The passenger’s comfort is of vital importance in today’s car, even more so in the future autonomous car. Its thorough analysis requires the use of objective measures, but the subjectivity of the passenger must also be considered. The relationship between both aspects is an uncommon research avenue, and could greatly benefit from the use of a vehicle simulator. This study use a simulator to replicate a journey with diverse characteristics that numerous passengers may experience (including engaging in different activities). The appropriate variables for a proper comfort analysis are determined in this study, resulting in a comprehensive database for the same purpose. Additionally, a passenger comfort based questionnaire is developed and applied to obtain the subjective assessment of different passengers. The results of these questionnaires help not only in the identification and study of the self perception of comfort by the passengers but also in the use of the simulator for comfort experiments. The detailed comparison between both the objective approach of signal analysis and the subjective approach of questionnaires that is performed thanks to the simulator creates a foundation for future research and narrows the existing research gap.
Hinweise
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
1 Introduction
In the era of autonomous cars, where humans relinquish control to technology, ride comfort takes on an increasingly vital role. As these self-driving vehicles revolutionize transportation, it is essential to prioritize passenger comfort and well-being while avoiding malaise like motion sickness. Ensuring a comfortable ride directly impacts user experience, safety, and the broader acceptance and adoption of autonomous vehicles [1]. This task is challenging, since comfort is a subjective sensation which changes depending on the person even if the utilized vehicle or route are the same. Nevertheless, there are standards that allow us to objectively study and evaluate this sensations. Knowing that both general comfort and motion sickness can be analyzed objectively and subjectively, researching a link between these aspects seems the next logical step.
Most of the objective comfort evaluation analysis are the results of the research done in the ISO2631 Standard [2] and from the Handbook of Human Vibration [3], where detailed metrics are defined in order to study the effects of vibration. However, as it is already mentioned, the severity and onset of ride comfort can vary widely among individuals, and the factors triggering it are often deeply personal. In order to take this into account, assessments tools are key, such as questionnaires, interviews or reports of subject experiences. In this work the subjective experiences related to the lack of comfort causes of the passengers will be explored, with a particular focus on motion sickness.
Anzeige
These questionnaires often employ Likert scales, ranking systems, or open-ended questions to elicit responses that shed light on the contributing causes, and are designed to capture a range of information, including the severity of symptoms, the duration of exposure to motion stimuli, and individual predispositions or triggers. Common scales and questionnaires include the MIsery SCale [4], the Motion Sickness Susceptibility Questionnaire [5] and the Simulation Sickness Questionnaire [6].
In [7‐12] the different biological or physiological factors that have a notable impact on the personal self perception of motion sickness are described attending to the answers of questionnaires. They concluded that factors that have the highest impact on motion sickness are: gender, lifestyle, being blind/mute, and suffering from migraines. However, these articles mostly study the physiological and social factors of motion sickness without subjecting the passenger to a specific path and do not take any objective metric into account.
Broadly, research that combines both the subjective experience of the passenger via questionnaires and the objective evaluation with parameters defined in the ISO2631 in order to assess comfort and motion sickness both ways is severely lacking. In [13], car accelerations and a variety of signals are analyzed but while the test subjects also offer insight of their subjective evaluation, both areas are not studied together. Craig C. Smith, David Y. McGehee and Anthony J. Healey offer in [14] a complete study that uses the ISO2631 with the UrbanTracked Air Cushion Vehicle (UTACV) Specifications to assess the risk of personal discomfort and motion sickness. This research also included other tools such as spectral vibration analysis.
The combined subjective and objective evaluation is complex due to the impossibility of replicating identical driving conditions for many different passengers, situations and routes. Under normal circumstances, drivers are exposed to numerous factors that are not under researcher control such as accidents, weather, road conditions, their actions, etc.
Anzeige
In this sense, a vehicle simulator can be a very useful tool in evaluating comfort. Driving simulators play an important role in research on vehicle dynamics, human comfort factors, and the development of new advanced driver assistance systems [15]. These platforms allow testing to be conducted at a much earlier stage of the development process at a lower cost, which means the vehicle is closer to production when physical prototypes are manufactured. The simulator then becomes an integral part of the vehicle development cycle as it provides a natural link between computer modeling phases, laboratory testing, and ultimately, the test track. If this specific simulator is able to correctly reproduce, emulate and link the routes used for data acquisition, a new experimental standard for comfort evaluation can be developed since reproducing a real modelled route could negate all external factors while shedding light on the subjective analysis of ride comfort.
The primary aim of this of this work is to accomplish an all-encompassing assessment of ride comfort, seamlessly combining subjective and objective dimensions. The comprehensive approach involves three key objectives: Firstly, to ensure the capability for precise emulation of the vehicular simulator and establish a specialized database to facilitate comfort analysis. Secondly, the application of standardized criteria in accordance with ISO standards, objectively evaluating motion sickness. Finally, the design of a comfort specific questionnaire and its administration during a experimental campaign, facilitating the subjective validation of the research. The final integration and comparison of all of these results lays the groundwork for a combined analysis of motion sickness.
To achieve this goal, main objective and subjective comfort evaluation methods used during the work are presented in Section 2. Section 3 describes both the vehicle utilized to acquire the data and the simulator used to replicate the driving scenario. The route specifications and data acquisition details are presented in Section 4, along with a brief summary of how the virtual scenario in the simulator has been developed to recreate the route. In Section 5, we evaluate the progress made and analyze the variables that will form part of a comprehensive database to objectively determine passenger comfort in the vehicle. Moreover, in Section 6, we conduct a quick analysis of objective comfort variables summarized in Section 2 and driver actions by zone are analyzed, building on the authors’ previous work [16, 17]. In Section 7, we analyze the results of the Questionnaires and its implications, comparing them to the previous objective approach. This analysis further validates the entire system we have developed. Finally, conclusions are presented in Section 8.
2 Ride Comfort Evaluation Methods
2.1 Objective Analysis via ISO-2631
The subjective nature of comfort poses a challenge when attempting to quantify and measure it objectively, however, due to this, a lot of research was conducted in order to present standard evaluations.
ISO-2631 is a standard that provides guidelines for the evaluation of human exposure to whole-body vibration and mechanical shocks. Moreover, it includes various objective metrics for general comfort evaluation that have been used in many research articles such as Vibration Dose Value (VDV) in road roughness analysis [18], or the Crest Factor (CF) as one of the factors to optimize in suspension design [19].
In this work, we will focus on specifically a subset of general comfort, which is motion sickness. Motion sickness, as defined academically, is a physiological reaction characterized by symptoms such as nausea, vomiting, dizziness, and discomfort. It occurs when there is a sensory conflict and vibrations, leading to an imbalance in the perception of motion. In the ISO2631 it is described that while certain frequencies are completely mitigated by our body, several ranges of vibration frequencies are not, and in fact are the responsible for causing maladies such as motion sickness (e.g. 0.1 Hz-0.3 Hz).
More importantly, it describes the metric that will be used in this article, the Motion Sickness Dose Value (MSDV). While this standard only provides information about using it for signals related to the vertical axis, our previous article has successfully applied it in the longitudinal and lateral axis with good results [16]. Furthermore, in [17] we define the criteria observed in Eq. 1 which combines the effect of the X and Y axis. Here, \(a_x(t)\) represents the longitudinal acceleration. After applying a motion sickness pondering signal filter described in the ISO2631 to the acceleration, \((a_{x, wf}(t))\) is obtained. The signal is integrated to take into account time exposure. The other half of the equation is the exact same but for the lateral (Y) acceleration signal.
To comprehensively evaluate the simulator experience for our test passengers, we have developed a new questionnaire with a multifaceted approach. We have based on previously designed questionnaires as [4‐6] but in order to effectively used the questionnaire as a tool in this experimental campaign, it has to be short and concise enough so that it is as less intrusive as possible while obtaining as many meaningful answers and insights as possible.
Leveraging the simulator’s capabilities, each driver will be exposed to each of the zones while being asked questions before starting, after each zone and finally after the simulation is over. The objective is to obtain information that allows us to compare the reactions and perception of each passenger for each different zone of the simulation. This information can then be used to draw conclusions regarding the route and its effects.
The questionnaire is presented in Table 1. The initial questionnaire is administered prior to the start of the simulation. This phase is instrumental in obtaining demographic information such as Age, Gender, and lifestyle details. Additionally, attending to the aforementioned biological and physiological factors that influence motion sickness, the questionnaire includes inquiries in this sense.
Table 1
Questionnaire designed for our experimental campaign
Questionnaire 1: Pre-simulation questionnaire
1. Age, gender
2. Perceived susceptibility of motion sickness now. [1-5]
3. Perceived susceptibility of motion sickness when you were a child. [1-5]
4. Likelihood of vomiting due to motion sickness when you were a child. [1-5]
5. Frequency of sports or physical activities per week. (None, Once, Twice, Three times or more)
6. Use of transportation during the week.
(None, Sometimes, Several Times, Everyday)
Questionnaire 2: Post-zone questionnaire
1. Amount of time needed to recover from the sensations induced by the simulation.
(Less than a minute, One to Two minutes, Two to Five Minutes, More Than Five Minutes)
2. Overall score of the zone regarding motion sickness. [0-10]
3. Symptom assessment following the MISC scale. [0-10]
4. Order the following factors by their impact on your motion sickness: (Throttle, Brakes, Steering Wheel and Verticality). [1-4]
5. Overall score for the impact of the Throttle. [1-5]
6. Overall score for the impact of the Brakes. [1-5]
7. Overall score for the impact of the Steering Wheel. [1-5]
8. Overall score for the impact of the Verticality. [1-5]
Questionnaire 3: Post-simulation questionnaire
1. Rank each travelled zone by their level of motion sickness. [1-6]
2. Additional information (Optional)
It can be devided into three separate parts, a pre-simulation questionnaire, a post-zone questionnaire and a final post-simulation questionnaire
Subsequently, the second questionnaire is conducted after each simulation zone, focusing on evaluating the characteristics of the specific zone and its impact on the passenger. This questionnaire encompasses questions about a general comfort score, a rating of motion sickness symptoms utilizing the MISC scale [4], and the recovery time following exposure to the simulated environment. Furthermore, it seeks the participant’s perception regarding various driving factors, including Brakes, Throttle, Steering Wheel, and the Verticality (inclination of the platform) of the simulation. The aim is to discern the influence of these factors on motion sickness. Lastly, participants are instructed to rank these factors based on their perceived importance in motion sickness generation for each of the zones.
The final questionnaire is administered upon the completion of all simulations. Participants are required to rank all zones based on their perception of motion sickness generation, ranging from the zone inducing the highest motion sickness to the one with the lowest impact. Furthermore, the passengers can write additional information that they deem necessary about the simulation itself and the experimental campaign. This approach ensures a systematic and insightful evaluation of the participant’s experience throughout the simulation study.
3 Dynamic Correlation Platforms
The tests have been carried out in two different platforms at Automotive Intelligence Center (AIC): a real car and a driving simulator. Both of them are equipped for experimental testings within the standards of the Automotive Industry.
3.1 Real Vehicle
The data acquisition experiment has been done using a 3rd generation sensorized Hybrid Toyota Prius, which can be observed in Fig. 1.
The representative variables of the vehicle are obtained through the OBD system by identifying the CAN bus data frames: Rotational speeds of the vehicle (RR, YR, PR), Accelerations (ACC_1, ACC_2, ACC_3), Vehicle Speed (VS), revolutions of the motor per minute (ERPM), Steering Wheel Angle (SWA), Steering Wheel Rotation Speed (SWRS), Brake Pedal (BP) and Gas Pedal (GP). These last six variables are variables of which the driver has complete control and a very important factor in controlling comfortable scenarios. They will be referred to as actuable or interpretable variables throughout the article
On the other hand, Triaxial IEPE Seat Accelerometer (Type 4515-B), located in the driver’s seat, is used to capture the accelerations received by the driver during the ride (Acc_x, Acc_y, Acc_z).
Fig. 1
The Toyota Prius used in the experimental data acquisition campaign
×
Using CAN Bus allows us to safely and non intrusively obtain information from the internal computer of the car in real time. However, the biggest challenge is to actually decode and obtain the information from the specific identifiers of the data. Besides, due to the internal message sending protocols, some messages are slightly delayed depending on the influx of messages of the bus itself. This results in the data not having a uniform sampling rate, which is fixed by processing the data to interpolate points between samples.
3.2 Driving Simulator
The driving simulator is a six-degree-of-freedom motion system based on a Stewart platform scheme that is shown in Fig. 2. Stewart platforms are of great interest in automotive research and development, especially in vehicle simulation and dynamic analysis applications [20]. The system consists of a dynamic upper platform connected to a static lower platform by six electric actuators.
Fig. 2
Photo of the Stewart patform scheme
×
The simulator works with IPG-Carmaker in cosimulation with Matlab-Simulink in order to control the motion platform. This way both software-in-the-loop (SIL) and hardware-in-the-loop (HIL) configurations are possible, as well as driver-in-the-loop (DIL) by introducing the human factor as driver or passenger. In terms of computational capabilities, the latest technological improvements implemented allow it to work in a Hard Real-Time (HRT) environment. HRT systems offer the capability to develop projects in vehicle simulation environments, including vehicle dynamics applications and advanced user monitoring studies for driving and comfort applications in autonomous driving scenarios. In order to carry out the implementation, it is necessary to configure the simulation environment with the optimal hardware equipment. This configuration uses the modular dSPACE SACALEXIO LabBox system which is compatible with the software arquitecture previously built. This system is highly scalable and provides high-performance processor technology for demanding real-time requirements as well as complete, accurate and fast I/O capabilities. Its automotive applicability and powerful real-time technology are ideal for the most demanding applications, where highly automated and autonomous driving is possible [21].
Fig. 3
Travelled route of this experimental campaign
×
The variables of the vehicle model are obtained directly from the software. The most representative ones for the study have been acquired: simulator accelerations (\(ACC\_Xs\), \(ACC\_Ys\), ACC_Zs), velocities and positions (both linear and rotational) around the centre of gravity (RRs, YRs, PRs), pedal positions (GPs, BPs) and steering angle and rotation speed (SWAs, SWRs).
Besides, the same Triaxial IEPE Seat-Accelerometer (Type 4515-B) used in the real car is located in the passanger seat, measuring the acelerations named Acc_xs, Acc_ys, Acc_zs.
4 Experimental Development Specifications
With the objective of performing an exhaustive comfort-based analysis, a complete diverse route has been designed. The car will travel this route so that we can obtain our dataset and at the same time obtain the necessary information so that the route is replicated within the simulator. Once the route is emulated, the route can be adequately simulated for different passengers.
The wide range of acquired variables enables the correlation of objective data from both real and simulated vehicles, with the ability to later obtain the personal assessment of the participants. This unique aspect, absent in other databases within this field, paves the way for new and significant avenues of research.
4.1 Route and Acquisition Specification
The journey, comprising segments of diverse nature, should be long enough to provide an ample number of samples for constructing a suitable database and subsequent utilization in machine learning techniques. Additionally, it is important to ensure a similar number of samples per segment to facilitate optimal learning and deriving meaningful conclusions.
Data was acquired in the Basque Country, Spain, around the city of Leioa which features a variety of road types. The complete route of Fig. 3 can be separated in 6 different zones (Zone0, Zone1, Zone2, Zone3, Zone4, Zone5), of approximately two kilometer each, labeled according to the type of road (i.e single or multi-lane roads, featuring urban, highway or interurban areas) with varying degrees of elevation, maximum speeds and road roughness.
Another crucial factor in data delimitation is the window size, which refers to the number of samples required to calculate the variables that assess comfort. This primarily depends on the data acquisition system.
In order to select the optimal window size, a comprehensive approach combining experimentation, human factor consideration, and expert knowledge has been employed. Considering the ultimate objective of providing recommendations to enhance driving, a system that generates guidelines every few seconds would be uncomfortable and inefficient for the driver. However, it’s important to consider that smaller windows will generate more samples, which can be beneficial for future model training.
Initial investigations explored a wide range of window sizes, spanning from 2 to 10 seconds. Participant feedback from the study, the research team’s prior experience with similar topics and familiarity with simulator signal characteristics made the 8-second window the optimal choice, striking a balance between capturing sufficient motion data for accurate comfort assessment and maintaining a useful update frequency for real-time communication. The selection of a 4-second overlap was motivated by the aim to enhance data continuity, provide smoother transitions between the recommendations and to ensure that no critical motion events are missed at the boundaries of the windows.
4.2 Virtual Scenario Building
Designing the scenario within the simulator has two different stages, the first one consists of reproducing the road as accurately as possible in terms of dimensions and roughness (critical from a vehicle dynamics point of view) and the second one regarding the placement of external objects (e.g. buildings or traffic) which is crucial for the immersion of the user.
Road design is handled on two different points of view: macroscopic (dimensions) and microscopic (roughness). Concerning dimensions, the route is set via Google Maps where it can be exported as a .gpx file (GPS Exchange format). This file contains a point list of geographic coordinates which can be imported and edited. The raw data obtained has to be smoothed in order to solve errors encountered that are related to repeated points, sharpness in the “XY” plane, sharpness in the “Z” coordinate and wrong points that are outside the route. On the othe hand, road roughness can be defined according to the spectral density level (PSD) [22]. ISO 8608 [23] defines an analytical expression for the PSD of the road profile. In our case each road section is calculated experimentally through acceleration correlation.
The value of the roughness corresponds to road class according to regulation. In our case each road section is calculated experimentally through acceleration correlation.
Finally, the scenario editor in IPG CarMaker is used to include road markings, traffic signs, traffic lights, terrain, buildings, vegetation and traffic, as well as weather and light. The aim of this stage is not to obtain a one-to-one representation but an immersive one.
Figure 4 shows two examples of the immersive experience of the emulated route.
Fig. 4
Example of the emulated urban and highway area respectively
×
5 Validation of the Developed Acquisition Systems and Analysis of the Measured Variables
To assess the obtained data, we have statistically analyzed the different variables and their relations to each other by calculating their distributions, correlations and observing their time series.
To properly assess our data with a objective criteria of comfort, it is essential to correctly measure the accelerations, and thus, to understand and research the usefulness of the position of the accelerometer inside the car and its implications (seat accelerations, center of mass accelerations, internal car sensor accelerations). Therefore, the first analysis of the car signals has been focused on linking the different groups of accelerations: seat accelerometer signals in the car (acc_x, acc_y, acc_z), CAN Bus decoded signals (ACC1, ACC2, ACC3), acelerations from the software that emulates vehicle (ACC_Xs, ACC_Ys, ACC_Zs) and seat accelerometer signals in the simulator (acc_xs, acc_ys, acc_zs).
Fig. 5
Comparison of the CAN bus longitudinal acceleration (ACC_X), the car seat accelerometer longitudinal acceleration (acc_x), the simulator emulated car longitudinal acceleration (ACC_Xs), and the simulator seat acceleration (acc_xs)
Fig. 6
Comparison of the car Gas Pedal (GP), the derivate of Vehicle Speed (VS), Longitudinal acceleration of the CAN Bus (ACC_X) and car seat longitudinal accelerometer signal (acc_x)
×
×
Figure 5 shows the comparison of the longitudinal accelerations for the first part of the route. It is found that ACC1 is in fact ACC_X as it perfectly matched the other signals related to the longitudinal acceleration. The seat accelerometer of the car (acc_x), the longitudinal acceleration obtained via CAN Bus (ACC_X), and the longitudinal acceleration of the driving simulation (ACC_Xs) are perfectly linked. On the other hand, when studying the signals of the seat accelerometer of the simulator (acc_xs), it is clear that the signals did not follow the expected values. Moreover, we observe how the patterns ranging from -2 to 2 meters per second squared of the longitudinal acceleration cannot be recognized in any shape or form in the accelerometer. It has been concluded that the mechanical rotational movement of the seat of the simulator cannot transfer linear acceleration values, since the movements of the platform try to emulate the real sensations of the car not its movements per se. While the signals obtained from this accelerometer (acc_xs, acc_ys, acc_zs) don’t correlate with real vehicle accelerations and thus will not be of use moving forward, they provide valuable insights into the simulator’s limitations in replicating actual driving dynamics. This finding is crucial for interpreting differences in simulated and real driving scenarios, and for improving future simulator designs.
At the same time, thought ACC2 and ACC3 seem to match the lateral and vertical accelerations of the car, upon further analysis has found that while its correlation is very high in some zones, it is nonexistent in others and thus we cannot use them as reference.
It is also interesting to study with a comfort-based perspective, the influence of variables related to driver actions, which is obtained by decoding the CAN Bus (GP, BP, VS, SWA, SWRS, ERPM, YR) or from the vehicle model of the simulator (GPs, BPs, VSs, SWAs, SWRSs, YRs) .
Fig. 7
Comparison of the car Steering Wheel Angle (SWA), the derivate of the Steering Wheel Angle (SWRS), Vehicle Speed (VS) and car seat lateral accelerometer signal (acc_y)
Fig. 8
Comparison between the normalized car Yaw Rate (YR), Steering Wheel Angle (SWA) and seat lateral accelerometer (acc_y). A time delay can be observed between the seat lateral acceleration and the car signals
×
×
Figures 6 and 7 showcase the most important driver controlled variables that can be linked to the longitudinal and lateral dynamics of the vehicle. Figure 6 shows the statistical analysis of the variables related to the longitudinal acceleration (X). Each main variable is represented in the first line as a distribution plot, Gas Pedal (GP), the first derivate of the Vehicle Speed (d1_VS), CANBUS decoded longitudinal acceleration (ACC_X) and car seat longitudinal accelerometer (acc_x), and then the most representative regression plots between variables are observed in the second line, r shows the Pearson Correlation Coefficient, with the number of stars next to the number being the obtained p-value (* = p<0.05, ** = p<0.01, *** = p<0.001). As it can be observed, the GP, the derivative of VS and the ACC_X all have positive to extremely positive correlations.
Since we do not have a confirmed lateral acceleration from CAN Bus, the analysis has been done using only seat accelerometer with variables related to lateral dynamics such as SW, SWRS and VS. As it can be seen in Fig. 7, signals related to lateral motion also show high correlations between them, with SW and VS having a negative correlation. However, with the SW being the primary and direct enabler of lateral acceleration in the car, the positive correlation between the SW and the accelerometer is lower than expected.
In order to better understand the relation between SW and acc_y, after a thorough analysis of different zones, it is observed that the accelerometer is out of phase, not due to the data acquisition systems but due to the fact that the body movements of the driver are anticipating the different turns of the route as it can be observed in Fig. 8. It can be deduced that in autonomous car, where the role of the participant is that of a passenger, this effect would not appear.
Fig. 9
Comparison of the simulator Gas Pedal (GPs), the derivate of Vehicle Speed (VSs), and the longitudinal acceleration of the emulated vehicle (ACC_Xs)
Fig. 10
Comparison of the emulated Steering Wheel Angle (SWAs), the derivate of the Steering Wheel Angle (SWRSs), Vehicle Speed (VSs) and the lateral acceleration of the emulated vehicle (ACC_Ys)
×
×
As we can see in Figs. 9 and 10, the obtained correlations are very similar to the ones obtained for the same car variables, so, the simulator emulates the driver actuable variables really well. Regarding the longitudinal axis, all correlations follow the same trends that are observed in the real car. In the lateral axis, the correlations are higher, as the internal SW and ACC_Ys are perfectly linked with no delays.
Fig. 11
Comparison of the CANBUS decoded angular motion of the Yaw Rate (YR) and the Yaw rotation velocity of the emulated vehicle (YRs)
Fig. 12
Comparison of the driver actuable variables of the car and simulator, the Brake Pedal (BP), the Gas Pedal (GP),and the Steering Wheel (SWA)
×
×
The fact that the simulator follows the same trends as the car shows promising results. However, with the goal of having a comfort-inclusive simulator, we need to be meticulous in the signal comparison and strictly compare different variables to confirm that the driver variables that can be linked to motion sickness are perfectly emulated in the simulator.
In Fig. 5 we compare the longitudinal axis, but we cannot strictly repeat this process for the lateral motion since we have not accessed the lateral acceleration of the CAN Bus. However, the comparison regarding lateral motion will be done with the Yaw Rate (YR and YRs) since it is a signal that was decoded for both the CAN Bus of the car and software of the driving simulator. As we can observe in Fig. 11, both angular velocities are very similar. The observed differences are attributed to the fact that the actions of the simulator are limited by its actuators and some spikes observed in the car are not properly emulated for either mechanical or signal limitations. Moreover, the differences that start around the 140-second mark are linked to the emulated route, which is slightly shorter than the real one and therefore it starts being slightly out of phase the longer the route is. This is also the reason why it stops at 185 seconds.
After having observed the motion dynamics of longitudinal and lateral movements, the next comparison is to observe if the variables related to driver actions are maintained. Figure 12 shows the comparison of the Brake Pedal (BP), Gas Pedal (GP), and Steering Wheel Angle (SWA) between the CAN Bus decoded signals and the driving simulator (BPs, GPs and SWAs). Again, the similarities are evident, BP and GP seem to perfectly match, with differences in magnitudes being attributed to the fact that the motor of the emulator is not the exact same as the car used in the experimental campaign, and the differences in the SWA variable being the same as those previously observed in Fig. 11.
After a strict analysis of all the available signals, the presented results are the following:
The accelerations obtained from the seat of the simulator can not be associated with any other signal, rendering them useless for comfort-based analysis.
Variables related to driver actions like BP, GP, SWA, SWRS, and Vs, are correctly emulated. However, due to having a different version of the car motor, some signals are overly saturated in order to maintain the speed profile of the original car (It can be observed in Fig. 12). Moreover, although SWAs and YRs are analogous to their respective car signals, both show limitations due to the mentioned restrictions in the actuators.
Regarding longitudinal acceleration (X), we can perfectly relate the acceleration of the seat of the car (acc_x), the acceleration obtained via CAN (ACC_X), and the acceleration values of the emulated vehicle (ACC_Xs).
Regarding lateral acceleration (Y), not having a CAN Bus decoded ACC_Y and factors like the driver anticipating the curves makes it difficult to directly correlate the other signals. Nonetheless, the relation of lateral angular velocities and steering wheel validates the good emulation.
6 Objective Comfort Evaluation Using Simulator Signals
Our previous rigorous analysis demonstrates that the route generated in the emulated environment is not the exactly the same and the emulated variables are useful. For the future subjective analysis, passengers will be subjected to the emulated route, so, the accelerations of the simulator (ACC_Xs, ACC_y, ACC_z) and simulated driver actions (GPs, VSs, SWAs, SWRSs) have been chosen as valid candidates for comfort analysis.
To briefly study and infer information related to comfort, we have applied some of the techniques we have previously used in [16] and [17] with success.
In Table 2 the mean value and standard deviation of calculated \(MSDV_{x}\) and \(MSDV_{y}\) for each of the specific zones are shown. The general criteria defined to evaluate motion sickness in accordance to the ISO-defined metric, MSDVxy, is also calculated. Additionally, in Fig. 13, the obtained kernel density distribution based on those data is shown. Different zones obtain completely different distribution centers. These different density shapes are related to the type of driving patterns, situations and road factors of each of the areas.
Table 2
Table showing the mean and standard deviation of the MSDVx, MSDVy and MSDVxy values for each of the zones
MSDVx
MSDVy
MSDVxy
Zone
Mean
Std
Mean
Std
Mean
Std
0
3.081
1.029
1.522
2.494
4.009
1.701
1
2.422
0.960
2.204
2.157
3.580
1.836
2
3.322
1.087
1.960
1.986
4.099
1.774
3
1.844
0.667
0.785
0.873
2.142
0.788
4
2.248
0.687
1.209
1.285
2.787
0.916
5
1.689
1.674
2.647
2.134
3.249
2.578
Examining Table 2 and Fig. 13, Zones 0, 1, and 2 stand out with their shared characteristics, displaying the highest MSDVxy values among all zones. Remarkably, in each case, the dominance of MSDVx over MSDVy is clear. This pattern can also be observed with the lateral component even though the values in the density distribution. On the contrary, Zones 3 and 5 are characterized by showcasing lower values of MSDVx. Moreover, Zones 3 and 4 also have low MSDVy values. Zone 3, apart from having the lowest values in both directions it also has the lowest deviation of all zones, which can be seen in the shape of its kernel distribution. Zone 5, in particular, stands out not only for having the lowest MSDVx but also for its skewed distribution towards the lowest values, coupled with the largest standard deviation among all zones. Which causes Zone 5 to also diverge significantly in its kernel distribution.
7 Subjective Evaluation via Questionnaire
In this section, following the responses of the questionnaire, a comparative analysis of distinct zones has been conducted based on the discomfort experienced by passengers in the simulation. Figure 14 illustrates the outcomes pertaining to General Score (a parameter that conveys the general feeling of discomfort and motion sickness on a scale from 1 to 10) , MISC score (which has its own set of scale based on motion sickness symptoms), and recovery time (the time it takes for the symptoms to disappear).
It can be deduced that even though passengers scored different medium General Score values for each zone, their physical symptoms were low, and thus, recovery time was almost nonexistent for a lot of the passengers. Since the route is composed by zones with significative different levels of motion sickness, it seems that the simulator itself attenuated certain events.
Fig. 13
(a), (b), (c), (d), (e), (f) Kernel density plots of MSDVx and MSDVy values of all six different travelled routes from Zone 0 to Zone 5 respectively
Fig. 14
Recovery time, general score and MISC score of the different subjects for each of the different zones
×
×
Regarding the results, Zone 5 stands out as the clear outlier, with a higher than 7 average General Score, it seems that it is generating a lot more motion sickness than all other zones. Zones 0, 1 and 2 can be discerned as zones with medium values, while Zones 3 and 4 score the lowest.
In order to shed light into the overlapping of standard deviations of the General Score of Fig. 14, the ranking of the zones made by the passengers in the final questionnaire has been analyzed. This result can be observed in Table 3. This table represents how the passengers ordered the zones depending on the motion sickness level felt during the simulation. A zone scored as (1) entails the most uncomfortable zone whereas the most comfortable is rated a (6). Therefore, the percentage of passengers that have assigned a zone with the same valoration is noted.
Table 3
Percentage of passengers that rank each zone with each specific score, where 1 signifies being the most uncomfortable and 6 indicates the most comfortable
Score (%)
1
2
3
4
5
6
Zone 0
0%
25%
25%
25%
12.5%
12.5%
Zone 1
12.5%
25%
12.5%
0%
12.5%
37.5%
Zone 2
0%
0%
25%
0%
50%
25%
Zone 3
0%
37.5%
0%
25%
12.5%
25%
Zone 4
0%
0%
37.5%
50%
12.5%
0%
Zone 5
87.5%
12.5%
0%
0%
0%
0%
Zone 5 has been noted primarily as a (1), resulting in being by far the worst qualified zone. Half of the passengers label Zone 4 and Zone 2 with a (4) and a (5) respectively with other passengers evaluating both of these zones with scores close to each of these, showing a generalized consensus. However, Zones 0, 1 and 3 have disperse score distribution, indicating discrepancies between passengers and even having completely opposite evaluations.
The expected rank of each area can be represented by an averaged motion sickness ranking, which is obtained by weighting the obtained scores. The results of the Expected Motion Sickness Score can be observed in Table 4, where clearly it can be concluded that Zone 5 is by far the most motion sickness inducing zone and Zone 2 the least, but others are very similar to each other. The order of zones by EMSS is the following:
This analysis can be further expanded by observing how different participants perceive and prioritize the factors contributing the motion sickness. Figure 15 shows the answers that have been transformed into distribution shapes to better visualize them. Lower rankings (closer to 1) indicate that the participants consider those factors to be the primary contributors to their motion sickness, while higher rankings (closer to 4) suggest that the factors have a lower perceived impact.
As an attempt to clarify the obtained rankings, it can be beneficial to observe other answers of the questionnaire, such as the ranking of different driving factors and their effects. Figure 15 shows how participants perceive and prioritize the factors contributing the motion sickness. The answers have been transformed into distribution shapes to better visualize them. Lower rankings (closer to 1) indicate that the participants consider those factors to be the primary contributors to their motion sickness, while higher rankings (closer to 4) suggest that the factors have a lower perceived impact.
Table 4
Expected motion sickness score obtained via ranking of questionnaires of all passengers
Zone 0
Zone 1
Zone 2
Zone 3
Zone 4
Zone 5
EMSS
3.625
3.875
4.750
3.875
3.75
1.125
Fig. 15
Distribution of the effect of the driving factors by the perceived influence on the generated motion sickness
×
The results are quite unanimous: Steering Wheel and Verticality are the dominant factors in the lack of comfort. Verticality is considered the worst generating factor, Steering Wheel ranks as the second worst factor, with some also noting it as the worst. On the contrary, the Throttle and the Brakes are consistently the most comfortable factors, ranking third and fourth respectively.
Besides, on the last question, several passengers noted their opinions about the simulator. Many of those are related to the sensations of each of the axis of the simulator and its comparison to real driving. These personal assessments can be summarized in the following points:
Longitudinal accelerations are not noticeably perceived in the simulator. It has been indicated that sharp accelerations or brakes as barely noticeable. This is related to the fact that the simulator itself is mechanically limited in the longitudinal axis.
In order to emulate longitudinal movements, the platform inclinates, which implies a much greater sensation of Verticality than the road slope.
Some passengers have noted that the sensations are smoother than real life driving attending to lateral assessment of the vehicle, which can be explained by observing the actuators in Fig. 11, where it can be seen that the Steering Wheel signal is attenuated by the platform.
8 Objective and Subjective Evaluation Comparison
When comparing the objective results observed in Section 6 and the results of the questionnaires that have been summarized in Section 7, it seems that a direct correlation is missing. Zones 0, 1 and 2 have the highest motion sickness criteria value (MSDVxy), but it is not reflected in the questionnaires. Zones 4 and 5 have medium motion sickness values, whereas attending to the questionnaires, Zone 4 is considered as comfortable yet Zone 5 is the worst by far. Moreover, even though Zone 3 the most comfortable zone if objective criteria is considered, this result cannot be outlined by questionnaire answers.
However, these discrepancies can be directly explained by using the limitations mentioned in Section 7. As stated by the passengers, considering both the final comments on the questionnaire and Fig. 15, the overall longitudinal effect is attenuated by the simulator itself and thus, the lateral calculation becomes the determinant factor in the answers of the questionnaire. According to the objective results shown in Table 2 and Fig. 13 pertaining the lateral effect, it can be concluded that the subjective valoration (General Score) assigned by the passengers is directly related to the MSDVy. The slight overscoring of Zone 0 may be related to the fact that it is the simulated first zone and there is no reference point.
On the other hand, the fact that passengers similarly evaluate all zones, makes it considerably difficult to achieve an proper process of ordering, which can be observed in Table 3, where the distribution of the answers is in some cases polarizing, or in Fig. 14 with the big standard deviations of the General Scores.
This is aggravated when taking into account that there is not a direct reference for the ranking itself, its order nor the distances between zones. As consequence, the Expected Motion Sickness Score is very similar for many zones. A exhaustive analysis of zone comparison in this area would need a much longer and complex experimental campaign.
9 Conclusions
In conclusion, the work validates the simulator’s ability to faithfully recreate real-world driving. The experimental campaign and dataset creation are highly successful, providing a comprehensive collection of valuable data comprising both objective and subjective information, which paves the way for further advancements in vehicle development and comfort assessment. Moreover, the simulator proved to be a reliable tool, as its results have been validated through careful analysis.
However, while having positive results, the unification of subjective and objective comfort evaluation presents some shortcomings when this simulator is used. It has been deduced that despite the good synchronicity with the environment and the adequate reproduction of signals, this simulator fails to replicate certain sensations that could be important in the feeling of motion sickness.
In this simulator, the inclination of the platform is the result of the slope and the motion cueing to simulate longitudinal acceleration. This approximation is adequate for many uses and applications, but seems critical in the evaluation of motion sickness. This renders verticality the main perceived factor of motion sickness generation, erasing the longitudinal factors and their effect (brakes and throttle). On the other hand, even though the lateral effect is more realistic, the saturation of certain signals can have implications in certain driving scenarios.
It is important to note that simulator sickness can have additional causes, most notably visual cues from the simulated environment. Future work in this area would benefit from adding analysis of visually-induced simulator sickness alongside vibration based factors to provide a more comprehensive understanding of overall simulator comfort and fidelity.
The work is the starting point for an complete experimental campaign. Future works will be focused on the expansion of the already successful dataset, by having different naturalistic driving styles in different kinds of road settings. Consequently, it is imperative to take into account the identified discrepancies when utilizing the simulator for these objectives. Many of the limitations could be addressed in the following works by utilizing a simulator designed with more realistic longitudinal accelerations, such as DiM50 and its related simulators. To explore signal saturation, future research could focus on conducting different maneuvers directly within the simulator, repeating the process for different subjects to study the reproducibility.
Declarations
Conflict of Interest
All authors certify that there are no conflicts of interest.
Ethics Approval and Consent to Participate
All authors certify that consent to participate was obtained from every subject that has taken part in the work.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
ATZlectronics worldwide is up-to-speed on new trends and developments in automotive electronics on a scientific level with a high depth of information.
Order your 30-days-trial for free and without any commitment.
Die Fachzeitschrift ATZelektronik bietet für Entwickler und Entscheider in der Automobil- und Zulieferindustrie qualitativ hochwertige und fundierte Informationen aus dem gesamten Spektrum der Pkw- und Nutzfahrzeug-Elektronik.
Lassen Sie sich jetzt unverbindlich 2 kostenlose Ausgabe zusenden.
Die im Laufe eines Jahres in der „adhäsion“ veröffentlichten Marktübersichten helfen Anwendern verschiedenster Branchen, sich einen gezielten Überblick über Lieferantenangebote zu verschaffen.