEG Features Calculation

3.1.1. Normalized Power

In this study, power spectral characteristics were used to quantify the current brain load, the EEG power spectral density of the subjects was calculated and normalized at the temporal level, and the average power spectral density of 10 subjects at the same temperature was computed. From Figure 4, it can be seen that the normalized α wave does not vary a lot at each ambient temperature, and shows a slight increase at 26 °C and 30 °C. As the ambient temperature rises, the average power of the β wave representing excitement and tension shows a decreasing trend, and the θ wave representing sleepiness and fatigue shows a rising trend. The test results show that as the ambient temperature rises, slow EEG waves are produced in large quantities in the brain, making it easier to have emotions such as fatigue and negativity.

3.1.2. Alertness Level

The results of the AL change are shown in Figure 5. The alertness is more significant at 20 °C compared to other temperature levels when subjects are mobilizing a lot of attention to complete the trial. A one-way ANOVA with LSD post hoc comparison revealed that the level of alertness was significantly different across ambient temperatures (p < 0.05).

3.1.3. Frontal EEG Asymmetry

A higher FEA indicates a more positive mood, and the average asymmetry of the subjects was calculated to give a final score at each particular temperature [13]. We quantified the current state of brain emotional activity by calculating the asymmetry of the forehead EEG. As seen in Figure 6, the prefrontal asymmetry showed a concentrated decrease after 23 °C, and the FAE reached its minimum value at 30 °C.

3.1.4. EEG Wavelet Entropy

The EEG signal is a non-smooth, non-linear signal, and the complexity of the EEG signal can change when a person is thinking of activities with varying difficulty. The complexity of the EEG signal can be described using wavelet entropy [42]. The higher the wavelet entropy value, the greater the alertness and the higher the degree of arousal. The embedding dimension parameter of wavelet entropy m = 2. Given the similarity tolerance r = 0.2 × SD, where SD is the standard deviation of the sample, the total wavelet entropy values of 16 channels in the first half and the second half of the experiment were selected, and the difference was calculated to explore the change in brain complexity. The calculation results are shown in

Table 3. EEG signal wavelet entropy.

Environmental Temperature	Wavelet Entropy in the First Half	Wavelet Entropy in the Second Half	Wavelet Entropy Difference
20 °C	0.86 ± 0.31	0.78 ± 0.42	0.08 ± 0.11
23 °C	0.89 ± 0.38	0.82 ± 0.56	0.07 ± 0.18
26 °C	0.72 ± 0.19	0.61 ± 0.32	0.11 ± 0.13
30 °C	0.57 ± 0.21	0.44 ± 0.42	0.13 ± 0.21

: with the passage of time, the wavelet entropy decreases, the complexity of the EEG decreases, and the fatigue level increases. In the first half of the test, the wavelet entropy was the most significant at 23 °C, and the wavelet entropy difference was the most important at 30 °C. In the second half of the test, the wavelet entropy at 30 °C was the smallest, and the fatigue was the largest (p < 0.05).

3.2. Cognitive Outcomes

Table 4. Changes in subjects’ cognitive abilities.

Testing Capability	Test Time (s)/Number of Errors
	20 °C	23 °C	26 °C	30 °C
Thinking	259.4 ± 41.1/1.7 ± 1.8	264.8 ± 44.3/2.1 ± 1.1	260.6 ± 36.5/2.1 ± 1.5	291.4 ± 54.3/3.0 ± 1.8
Memory	197.4 ± 37.8/1.5 ± 1.3	204.6 ± 37.6/1.8 ± 1.8	226.8 ± 62.4/2.4 ± 1.3	233.1 ± 46.4/3.7 ± 2.8
Perception	191.1 ± 13.8/1.1 ± 0.9	204.2 ± 44.5/1.9 ± 1.5	194.6 ± 27.4/1.1 ± 1.3	217.6 ± 42.7/2.5 ± 2.0
Total	647.9 ± 62.8/4.3 ± 2.9	673.6 ± 83.5/5.8 ± 3.8	682.0 ± 88.3/6.1 ± 3.2	742.1 ± 108.6/9.2 ± 5.6
Vigilance	8.400 ± 0.596	8.785 ± 0.636	8.753 ± 0.854	8.288 ± 0.608

shows the test time and number of errors for the subjects at different ambient temperatures.

As shown in Figure 7, in the thinking ability test, there was no significant change in the test time and the number of errors at 20 °C, 23 °C, and 26 °C, while there was a substantial increase in the test time and the number of errors at 30 °C, when the thinking ability decreased. In the memory ability tests, the test time and number of errors increased with increasing ambient temperature, and higher temperatures reduced the subjects’ memory effects. In the perceptual ability test, the time at 30 °C showed a significant increase compared to the rest of the temperatures, with no substantial change in the number of errors.

In addition, the total time spent on the test items and the total number of errors was used to determine the cognitive level of the subjects. As shown in Figure 8, the total test time, and the number of errors for the thinking, memory, and perception items, showed an increasing trend. One-way ANOVA and LSD posthoc comparisons revealed significant changes in the total test time and the total number of errors at 20 °C and 30 °C (p < 0.05).

Reaction time can be used to assess a subject’s alertness. The reaction times at different ambient temperatures are shown in Figure 9. The reaction time was shorter at 20 °C and 30 °C compared to 23 °C and 26 °C. The reaction time of subjects increased and then decreased with the rising of ambient temperature, possibly because high temperature increases the secretion of norepinephrine, which is associated with alertness, causing a decrease in reaction time [10,31].

3.3. Subjective Questionnaire

The results of the subjective questionnaire for the subjects are presented in

Table 5. Subjective psychological feelings of subjects at different ambient temperatures.

Test Project	Scale Range	20 °C	23 °C	26 °C	30 °C	p
Thermal sensation	Unbearable cold (−3) → Unbearable heat (+3)	−1.5 ± 0.5	0.3 ± 0.5	2 ± 0.7	2.5 ± 0.5	**
Thermal comfort	Cold discomfort (−3) → Heat discomfort (+3)	−1.4 ± 0.7	0.4 ± 0.5	1.7 ± 0.5	2.9 ± 0.1	**
Environmental acceptance	Unacceptable (−1) → Acceptable (+1)	0.5 ± 0.2	0.9 ± 0.2	−0.5 ± 0.4	−0.9 ± 0.2	**
Dizzy	No (−3) → Yes (+3)	−2.7 ± 0.5	−2.2 ± 0.9	0.6 ± 0.8	2.2 ± 0.6	**
Fatigue	No (−3) → Yes (+3)	−2.2 ± 0.8	−1.2 ± 1.0	1.2 ± 0.9	2.4 ± 0.7	**
Sleepy	No (−3) → Yes (+3)	−2.8 ± 0.4	−2.1 ± 0.6	−1.3 ± 0.8	0.3 ± 0.7	**
Lack of Concentration	No (−3) → Yes (+3)	−2.6 ± 0.5	−2.1 ± 0.7	−0.2 ± 0.8	1.3 ± 0.3	**
Thinking ability	No (−3) → Yes (+3)	−2.7 ± 0.5	−2.3 ± 0.8	−0.7 ± 1.2	0.7 ± 0.5	**

. All test items changed significantly as the ambient temperature rose, thermal sensation and thermal comfort scores increased, and environmental acceptance scores decreased; subjects experienced increased dizziness, fatigue, sleepiness, lack of concentration, and decreased thinking ability. In Table 5, ** indicates p ≤ 0.05.

A third-order polynomial was used to fit the thermal sensory score to the ambient temperature, where R² = 0.899, and the fitted curve is shown in Figure 10. As the ambient temperature increased, the overall thermal sensation of the subjects showed an upward trend. When the ambient temperature was 20 °C, the subjects’ TSV was concentrated at −2 (50%) and −1 (50%); when the ambient temperature was 23 °C, the subjects’ TSV was focused at 0 (70%) and +1 (30%); when the ambient temperature was 26 °C, the subjects’ TSV were +1 (20%), +2 (60%) and +3 (20%); when the ambient temperature was 30 °C, the subjects’ TSV were concentrated at +2 (50%) and +3 (50%).

As shown in Figure 11, the NASA-TLX scores varied significantly (p < 0.05) across different ambient temperatures. As the ambient temperature increased, the subjects showed varying degrees of increase in brain load, time demand, effort, and physical demand.

We collected subjective questionnaires on participants’ job satisfaction and work willingness, and the score range was (−5, +5) representing the satisfaction and acceptance of work status. The numbers in the figure represent the subjects’ subjective questionnaire survey scores, with −5 indicating the worst acceptance and +5 representing the best acceptance. A higher score indicates higher acceptance.

The survey results on the subjects’ willingness to work and their job satisfaction are shown in Figure 12. The graph reveals that at 20 °C, 60% of the subjects had a high willingness to work (job willingness score (+3, +5)), and the remaining 40% felt that they had the idea to continue working (job willingness score (+1, +2)), while they were all satisfied with the performance of the task (job satisfaction score ≥ +1). There was no significant change in the willingness to work and satisfaction of the subjects at 23 °C compared to 20 °C. At 26 °C, there was a decrease in the subject’s willingness to work, as well as a perception that they were not doing a good job (negative job satisfaction score). At 30 °C the subjects’ willingness to work was negative, i.e., the subjects had the idea to escape from the environment and were not satisfied with the work (job satisfaction ≤ 0).

4. Discussion

4.1. Correlation Analysis

The Pearson correlation coefficient can be used to measure the linear relationship between variables and is an essential indicator for correlation analysis. In this work, Pearson analysis was used to calculate the correlation between ambient temperature and each parameter (two-tailed).

Table 6. Correlation analysis between ambient temperature and various parameters.

Vehicle Air Temperature	r	p
Test time	0.371	0.019
Number of errors	0.391	0.013
TSV	0.89	<0.001
Motivation	−0.52	<0.001
Mental Demand	0.91	<0.001
Physical Demand	0.95	<0.001
Temporal Demand	0.96	<0.001
Performance	0.66	<0.001
Thermal sensation	0.973	<0.001
Thermal comfort	0.954	<0.001
Environmental acceptance	−0.897	<0.001
Brain load	0.909	<0.001
FEA	0.453	<0.05
Vigilance	0.692	<0.05
Wavelet entropy	0.661	<0.05

shows the correlation analysis results of ambient temperature with different parameters. The analysis revealed that there was a significant negative correlation between ambient temperature and motivation and environmental acceptance. In contrast, we observed a positive correlation between vehicle air temperature and test time and the number of errors. There was a significant positive correlation between TSV, mental demand, physical demand, temporal demand, performance thermal sensation, and vehicle air temperature. Meanwhile, there was a significant and strong positive correlation between brain load and ambient temperature, and there was a weak positive correlation between FEA, vigilance, wavelet entropy and ambient temperature.

4.2. Correlation Analysis between EEG and TSV

4.2.1. Correlation Analysis between AL and TSV

The suitable ambient temperature can enhance the driver’s alertness, while high temperatures can make drivers irritable and other negative emotions that are not conducive to safe driving. At different ambient temperatures, the limit of the driver’s mood can be different. The higher the temperature, the greater the impact on the driver’s mood [39]. TSV is used to measure the driver’s sensitivity to ambient temperature. The changes in AL with TSV are shown in Figure 13. AL is optimal at TSV < 0. The AL declines as the TSV value gradually increases. At TSV = 2, AL shows a slight upward trend. The possible reason is that as temperature rises, fast EEG waves are produced which cause the brain to be excited. At TSV = 3, the AL drops, possibly because the temperature exceeds the body’s regulatory range and affects the state of the brain.

4.2.2. Correlation Analysis between EEG Wavelet Entropy and TSV

In the awake state, the driver’s brain is more active and the EEG signal is mostly at high frequency. In the fatigue state, the brain is in a comatose state, and the EEG signal is mostly at low frequency. To investigate whether the subjects’ heat sensation affects brain load, this paper analyzed the relationship between heat sensation scores and wavelet entropy. According to the analysis of the experimental results, the wavelet entropy of EEG signals in different TSV states is significantly different. The correlation analysis between EEG changes and TSV is shown in Figure 14. From the figure, it can be seen that the EEG signal is more active when TSV = 0, and its wavelet entropy is at a higher level. When TSV > 0, the wavelet entropy value decreases significantly, and the driver is prone to significant drowsiness, fatigue, and inattention at this time [43].

4.3. Correlation Analysis between Cognitive Ability and TSV

In this paper, it is found that the cognitive ability of subjects was at its best when the ambient temperature was 20 °C. This is in contrast to some studies conducted in indoor thermal environments, which concluded that indoor cognitive performance was the best at 26–27 °C [9,41]. Differences in space size, the intensity of personnel activity, the nature of labor, and the thermal resistance of clothing leads to differences in the thermal neutral temperature between indoor and vehicle environments. Drivers need to maintain a high concentration level when driving for long periods. Lower interior temperatures can promote activation of the pituitary-adrenal cortex and sympathetic nerves, increase cortisol and catecholamine secretion, and enhance the cognitive ability of drivers.

In order to investigate whether subjects’ heat sensation affects their cognitive ability, this paper further analyzed the relationship between heat sensation scores and mental ability, as shown in Figure 15. The cognitive test time and the number of errors were the most significant when the heat perception score was +2 and +3. This is similar to the findings of some studies where heat dissatisfaction due to high temperatures significantly reduced cognitive speed [28,38]. When the subject’s heat sensation is high, it is often accompanied by mental phenomena such as distraction, dizziness, and fatigue, thus increasing the time spent on cognitive ability tests and the number of errors.

As can be seen from Figure 10 and Figure 15, the thermo-neutral (TSV = 0) vehicle environment was 23 °C when the cognitive performance was not optimal. Therefore, for the driver to maintain good cognitive performance, it is recommended that a slightly colder (20 °C–23 °C) interior environment should be kept while performing driving tasks.

5. Limitations and Future Work

Although this investigation was carried out in relatively controlled circumstances, some potential limitations were encountered. In driving tests, solar radiation affects the average radiation temperature. To better control for the relevant variables, we assumed that the air temperature and the mean radiation temperature were equal, which is not realistic. Air inequality and air flow rate will affect the local discomfort of the driver. Furthermore, the participants were predominantly young and all graduate students, which differed greatly from the level of education, proficiency, thermal acclimatization, etc., of the operators. However, although the subjects all youngsters, the results of this study can also provide some basis for changes in the cognitive abilities of young drivers in driving safety. At last, despite the sample size being sufficient to investigate EEG activity, to be able to promote the consequences to the general population, larger sample, female subjects, and solar radiation and air inhomogeneities should include in future studies. It is necessary to study the effect of lower internal vehicle temperature on driver EEG. This helps us explore how hypothermia is induced in the driver brain. In the future work, we will explore more about the lower temperatures (<20 °C). Following this method, targeted and individualized vigilance management strategies can be established to deliver practical benefits in the management of in the management of safe driving.

6. Conclusions

In order to investigate the effects of EEG-based vehicle thermal environment on the driver’s cognitive ability, a simulated experimental driving platform was established. This experiment was used to obtain the physiological data, subjective perception, and cognitive ability results of drivers under different ambient temperatures. The conclusions are as follows:

• Lower vehicle interior temperatures increase driver alertness and arousal. As the ambient temperature rose, the normalized power of θ activity and α activity increased, and the vigilance and frontal EEG asymmetry declined. At 20 °C, the β and α wave components of the EEG signal strengthened, and the level of alertness (AL) and frontal EEG asymmetry (FEA) elevated. At 23 °C, the wavelet entropy values of the two segments before and after the experiment were the largest, and the wavelet entropy difference was also the greatest.
• The drivers’ cognitive test time and the number of errors increased, while their cognitive ability decreased with the rising of ambient temperature. At 20 °C, the driver could maintain a high level of cognitive ability, and at 23 °C, the subjective thermal comfort of the driver was optimal. Therefore, to achieve high cognitive performance and thermal comfort, the vehicle’s ambient temperature should be maintained in the range of 20–23 °C.
• As the ambient temperature rose, the thermal sensation, thermal comfort score and brain load increased in the subjective questionnaire; and environmental acceptance, job satisfaction and willingness to work decreased. At uncomfortable ambient temperatures, drivers experienced the uncomfortable states of dizziness, sleepiness, and weakness.
• The correlation analysis showed that there was a significant negative correlation between ambient temperature and motivation and environmental acceptance (p < 0.001). In contrast, there was a positive correlation between vehicle air temperature and test time and the number of errors; there was a significant positive correlation between TSV, mental demand, physical demand, temporal demand, performance thermal sensation, and vehicle air temperature (p < 0.001); there was a significant and high positive correlation between brain load and ambient temperature (p < 0.001); there was a weak positive correlation between FEA, vigilance, wavelet entropy and ambient temperature (p < 0.05).

To ensure the safety of the subjects, the test in this paper was carried out in the simulated driving platform. In future, some experiments will be carried out on a real road. At the same time, more experiments will be carried out in a colder environment (<20 °C) and with a longer driving time. Based on the physiological parameters and subjective perception data of this paper, the physiological and psychological mechanism of the change of vehicle environment temperature to the driver’s cognitive ability will be further analyzed.